Computer Music Using SuperCollider and Logic Pro

Computer Music With Examples In SuperCollider (3.6.5) and Logic Pro (9) David Michael Cottle

Thursday, July 11, 2013

Table of Contents Table of Contents

2!

SuperCollider Examples

8!

Table of Figures

10!

1.! The Introduction •! OS X (10.8) •! Logic Studio •! SuperCollider •! Max/MSP •! Amadeus Pro •! Target audience •! Logic Pro Certification •! Terms of Use •! Content, Format, Conventions •! Legal Stuff

11! 11! 11! 11! 11! 12! 12! 12! 12! 13!

2.! OS 10.x •! The Finder •! A Finder Window •! Power •! System Preferences •! Documents, Folders, and Info •! Storage Devices •! Server Space •! Sessions and Assets •! OS X Survival •! Exercises

14! 14! 15! 16! 16! 17! 17! 18! 18! 20!

3.! Logic Pro Quickstart •! New Project •! Video Overdub •! I/O Buffer Size •! Comping •! Apple Loops •! Looping •! Mix •! Automation •! Audio, Instrument, and MIDI •! Inserts •! Bounce •! Exercises

21! 22! 22! 23! 24! 25! 25! 25! 26! 26! 27! 27!

4.! The Nature of Sound •! Vibration •! Patterns, Periods, and Noise •! Frequency, Duration, and Tempo •! Amplitude •! Shape •! EQ •! Phase •! Velocity and Wave Length •! Expression

28! 28! 29! 31! 31! 32! 33! 33! 33!

2

11!

14!

21!

28!

•! Exercises

35!

5.! Music Languages •! Binary, Hex, Decimal •! Frequency •! Cents •! MIDI •! XML •! Pitch Class, Intervals, Notation •! PDF and Screen Shots •! AIFF, Wave, MP3, AAC •! Sample Rate and Bit Depth •! File Size •! Exercises

37! 38! 39! 39! 40! 40! 41! 41! 42! 43! 45!

6.! Mics, Cables, Placement •! Connectors and Cables •! Microphone Types •! Microphone, Cable Care •! Mic Placement •! Proximity •! Phase •! Axis •! Exercises

47! 48! 48! 49! 50! 51! 51! 52!

7.! Recording Techniques •! Stereo/Mono •! Overdubbing •! Big Mono •! Patterns, Stereo Pairs •! Sample Rate •! Noise •! Balanced Cables •! Distortion •! Optimum Levels •! Aliasing •! Edits •! Optimum Engineering •! Exercises

53! 53! 53! 54! 55! 56! 57! 58! 58! 59! 60! 61! 63!

8.! The Way Music Works •! Meter •! Harmony and Melody •! Music is Math •! Quality of Music •! Set Theory •! Consonance and Dissonance •! Tuning •! Pythagorean Comma •! Tuning by Ear •! Music is •! Exercises

65! 67! 69! 70! 71! 72! 74! 74! 75! 76! 77!

9.! Mix: Style, Dimension •! Music, and Machines •! Mix Styles •! Mix Dimensions and Elements

79! 79! 80! 3

37!

47!

53!

65!

79!

•! •! •! •! •! •! •! •! •! •! •! •!

81! 84! 84! 84! 85! 87! 88! 88! 89! 89! 89! 90!

EQ and Filtering (Height) Boost, Sweep, Cut Noise Reduction Algorithms Panning (width) Compression Reverbs and Delays Doubling Delay Synced Delays Asynchronous Delays Chorus Serial Routing Exercises

10.! Auxiliaries •! Signal Chain •! Groups •! Busses and Auxiliary Routing •! Sub-masters •! Parallel Effects •! Freezing and Distributed Processing •! Auxiliaries for Monitors •! Exercises

91! 91! 92! 92! 94! 96! 96! 98!

11.! Adding Sound Effects •! London Effects •! Media/Library, Shortcuts, User Settings •! Delay Designer •! Evoc 20 Filterbank •! Evoc 20 TrackOscillator •! Spectral Gate •! Flanger, Modulation Delay, Phaser •! Ring Shifter •! Side Chains •! EVOC 20 PolySynth •! Spatialization •! Automation •! Latch •! Exercises

100! 100! 100! 101! 101! 101! 101! 101! 101! 102! 102! 103! 104! 105!

12.! Review, Examples •! Review Session •! Electro-acoustic mix •! Examples and Analyses •! Exercises

106! 109! 109! 112!

13.! SC 3 Quick Start •! SC Intro •! Hello World •! Error messages •! Elements of Code •! Enclosures (parentheses, braces, brackets) •! Arguments •! Object.Message •! Exercises

113! 114! 116! 116! 117! 119! 119! 121!

14.! Concrète Procedures •! Concrète

123! 4

91!

100!

106!

113!

123!

•! •! •! •! •! •! •! •! •! •! •! •! •!

124! 125! 125! 125! 125! 126! 127! 129! 129! 131! 132! 133! 134!

Looping Pitch and Time Shift Reversal Abrupt Edits EXS24 UltraBeat Concrète using SC Envelopes Variables, Frequency and Duration User Defined Arguments Multi-Channel Expansion Experiment Exercises

15.! SC to Logic •! KeyState.kr •! Recording to Disk •! SC to Logic using Soundflower •! Real-Time Analysis •! Come Out •! Granular Synthesis •! Sequenced Granular Synthesis •! Exercises

135! 136! 137! 139! 139! 142! 143! 144!

16.! Wave Generators •! Frequency, Amplitude, Timbre •! Voltage Control •! Scale and Offset •! Precedence •! Keyword Assignment •! Vibrato •! Unipolar/Bipolar •! Audio UGen, Control UGen (Carrier, Modulator) •! Theremin •! Exercises

145! 145! 146! 148! 149! 150! 152! 153! 154! 155!

17.! Sources of Control •! Periodic Waves •! VCA •! Aperiodic Triggers, Complex Envelopes •! Dseq •! Array.fill •! Function Arguments •! VCO, VCF, VCA in Logic Instruments •! Control Sources in logic •! Synchronized Controls •! Exercises

156! 157! 158! 160! 161! 162! 163! 164! 165! 169!

18.! SynthDef and GUIs •! SynthDef •! SynthDef arguments •! GUI slider •! Iteration •! Synth Browser •! Startup File •! Exercises

170! 171! 173! 178! 180! 183! 184!

5

135!

145!

156!

170!

19.! Filtering, SoundIn •! RLPF •! Logic ES1, ES2 •! Complex Resonance •! Klank •! Bells •! Debugging •! Monitoring Data •! Exercises

185! 186! 186! 187! 188! 190! 192! 194!

20.! Phase Modulation •! Frequency Modulation •! Phase Modulation •! EFM1 •! Logic Pro Modulators, Ringshifter •! Latch (Sample and Hold) •! Control Source Review •! Exercises

195! 198! 201! 201! 201! 207! 211!

21.! Physical Models •! Chaos •! Karplus-Strong Pluck Instrument •! Pluck •! Physical Model Interface •! Physical Model Results •! Sculpture (Modeling Synth) •! Jitter •! MIDI Input •! MIDI In •! Exercises

212! 214! 215! 216! 216! 217! 219! 220! 224! 225!

22.! Additive Synthes.. •! Harmonic Series •! Additive Synthesis •! Syntax Shortcuts •! Inharmonic Spectra •! Noise •! White and Pink Noise •! Additive Examples •! Synthesis Review •! Exercises

226! 226! 231! 232! 235! 236! 237! 241! 242!

23.! K and A Busses •! Audio and Control Busses •! Auxiliary Sends. •! Control Busses •! Exercises

243! 244! 247! 250!

24.! On Random Numbers •! Computers Cannot be Random •! Random Number Generator •! Random Seed •! Random Bias •! Aleatory and Generative Composition •! MIDI Out •! Random Walk •! Total Randomization

251! 252! 253! 258! 259! 260! 261! 263! 6

185!

195!

212!

226!

243!

251!

•! Conditional Control •! Structural Control •! Exercises

265! 267! 269!

25.! Computer Music •! Speed •! Accuracy •! Methodical •! Obedient and Obtuse •! Escaping Human Bias •! Autonomy •! Absolute or Proportional •! Pitch •! Duration and Next Event •! Next Event •! Non-Sequential Events •! Rhythmic Inversion •! Pbind •! Legato

270! 271! 274! 276! 276! 278! 279! 279! 280! 281! 281! 281! 282! 284! 285! 290!

Prand, Pseries, Pseq •! Exercises

26.! Working With Data •! ASCII •! 12-Tone Matrix •! Kinderstuck Variations •! Working With Files •! Path Names •! Set Theory •! Cyphers •! Exercises

291! 292! 296! 297! 297! 299! 301! 305!

27.! Maps, Masks, and Markovs •! Music from Non-Musical Ideas •! Mapping •! Markov Process •! Control Source Review •! Music of the Spheres •! Exercises

306! 306! 312! 318! 318! 320!

7

270!

291!

306!

SuperCollider Examples 1.1::! Code Headings __________________ 13! 7.1::! Balanced Lines ________________ 57! 7.2::! Balanced Lines With Noise ________ 58! 7.3::! Balanced Lines Re-Inverted _______ 58! 13.1::! Hello World __________________ 114! 13.2::! Hello Hal ____________________ 114! 13.3::! Music Math __________________ 114! 13.4::! First Patch __________________ 115! 13.5::! Second Patch ________________ 115! 13.6::! Balanced Formatting ___________ 118! 13.7::! Arguments __________________ 119! 13.8::! SinOsc defaults _______________ 120! 13.9::! Experimenting with a patch ______ 121! 13.10::! Rising Sines __________________ 122! 14.1::! Loading an Audio Buffer _________ 127! 14.2::! Pitch and Direction ____________ 128! 14.3::! Loop, Playback Position, Triggers___ 128! 14.4::! Envelopes ___________________ 129! 14.5::! Variables ___________________ 130! 14.6::! Arguments __________________ 131! 14.7::! 60 Second Phase ______________ 132! 14.8::! Generative Choices ____________ 133! 15.1::! Keyboard Triggers _____________ 135! 15.2::! Options for Record ____________ 137! 15.3::! SC to Soundflower to Logic ______ 138! 15.4::! Come Out Variations ___________ 139! 15.5::! Granular Synthesis ____________ 142! 15.6::! Grains with Pan and Speed _______ 143! 16.1::! VCO SinOsc __________________ 146! 16.2::! VCO, VCF, VCA using Blip ________ 146! 16.3::! Scale and Offset Examples _______ 147! 16.4::! Scale and Offset Patch _________ 148! 16.5::! Keywords ___________________ 149! 16.6::! First Patch using Keywords _______ 149! 16.7::! SinOsc poll __________________ 150! 16.8::! VCO Vibrato _________________ 151! 16.9::! Rate, Scale, and Offset Sliders ____ 151! 16.10::! Vibrato Envelope ______________ 152! 16.11::! Theremin ___________________ 154! 17.1::! Wave Shapes _________________ 156! 17.2::! VCA _______________________ 157! 17.3::! VCA With LFNoise and Line ______ 157! 17.4::! VCO, VCA With LFNoise And Line __ 157! 17.5::! Aperiodic Triggers _____________ 158! 17.6::! VCA and VCF Envelope __________ 159! 17.7::! Dseq _______________________ 160! 17.8::! Filling Arrays ________________ 161! 17.9::! Array.fill counter _____________ 162!

17.10::! Complex Demand Sequence _______ 163! 17.11::! Synched LFO ________________ 165! 17.12::! Mix.fill _____________________ 166! 18.1::! SynthDef ___________________ 170! 18.2::! SynthDef arguments ___________ 171! 18.1::! Ad Hoc Arguments _____________ 172! 18.2::! Single Slider _________________ 173! 18.3::! GUI Button, Parent Window ______ 173! 18.4::! Sines and Theremin with Interface _ 174! 18.5::! Synth with GUI: 2666 __________ 176! 18.6::! No GUI: 138 Characters _________ 178! 18.7::! Array, Do, Counters ____________ 179! 18.8::! Array Iteration _______________ 179! 18.9::! Sliders Using Do() _____________ 180! 18.10::! SynthDef.load.play ____________ 180! 18.11::! Add Control Specs _____________ 182! 18.12::! Modify Control Spec ___________ 183! 18.13::! Startup.rtf __________________ 183! 19.1::! SoundIn, RLPF ________________ 185! 19.2::! RLPF With Dust ______________ 186! 19.3::! Klank ______________________ 187! 19.4::! Chimes _____________________ 189! 19.5::! Isolate, Evaluate ______________ 190! 19.6::! Safe Alternatives _____________ 191! 19.7::! Post And Poll ________________ 192! 19.8::! Nested Postln ________________ 192! 19.9::! Formatting Posted Information ____ 193! 20.1::! From LFO to FM ______________ 195! 20.2::! AM Synthesis ________________ 196! 20.3::! AM, FM, Bird Song _____________ 196! 20.4::! FM Synthesis ________________ 197! 20.5::! Complex Sets of Sidebands ______ 197! 20.6::! Phase Modulator Efficiency ______ 198! 20.7::! Phase Modulation _____________ 198! 20.8::! PM Ratio ___________________ 199! 20.9::! Voltage Controlled PM Ratio _____ 200! 20.10::! PMOsc and Demand Sequence_____ 201! 20.11::! Sample and Hold Patterns _______ 202! 20.12::! Sample and Hold Improvisation ____ 203! 20.13::! Complex Wave as Source ________ 206! 20.14::! Sample and Hold, PM ___________ 207! 21.1::! Sample vs. Modeled Bell _________ 213! 21.2::! Karplus-Strong with CombL ______ 215! 21.3::! Pluck ______________________ 215! 21.4::! SynthDef and Triggers __________ 221! 21.5::! Pluck and Task _______________ 222! 21.6::! Logic to IAC to SC to Pluck ______ 224! 21.7::! Logic to IAC to SC to Logic ______ 224! 8

22.1::! 22.2::! 22.3::! 22.4::! 22.5::! 22.6::! 22.7::! 22.8::! 23.1::! 23.2::! 23.3::! 23.4::! 24.1::! 24.2::! 24.3::! 24.4::! 24.5::! 24.6::! 24.7::! 24.8::! 24.9::! 24.10::! 24.11::! 24.12::! 25.1::! 25.2::! 25.3::! 25.4::! 25.5::! 25.6::! 25.7::! 25.8::! 25.9::! 25.10::! 25.11::! 25.12::! 25.13::! 25.14::! 26.1::! 26.2::! 26.3::! 26.4::! 26.5::! 26.6::! 26.7::! 26.8::! 26.9::! 26.10::! 26.11::! 26.12::!

Additive Saw, Square __________ 227! Additive Graph ______________ 229! Additive Saw Modulation _______ 230! Syntax Shortcuts _____________ 231! Additive Inharmonic Spectrum ___ 232! Additive Noise _______________ 235! Linear/White, Exponential/Pink __ 237! Additive Synthesis Examples _____ 237! Busses as Sub Masters _________ 244! Three Simple Independent Synths _ 245! Auxiliary Send Using Out and In __ 246! Control Busses _______________ 248! Random Seed ________________ 254! Client/Server Seed ___________ 255! Pick a Card, but Show Me ______ 257! Random Bias ________________ 258! MIDI Init __________________ 260! Constant Monkeys _____________ 261! Random Walk _______________ 262! Random Walk _______________ 263! Control Using If _____________ 265! Range Control with If _________ 266! More If ___________________ 267! Structural Control ____________ 267! Fast ______________________ 270! Mixed Tuplets, Atonal Just Tuning 272! I Don't Dig You Work __________ 274! As Fast As Reasonable _________ 275! Proportional MIDI inversion _____ 280! Rhythmic Inversion ___________ 282! Environment Variables and Defaults 283! Read global library ___________ 283! Pbind with frequency function ___ 283! Pbind, Previous Instrument Def ___ 284! Pitch Model for Mutation _______ 285! Patterns ___________________ 286! Parallel Pbinds ______________ 286! Serialism ___________________ 288! String as Array ______________ 292! 12 Tone Math _______________ 293! T and E____________________ 294! Kinderstuck Segment __________ 295! Reading a File _______________ 297! Interpret __________________ 298! Open File __________________ 299! Set Versions ________________ 300! Locate Pitch Class Set _________ 300! Cypher ____________________ 302! To Be _____________________ 303! Vigenere Cipher ______________ 303!

26.13::! 27.1::! 27.2::! 27.3::! 27.4::! 27.5::! 27.6::! 27.7::! 27.8::!

9

Extracting Code ______________ 304! Pitch Map __________________ 307! Mapping Array _______________ 307! Extra-Musical Criteria, Pitch Only _ 307! Text Map, Intervals ___________ 309! Transition Table ______________ 314! Probability chart _____________ 315! Foster Markov _______________ 316! Solar System ________________ 319!

Table of Figures Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig.

1-1 Figures Look Like This ___________ 13! 2-1 Modifier Keys for Shortcuts _______ 15! 2-2 Logic Session Folder vs. File _______ 18! 3-1 Logic's Unmodified Shortcuts ______ 21! 3-2 Comp Tracks ________________ 23! 4-1 Waves of Alternating Density ____ 28! 4-2 Periodic, Complex, Aperiodic _____ 29! 4-3 Amplitude ___________________ 31! 4-4 Wave Shape and Timbre ________ 32! 4-5 Sonogram Showing Harmonics _____ 32! 4-6 EQ Parameters _______________ 32! 4-7 Phases _____________________ 33! 4-8 Four Score __________________ 34! 5-1 Nyquist Theorem ______________ 42! 5-2 Sampled Triangle Wave _________ 42! 5-3 Low Bit Depth Terracing ________ 43! 5-4 Relative File Sizes ____________ 44! 5-5 Bulgarian Download Speeds ______ 44! 6-1 Connectors __________________ 47! 6-2 Waves Out of Phase ____________ 51! 7-1 ORTF ______________________ 55! 7-2 Bit Depth Noise ______________ 56! 7-3 Balanced Lines _______________ 57! 7-4 Saturation, Clipping ___________ 58! 7-5 Aliasing ____________________ 59! 7-6 Non-Zero Crossing Edits ________ 60! 7-7 Descending, Zero-Crossing _______ 60! 8-1 Repeated Noise = Piano _________ 66! 8-2 Phase and Constructive Interference 67! 8-3 1:1 Ratio ___________________ 67! 8-4 2:1 Ratio ___________________ 68! 8-5 3:2, 4:3 Ratios _______________ 68! 8-6 Constructive, Destructive Waves __ 69! 8-7 Interval Ratios _______________ 69! 8-8 Comparative Dissonance ________ 72! 8-9 Patterns of White Key Intervals __ 73! 8-10 Movable Frets _______________ 75! 9-1 Channel EQ___________________ 81! 9-2 Clean Sine Wave From Low Roll Off 82! 9-3 Sonogram Of R, K _____________ 83! 9-4 Pan, Balance, Directional Mixer ___ 85! 9-5 Gain Control, Mono ____________ 85! 9-6 Compression and makeup gain ____ 86! 9-7 NY, LA, Classical Styles ________ 87! 9-8 Reverb ____________________ 88! 9-9 Slap Echo, Doubling ___________ 88! 9-10 Channel Strip, Serial Signal Flow _ 89! 9-11 Signal Flow _________________ 90!

Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig.

10

10-1 Sub-Masters ________________ 93! 10-2 Reverb Mix vs Dry Mix _________ 94! 10-3 Auxiliary Sends vs. Sub-masters __ 95! 10-4 Routing for Headphone Sends ____ 97! 11-1 Loop Browser _______________ 100! 11-2 Channel Strip Focus __________ 100! 11-3 Binaural Panner _____________ 103! 11-4 Automation Menu ____________ 103! 11-5 Multiple Timelines ____________ 104! 13-1 SuperCollider IDE Shortcuts _____ 113! 13-2 Error Message _______________ 116! 13-3 Matched Enclosures ___________ 117! 14-1 UltraBeat _________________ 127! 15-1 Keycode map _______________ 136! 15-2 Server GUI ________________ 137! 16-1 Scale and Offset ____________ 148! 16-2 Unipolar Output _____________ 153! 16-3 Unipolar Scaled by 5 _________ 153! 16-4 Unipolar Offset by 100, Scaled by 5153! 17-1 Wave shapes _______________ 156! 17-2 AR, ASR, ADSR, Aussie Opera ___ 158! 17-3 Logic Voltage Control Instruments 163! 17-4 Synchronous LFOs ___________ 166! 18-1 Synth Browser Windows ________ 181! 19-1 ES1, Filter _________________ 186! 19-2 Single Pole Filter Sweep _______ 186! 19-3 Complex Resonance __________ 187! 19-4 Noise Burst ________________ 188! 20-1 LFO to FM _________________ 195! 20-2 Sample and Hold (Latch) Patterns 203! 21-1 Chaotic vs. Static Excitation ____ 213! 21-2 Chaotic Pluck Waveforms ______ 215! 21-3 FM vs. Physical Model ________ 216! 21-4 EVD3, EVD6, EXS24 Waves _____ 217! 21-5 Sculpture Interface __________ 218! 21-6 Sculpture Controls ___________ 219! 21-7 Jitter ____________________ 219! 21-8 CPU Overload ______________ 223! 22-1 Harmonic Series _____________ 226! 22-2 Additive Saw Sonogram _______ 228! 22-3 Additive Sine Scope __________ 228! 22-4 Additive Sine, Random Phase ___ 228! 22-5 Additive Saw Modulation ______ 230! 22-6 1000 Sines vs. 100 PM Osc's ____ 236! 24-1 A Million Random Digits (Excerpt) 251! 24-2 When Repeated, Not Random ___ 252! 26-1 Set Theory Clock ____________ 300!

1. The Introduction Which starts from the beginning. First year tech students. This text begins with OS X, the nature of audio, different qualities of audio and best practices for recording, mix elements, styles and techniques. It then applies those techniques in Concrète and electro-acoustic projects, and synthesis techniques with examples in Logic Studio, Amadeus Pro, and SuperCollider. •

OS X (10.8)

… is invariably the choice of professional musicians. The reasons are sometimes personal and subjective, such as pride, erudition and loyalty, but many are practical; compatibility, stability, interface, hardware consistency, security, and software. Resistance is futile. Logic Studio is only available on OS X, but knowledge and techniques are transferable. Amadeus Pro and SC (as insiders call it) works on a number of platforms, but I won't address the differences. •

Logic Studio

… has come to dominate new composition at academic institutions. I believe it's the future. Other developers now follow their lead (with no external hardware requirement, comping features, on board instruments and plug-ins), or certainly will (physically modeled instruments and reverbs). •

SuperCollider

… is open source, with a dedicated cult of contributing international geniuses. It is powerful and deep, with a solid academic pedigree; the logical progeny of Music N languages, including Csound. You can do things right away, but will spend years, maybe even a lifetime mining the possibilities. But ultimately, it's the sounds. Twenty years ago I launched SC 1 and hit command-r and was immediately hooked by the rich textures. You can do everything, including that new, previously unknown method of synthesis you are going to invent. •

Max/MSP

… is the other (and probably more common) choice for electro-acoustic composition. The key difference between the two is code (lines you write and execute), vs GUI (icons you drag around and connect with lines). They both have a steep learning curve, but SC, once mastered, is more flexible and versatile. As with any user interface, including Logic Pro, you will always be limited to the choices on the GUI. I've composed with and taught these courses using all three, and SC is now my first choice for teaching and exploring new ideas. •

Amadeus Pro

… is an inexpensive, full-featured, two track digital editor. Many of the examples and figures in this text were generated in Amadeus, and we will use it in the early chapters for audio experiments. 11

•

Target audience

… is second year music tech and composition students with no previous experience with synthesis or code. Knowing music or programming might be handy, but not necessary. The topics assume some individual research and reading. Each of these programs has an online manual and help files that you will need to dig into. My goal is not to teach the applications themselves, but rather synthesis, quality audio, and computer assisted composition with step-bystep examples using one or more of these tools. The same is true for music theory topics. For example, 12-tone theory receives only a brief description, then we dive into examples using SC. Most of the SC examples, while interesting, are not performable works. I offer them as starting points, illustrations, ideas that can be expanded into full-fledged performances. While I try to represent a broad spectrum of techniques and styles, they do lean toward generative, minimalist, aleatory, algorithmic, and abstract methods. •

Logic Pro Certification

… is not a priority. Though I am T3 certified (certified to train trainers), preparing you for certification is not the goal of this text. The certification people are very picky about terms, and the exam is hard, really hard, I-failed-by-0.3%-during-my-second-certification hard. To certify you need to study the official certification text, page by page. You can assume there is a very specific question taken from every page of the text. For example, the white line with the green triangle at the top: Is that a playback bar, a scroll bar, a playhead, a scroll position, a playback cursor, an edit cursor, a select tool, a position locator, or a reflex phase induction modulator unit? You have to know. And rightly so. If you're certified you should be able to offer user support over the phone away from a computer. That means knowing where everything is and the correct terms for everything on the screen. •

Terms of Use

… reflect these goals: First and foremost I want this text to be used by as many people as possible, so as to generate feedback to improve it for my own use. I prefer it remain digital but also am opposed to DRM. You can share it with colleagues as you might lend a hard cover text. But if you use it regularly, purchase the original. I'll try to keep the cost within a graduate student's lunch budget. Benefits of buying (aside from encouraging me to continue) include being on the mailing list for notifications of updates, and TBD extras such as the answers to assignments, access to code files, corrections, more detailed break downs of patches, and links to podcasts of lectures and assignments. Examples in these chapters are designed to push the envelope. Turn down monitors before trying something unusual. I take no responsibility for damaged gear, locked up machines, corrupted systems, missed meals, incensed co-habitants or startled pets. •

Content, Format, Conventions

The text is divided into two sections. Chapters 1—12 focus on Logic Pro, the goal being a multi-track mix in a popular style. Chapters 13—29 build on this foundation with electroacoustic composition topics. The first half teaches how to capture, compile, organize, and compose elements of composition. The second digs into the sounds themselves, creating the raw materials from scratch. We split these two sections into a year course, a semester each. 12

• Section Headings are in Bold …and part of the next or previous sentence, for really no good reason. Text that you should read quietly to yourself is in the Times font with no other formatting. Bold text indicates items on the computer screen; Save As… under the File menu, Freeze in the contextual menu, or the Sample Editor button. Keys that you need to find and press on the keyboard are bracketed (to imitate a square key): [A], [ [ ] [return], [I]. They are spelled and capitalized as they appear on the key. Keyboard shortcuts are identified by the preceding modifier; ⌘R. Modifier keys are: ⌘ (command) ⇧ (shift) ⌥ (option) ⌃ (control). Italics are used to offset code in the paragraphs you read quietly—bizarre looking words or phrases such as SinOsc.ar, LFNoise0.kr(1, 0, 100)—but also normal looking words that have a particular meaning, different from their normal meaning, such as the variable myString, the argument offset, the if, play, or do message. Numbers taken from code examples are not italicized (because they have an intrinsic meaning), unless they are parameters inside a section of code. Italics are also used for emphasis or introduction of a new term: The frequency of a wave is different from its output. Changing the range of output is called scaling. Some common SC terms, that stand on their own, are not italicized, but often capitalized such as UGen, Synth Server. Fig. 1-1 Figures Look Like This

Figure numbers are tiered with the chapters (see Figure 1.1) and will float close to the text where they are explained. Images that illustrate a set of instructions (like the image on the right) are not useful out of this context, and are not labeled as figures.  Paragraphs with information that may appear on an exam are check marked. 1.1:: Code Headings "Code intended to be run in SC appears in Chalkboard Font.".post; "Code that is broken in the current version of SC will be crossed out.".speak; ["I", "leave", "it", "in", "presuming", "an", "eventual", "fix"].plot; ::

1:1 Exercises are tiered with chapters. •

Legal Stuff Apple screen shots reprinted with permission from Apple Inc. Amadeus Pro screen shots reprinted with permission from HairerSoft Inc.

13

2. OS 10.x It's Unix under the hood. Named after big cats. •

The Finder

… is an application like MS Word, Internet Explorer, or iPhoto. It launches automatically when the computer boots and it remains open and running at all times. You use it to navigate the contents of the hard drive as well as connect to and copy between other storage devices. It also has menus and tools that change superficial preferences such as appearance, keyboard function, screen backgrounds, but also essential preferences such as the input and output device for recording. •

A Finder Window

… can be opened by pressing ⌘N. (You can also double click on the hard drive icon or click on the icon in the Dock.) Several finder windows can be open at once, and often need to be, for example to drag files from one to the other. The default layout of a window has a customizable sidebar (drag items to or off of it) with frequently visited folders and locations, a customizable menu bar at the top, and a list of items in that folder. You can change how the items are listed using the View menu, or by clicking on one of the view buttons in the menu bar. I find icon view (far left) the least useful, although with preview icons, it seems better. Columns view shows the hierarchy of folders, but you can also turn on View/Show Path Bar and the path will be displayed at the bottom. I prefer list view or cover flow because they show information such as creation date and size as well as a preview. You can also have them ordered by name, date, and size by clicking on the appropriate label at the top of each column. I keep my folders ordered by date. This brings things I’m working on to the top of the list. I switch to size if I’m cleaning up my drive. Quick Look (space bar) is useful for auditioning audio files or peeking into documents. Spotlight is a powerful search engine that not only looks for file names, but also inside files. In some cases it's a little too powerful—useless, for example, if I type in "cottle." Pretty much every email I send has "cottle" in it, so I gravitate toward the find by name when looking for files, then the search options in a particular application for content. That said, I’ve found documents several years old, buried deep in an archive folder by just typing a few key words of a Webern Quote: "The further one presses." Once in spotlight, the down arrow will reveal quick looks at documents. ⌘-return will open the enclosing folder rather than the file itself. If you open a search window (File/Find, or ⌘F) you have many more options. You can find by content or name, but also modification, type, and so on. Clicking the "+" button allows multiple criteria. Holding down the option key (⌥ ) while clicking the button allows combined criteria such as "music" and "created today." You can also use keywords in both the spotlight and find fields. For example "name:cottle kind:folder " will bring up all folders with cottle in the name. (This looks like it needs a comma between cottle and kind, but you have to type it exactly as it is. The name includes cottle, the kind is folder.) 14

Another great search resource is the Help menu (available in each application). It even shows menu items. Try searching for shortcuts, covered next. •

Power

… users have become a Hollywood action movie cliché: The (mandatory) geek character sits at a computer and uses his special knowledge to execute a key plot point. A detail you’ll notice from now on is that they never use a mouse. Before the development of graphic user interfaces (GUI), computer commands were text based, and had to be typed out on a command line. Once learned, they are very efficient, but at the expense of intimidating language such as ls, grep, or -mtime +3 \\! -name '.*' -exec mv -n {} " & file_path & "/Archive/. Command lines are faster. GUI's are accessible. Even two year olds understand an elephant icon. The fact is you can have both. Learning shortcuts doesn't mean you have to give up the user interface. You can have both power and accessibility, relying on the GUI when navigating unfamiliar waters. Fig. 2-1 Modifier Keys for Shortcuts

The fastest way to learn is to notice them next to each menu item. Figure 2.1 shows the View/Arrange by submenu. Each menu item has a shortcut. To view by icon pres command-1. The Arrange By list of shortcuts is a nice summary of the symbols used for three modifier keys; control, option, and command, in that order. Other key icons are shift (⇧), caps lock (⇪), delete, return (↩), escape, and arrows. As a matter of fact, I will often pull down the menu just to learn or remind myself what the shortcut is, close the menu, then use the shortcut.  Below are the shortcuts I use most often. I suggest you try each as you read, then note the ones you like, then review and practice them so that they become second nature. - Type to find: In most dialogs and windows you can type a name to select an item. Also, tab or arrows will scroll through a list of items. When I tell people this they reflexively click in the find field. Resist the temptation, just start typing. - Single click (select the item), double click (open the item), ⌃ -click (contextual menu) - ⌘N: new finder window - ⇧⌘N: new folder in a given window - ↩ (return): rename a selected item - ⇧⌘A (D, O, U, or H): go to Applications, Desktop, Documents, Utilities, and Home - ⌘K: connect to remote device - ⌘O, W, Q: open, close windows, quit applications - ⌘↓, ↑: move into a folder, or back up a folder level - ⌘ delete: move to trash - Fn delete: forward delete 15

- ⌘space: spotlight, then ⌘↩ to show item in the Finder (rather than open the item) - ⌘ tab and ⌘`: rotate through applications or windows - ⌘1, 2, 3, 4: change view to icons, list, columns, cover flow - ⌘⇧4: snapshot of screen. Add the control key to copy the snapshot to a clipboard, press the space bar after you get the cross hairs to isolate a window. - Not really a shortcut, but related to screen shots, you can create a PDF of any file from the print dialog by choosing PDF. - Esc (cancel out of any dialog) - Finally, though these are usually disabled, I often use the "move focus to" menu and dock as well as full keyboard access (all controls). Look at System Preferences/Keyboard for a more complete list or to create your own. There are many more. The few seconds it takes to commit a command to memory pays off quickly in terms of workflow, but also slack jawed onlookers. •

System Preferences

… allow customization of the computer's look and default settings. I’ve discovered some clever settings just poking around in my spare time. For example, users usually overlook universal access. I keep sticky keys on all the time. It doesn’t change the way I normally type, and is handy for those occasional situations where you’re typing with one hand. (It also displays modifier keys on the screen during a lectures and podcasts.) Zoom is indispensable in lectures. The preference you will interact with most often is Sound for setting the input and playback device. I can’t tell you the correct device to choose because it will change depending on your project. In our studio we regularly switch between the internal speakers, the Mackie or Presonus firewire return, SoundFlower, or a variety of inputs. The recording software you use will also have a preference file that allows you to choose an input. Sometimes the software inherits the system settings, sometimes not. Here is a real life example of not using this dialog correctly: after spending hours setting up mics and interfaces, you check the levels and hit record. Everything looks great. But then it sounds awful: in mono, noisy, distant. The error could be that the input is set to the internal mic on the computer rather than the firewire mixer input. •

Documents, Folders, and Info

… are hierarchical; folders in folders. There are usually several ways to get to important locations. The Desktop, for example, though immediately available by clicking on the background, is in fact buried in the hard drive. It’s in your account folder, which is in the Users folder, which is on the Hard Disk. Folders can go pretty deep. A song in your iTunes collection will probably have a path that is eight folders deep: Hard Disk/Users/YourAccount/Music/iTunes/iTunes Music/Tom Waits/Greatest Hits/They are all great/But How About Swordfishtrombone

Files are documents containing data. These can be opened by double clicking, pressing ⌘O, ⌃-clicking and choosing open or open with, or using File/Open from inside an application. There are usually two or three, sometimes dozens of applications that can open and edit a single file or file type. If you double click, the OS will use the default application. You can use 16

File/Open with, or Always Open With (using the ⌥ key) to force it to open in something other than the default. To move or copy a file, you can drag it to a new location (after opening both locations and arranging the windows so that you see both). If you drag the item to a location on the same physical device (e.g., hard disk, flash drive, server location), it moves. If another device, it copies. Use ⌥-drag to make a copy on the same physical device. A faster method for copying is to select one or several items (in list view click and drag to the right of—not on—a list of names, or use ⇧-click to select everything between a selected item and the one you click on, or ⌘-click to add non-contiguous items) then press ⌘C to copy, navigate to the new location, then ⌘V to paste. It is often useful to inspect the attributes of a file (or folder) using File/Get info for a single item, or even more useful, File/Show Inspector (hold down the ⌥ key). This window stays open as you change your selection or add to a group of items. You can then see the size of all items, or change attributes such as the default application, the label, or permissions. Permissions are a part of the security of an operating system; who can read and write to the file, who can just read, or no access at all. Most of this takes place behind the scenes, but what is intended as a feature often interferes with the function of a multi-user lab. For example, when a student hands in a project, or if copying an item from a shared folder, the permissions may not transfer correctly. •

Storage Devices

… contain data. These include USB jump drives, CDs, internal hard drives, external hard drives, or servers. Unmount discs before you remove or disconnect them. You may damage the disc and in the case of firewire drives you could damage the internal firewire card. Unmount a disk with the eject button or by selecting it and choosing File/Eject, or by dragging it to the trash. If you are working on one machine with one hard drive, such as your laptop, these issues are moot. But in a multi-user lab, four out of ten students (uh, that's two fifths I guess) struggle with storage issues all semester. If you're one of those (you know who you are), read this next section carefully. If confident with storage management, skip to the next section. •

Server Space

… is a remote machine where you can store and access items from on or off campus. This is handy when moving between machines or between labs. But don't rely on it as your master copy. The copy you own is the one you can carry out of the classroom. Servers are for storage, not work. You shouldn't open the file directly from the server, particularly in the realm of audio, because the connection between the two machines is too slow and/or the permissions may be wrong. Working from the server will result in oblique errors. By this I mean the error dialog will not read "hey, stop working from the server." It will complain about something seemingly unrelated, such as slow disk speed or permissions.

17

•

Sessions and Assets

… are contained in a single session folder. Logic (as with many other programs) uses possibly hundreds of components or assets; audio files, of course, but also instrument samples, reverb impulses, backups, movie clips, instrument and plug-in presets, and so on. All of these are individual files in different locations. (The advantage to this scheme is that one file can be shared by unlimited sessions.) If you are working with only one machine in a studio, these assets can be spread around the machine, nearly anywhere. But if you are moving from one machine to another, as in a lab, it is important to keep all of these items together. For that reason you should think of a Logic session not as a file, but an entire folder, with sub-folders filled with assets. Figure 2.2 shows a Logic session folder. The session file (what you open to start working) is named Spaceman.logic. This project has 81 tracks, many of them stereo, instrument samples and a video. But note the size: 6.1M. That's hardly enough to hold a few Mp3 files. It's small because the session file itself contains no audio. The audio is in the Audio Files folder. Fig. 2-2 Logic Session Folder vs. File

This bundled approach is used for most applications (Mail, iPhoto, Calendar, TextEdit), but the structure is buried deep inside the application itself, chiefly to keep you from fiddling with them. You can reveal these assets using ⌃-click, Show Package Contents. Even system applications are folders with assets that you should routinely peruse and modify. Kidding! Do not under any circumstances navigate to Hard Disk/System/Library/Core Services/CoreTypes.bundle, and open package contents. Are you nuts? Instead, make a copy of TextEdit on the Desktop, ⌃-click to open its contents. Continue to dig down into Contents/Resources to open the English.lproj/InfoPlist.strings. Open this file in any text editor (e.g., ironically, TextEdit) and change the wording in dialogs. The brave and/or reckless CoreServices people could check out some of the .icns files used by the OS. Use quick look to scroll through them. These are handy when writing texts on operating systems. They could be modified or changed, but I wash my hands of any fallout. Logic Pro session folders have a similar structure, but it's visible, and much easier to screw up. Just wait and see. If you use your own personal external firewire drive you can open the sessions directly. The connection is fast enough for audio.  To review: Work locally (on the internal hard drive), but save remotely (to the server). When creating Logic sessions choose Include Assets and check all assets in the save dialog. •

OS X Survival

The good news is OS X is very stable. Even if the Finder locks up you can continue to operate other programs. Beneath the interface of icons and menus is the Unix operating system, which has been used in government and university systems, where security and stability matter, for over 35 years. (For old school command line control launch the Terminal application. Unix is more cryptic, but once mastered, more efficient.)  Even with this stability, if you get past chapter 14, you will run into problems. Here is a survival guide. 18

- Save often. Save every 15 seconds. Each day use Save As… to save versions (with a different name) so that a previous version is preserved. Saving is a knee jerk strategy whenever I'm trying something new. Saving costs you nothing. Not saving could cost you hours of work, maybe hundreds of dollars. The machine could die at any moment: power failure, power manager failure, hard drive failure, motherboard failure, heck, the instructor unplugging the wrong power chord or bumping the power strip (true story). - Backup often. My rule of thumb is four copies, four physical locations, daily; my laptop, a hard drive at home, a cloud, and the university's backup system. Paranoid? Sure. I like SafeCopy and SyncSyncSync. - Learn how to locate and read online help and manuals. - Learn effective Googling. For example, take a snapshot of an error dialog and use the exact wording (especially numbers, or cryptic messages) in a search. Thousands of people around the country have no doubt seen that dialog, tried to trouble shoot it, and may have posted a solution. Also search user's groups. - Maintain a "Tech Solutions" file where you document your own solutions. - After trying to solve the problem on your own don't hesitate to lean over and ask a classmate or the instructor, or even fire off an email or text. I realize this feels like an admission of incompetence, but you have to be fearless, in two ways: Don’t be afraid of looking like a novice (as I do, I’m sure, on many discussion groups), or be intimidated by the machine. It was, after all, put together by nerds. The majority of solutions I eventually pass on to other users aren't things I knew at the time the question was asked. They are things I learned after being asked. - If the operating system or an application is acting strangely, you should first try saving everything then re-launching the application or the entire system. - Install a system monitor similar to Menu Meters or iStat Menus. Many problems first appear as absurdly high CPU, disk, or network activity. - Try discarding the preferences. The preferences for each application are in the folder [yourusername]/Library/Preferences, and/or [HardDriveName]/Library/Preferences. - If the system or an application does not respond to any commands, follow these steps: 1) Try typing ⌘-period. This is a standard key combination for "stop." (It never works, but I always try it.) 2) Wish you had saved more often. 3) Give the computer a time out, take a coffee break, go get lunch (don't laugh, this works about 80% of the time). Maybe the solution is you being patient. 4) For example, networking confusion can cause both slow performance or the spinning beach ball of death. Try turning off wireless and disconnecting Ethernet. Networking can hypnotize applications into flakey behavior and lock the entire system up for many, many minutes, which seems like hours. Removing the connection often snaps them back to reality. 19

5) Force quit the application (⌘⌥-esc, click-hold on the Dock icon). This brings up a dialog that allows you to force any application to quit. 6) Check connections. We laugh afterward, but often an unresponsive machine is just an unplugged keyboard. 7) A moment of silence to remember the work you lost because you didn't save. 8) If it turns into a zombie, cut off its head. This is different for each design and model, but for most machines hold the start button down for 15 seconds. If nothing else, unplug it. •

Exercises 2:1 Find and launch TextEdit, Quicktime, Amadeus Pro, Safari or Firefox, and PhotoBooth using as many different methods as you can. Generate content in each (i.e., record a movie, audio, take pictures, write something, connect to a site). Take snapshots of the web page and paste it into TextEdit. Switch between the two using ⌘-Tab. Create a PDF of the web page. Save all to the Desktop. Force-quit TextEdit. Select all the files and use Get Info or Inspector to examine their sizes. Use Get Info to change the permissions and the default application for the video. 2:2 Without using the mouse: Switch to Finder, open a new Finder window, navigate to the Desktop, create a new folder, and name it. Copy all the files you created and paste them into the new folder. Move back to the Desktop, rename the folder, copy it and paste it to another external device, or a server. 2:3 Open system preferences and change the desktop image, the alert sound effect, time until the system sleeps, and input device. Also find and turn on sticky keys and one finger tap and drag (if using a track pad). Find keyboard shortcuts and create a new shortcut for Show Clipboard in the Finder.

20

3. Logic Pro Quickstart Nice project for beginners: a movie soundtrack. •

New Project

The first dialog after launch shows a set of templates. Be sure to explore these, but we will start from scratch with an Empty Project. The next dialog asks for the number and type of tracks. For now we will use 4 mono audio tracks. More can be added later. Also, the options for each track can be changed later. If your project is a large session with a 16-channel interface, check Ascending next to the Input menu. This will assign each track to successive inputs of the interface. The next thing you see is the Arrange Window. Most projects can be done with this window. Click on Mixer (lower left) to open the mixer area. If you don't see the Media area to the right, select the Media icon in the upper right corner. There are five sections to the arrange window. The inspector on the left, the arrange area with tracks and timelines, the mixer with volume sliders, the media area to the right, showing the Library, Loops, Bin, and Browser, and the transport at the bottom. Key presses and shortcuts such as the space bar or arrows have different meanings depending on which area has focus. The focused area is highlighted and can be changed with the tab key. Press the tab key a few times to see how to change focus. You can toggle the Inspector with the icon in the upper left corner or by pressing the [I] key with no modifier. In most text editors shortcuts include modifier keys because the unmodified keys are require to enter text. Since most Logic operations never involve unmodified keys (aside from changing names or numbers in text and number boxes) they are free to be used as shortcuts. This sometimes gets you into trouble. If you start typing text, thinking you're in a text box or the caps lock keyboard, but are in fact focused on one of the arrange areas, you will inadvertently turn things on and off with each key press. In particular [S] and [M] will solo or mute tracks if pressed accidentally. Figure 3.1 shows a list of unmodified shortcuts. Those I use often are in bold. Fig. 3-1 Logic's Unmodified Shortcuts

[1-9] [A] [B] [C] [D] [E] [F] [G] [H] [I] [J]

screen recall automation bounce metronome delete duplicates Event List Finder (Browse) global tracks hide view (mode) inspector swap left, right locator

[K] [L] [M] [N] [O] [P] [Q] [R] [S] [T] [U]

create marker loop folders mute score editor loop browser piano roll quantize library solo tempo list time signature

[V] [W] [X] [Y] [Z] [/] [< >]

plug-in windows sample editor mixer hyper editor track zoom cycle mode rewind, forward

The arrange area shows a timeline with regions for each track. The mixer shows all channel strips. The inspector shows information about items that are selected, but also two channel strips; the one that is selected, and the next strip in the chain (chosen by context—it's 21

pretty smart). The Media, Lists, and Notes area on the upper right show the location and disposition of assets. The bottom, of course, is the transport. In addition to the mixer, the lower section can display the Sample Editor, Piano Roll, Score, and Hyper Editor. Most of the areas can be opened in a separate window, if you prefer, using the Window menu. A session can quickly become cluttered with information. Press keys 1 through 9 to recall a new screenset. Each screenset can be arranged with a particular task or workflow in mind. Once a screenset is arranged it will retain that layout when recalled. One of my favorite projects for illustrating the features of Logic is a… •

Video Overdub

… with music and sound effects. Find a short segment of an action filled movie (may I suggest a Roadrunner cartoon?) and drag it from to the arrange window. Be sure the yellow insert region shows 1 1 1 1. Logic will ask if you want to import the movie and extract the audio. Do both. The movie should appear in a floating window and the audio should appear in a new track. You can resize the video window or close it, which then moves it to the upper left corner of the arrange window. Save the session (maybe to the desktop). Be sure to check include assets and check all assets. Turn on cycle mode by pressing the cycle icon to the right of the transport. A green bar will appear at the top of the timeline window, set to the size of the audio region that was imported from the movie. Adjust the length of the cycle region by dragging the end of the green bar. Mute the original audio by pressing [M] when that track is selected. Press the play bar to start playback. We are going to record a vocal overdub (serious or tongue-in-cheek), where you add an action commentary or dialog for the actors. For this example use the internal microphone as an input source. To check these settings, choose Logic Pro/Preferences/Audio… and set the input to internal microphone (or your interface). •

I/O Buffer Size

 … the number of sample that are sent to the hard drive in one chunk. A larger size is more efficient, but adds latency, or a delay. A small size is more work for the CPU, which can result in dropped samples (pops or clicks in the recording) but less delay. A large buffer adds delay, but is more reliable DA conversion. (Imagine delivering telegrams to your boss. Doing them one at a time allows him to read each one the minute it arrives, but more work for you. Collecting 50 of them in a box is less work for you, but your boss reads them late.) If using the internal mic you should mute the speakers or use headphones to avoid a feedback loop. At the bottom of the inspector are two channel strips. The first is the track that is currently selected (Warp Engi) in the arrange window. The 22

second track shows the next track in the signal chain, in this image, the master output. Double the name of the track at the bottom to change it.  Just above the name are the stereo (two circles) and mono (once circle). Click to toggle. M is mute, S solo and I switches the monitor to input (rather than automatically switching from input to playback according to context), and R enables that track for recording. Above that are the meters and the volume adjustment slider. Next the pan knob, automation (off in this example), grouping (covered later) and finally input source and output (I/O) for that track. On one of the empty tracks click on the R button. This arms only that track for recording. The input in this example shows 3-4 because it was taken from a stereo mic setup (covered later). Since we are using the internal microphone we only have access to one input. If not already selected, change the input to input 1. We won't use the metronome, which is on by default. To turn it off, open File/Project Settings/Metronome and uncheck Click While Recording.  Logic has two sets of preferences: Project Settings under the File menu and Preferences under the Logic Pro menu. Preferences are globally applied to all projects. Project settings apply to the current project only. This may seem redundant, but is actually a smart design. Take a deep breath; you're about to enter the world of digital audio recording. Press the record button and start recording. Don't worry if you make a mistake. Just continue to lay down multiple tracks of dialog. With cycle mode on, the armed tracks will continue to record multiple takes, layering them in a comp folder. You will see each new comp added to the track as the cycle returns to the beginning. If you don't like any of the takes and want to start over, press ⌘period. This stops playback and discards those audio files. •

Comping When finished, use the zoom controls in the lower right corner of the arrange window to zoom your takes.

You can choose complete takes by clicking on the reveal triangle in the upper right corner of the new comp tracks. But you aren't stuck with one complete take. The real power of this feature is in Quick Swipe Comping. This allows you to click and drag anywhere in any take to select just that section of that take. The program will automatically switch between takes during playback, allowing precise control of edits. You can select a single word from one track then switch back to another. Spoken dialog is sometimes difficult to comp, if takes are very different, but if you were doing a musical track, especially if using a metronome, it is easy to isolate a single correct note from 10 or so takes. Figure 3.2 shows a comped track. Fig. 3-2 Comp Tracks

There is certainly a lot more to know about recording (low latency mode, auto punch, replace, MIDI recording), but this is a start. Feel free to explore on your own. (I just figured out what the audio wave image next to the zoom buttons does.) Once you've settled on an overdub, click the reveal triangle to close all the comped tracks into a single 23

track. (You can also choose "flatten" to reduce it to a single audio file.) Review the previous section describing the Logic Pro session folder/assets, then switch to the finder, locate your folder and open it. Confirm that in addition to the logic file a folder for your movie and newly created audio files has been created. All these combine make up your session. •

Apple Loops

… are a sizeable collection of high quality, creative, varied, interesting, and well performed musical components that can be snapped together quickly and efficiently. They are invaluable when putting together quick overdubs for video or voice over. The sound effects section alone (about 5% of all loops) would be worth thousands of dollars in the analog days. The right to use any of these effects and musical loops comes with the Logic license. To access loops, click on the Media icon in the upper right corner and choose the Loops tab. Select FX, located in the top right of the loops tab. The buttons below are filters. You can click on several to narrow down your search. For example, clicking both SciFi and Processed will give you a collection of processed sci-fi effects. In the list below, click once to hear the effect, space bar to replay it, and use the up or down arrows to scroll through each effect. Note that when the Media area is in focus, space bar plays back the selected effect. If you tab to the mixer or arrange window, space bar will start playback. Once you have found an interesting effect, drag it and hover over and empty track to place it on that track or over the blank area of the arrange window (not over an existing track). This will create and place the sound effect in a new track. You can certainly collect groups of effects on a single track, but if you decided to add reverb, pan position, or volume automation it will apply to everything on that track, so it is better to group sound effects logically. As you hover over the arrange window you will notice that the movie updates to show the frame at the drop location. It will also try to pop to a grid. Drop it on the grid closest to where you want the sound. Once placed, you can zoom in further and select the newly created region for a more finetuned location. All the while the video will show the exact frame corresponding to the left side of the dragged region. Continue to add sound effects. In the loop browser, switch from sound effects to music. This section contains two or four measure excerpts of drums, piano, guitar, bass, and synth music. Likewise, you can narrow down the search using the filters. I usually start with Beats. Note the green and blue icons to the left of each loop. The blue icons represent audio loops—recordings of real sound. The green are MIDI loops that use Logic's integrated synthesizers.  You can drag MIDI loops into an audio track and it will convert it to audio. You cannot place an audio loop in a MIDI track. It won't play back. For now, drag all loops to an empty space in the arrange area. This will create a new track with all the plug-ins and settings for that loop. Find both blue and green loops to use. Each loop includes columns showing the original tempo and key (where applicable), number of beats, match, and a favorite button. This allows you to choose loops that match the tempo of the original session, but also reveals how flexible this system is. If the original tempo 24

or pitch does not match the one you are currently using, it will adjust the loop to match, not only while browsing, but also after you have dragged it into the session. With MIDI tracks this is a simple adjustment to the timing. With actual audio it is done through sophisticated algorithms developed in the past decade. Previously it was possible to change the pitch and the speed, but not just one without changing the other. As amazing as this is (and yes, it's amazing) it's not infallible. If you try to match something that differs too far, artifacts will appear in the playback. Loops, of course, are designed for… •

Looping

To extend a loop, hover the mouse over the upper right-hand side of the region. The cursor will change into a loop icon. Click on the edge of the region and drag as far as you need the loop to play. If you know exactly how far, press ⌘R and type in the number of repetitions needed. Finally, in the inspector you can click the loop button, or press [L]. This will loop it indefinitely, or until it encounters another region. Next, you can browse for a guitar, piano, or bass loop to match the drum selection. Once located, drag it to a new track and loop it appropriately. Do the same with another backup instrument, but this time start playback (with cycle turned on) before clicking on the loop in the browser. This will playback the auditioning loop along with what you've gathered together already. Remember that space bar applies to the focused window. If the loop window is focused, it will start and stop the loop only. Continue this process until you've cobbled together an interesting background track to the action in the movie. In a later chapter we'll cover the elements and styles of creating a popular… •

Mix

This section will serve as a quick overview. There are a number of approaches to the layout of a mix. If you have a lot of tracks, you may want to close the mix section of the arrange window (press [X] to toggle on and off). You can still adjust mix components on the currently selected track using the inspector. Alternately, you can open a second window showing just the mixer (⌘-2). Finally, you can use screen sets to switch between a mix screen and an arrange screen. Play your project and explore the following components: Move the volume sliders up and down to adjust the relative volume of each track. Adjust the master volume if the overall sound is too low or too high. It's ok if you peak (the top of the level meter will show red) once in a while, but it should not be in the red the majority of the time. Use the pan knob to move the sound to the left or right of the virtual space. Use mute and/or solo to isolate a single track or tracks. Solo or mute to group similar items, to adjust them in isolation. For example, you might want to balance the drums, bass, and piano (volume and pan position) then add the guitar and adjust it. •

Automation

… allows you to adjust each of these (and pretty much any adjustable parameter of any instrument or plug-in) automatically, as if you had a handful of assistant engineers, each one managing a single track (which is how we used to do it). Press [A] to switch the arrange window to automation. Each track will show a timeline that can be used to change the displayed 25

parameter automatically. The track defaults to volume. Click on the volume pull down menu to select another item to automate.  To add a new breakpoint, click on the timeline. Note that you can create and locate a breakpoint by clicking above or below the existing timeline. Use shift drag to select a portion of the timeline, then click and drag the line up or down. This will place two breakpoints at each end of the selection, allowing you to adjust the entire segment. You can also copy and paste or drag automation along the timeline once selected. After adding automation, select playback and the volume and pan knobs will move on their own. Select a track that has automation and note the green Read button. This means that the track has automation and it is engaged. Change this button to Latch, start playback, and adjust any item in real time. This will record your movements precisely. There are three types of tracks in a session… •

Audio, Instrument, and MIDI

 Audio regions are colored blue, and contain real audio recorded in the real world. The next are instrument tracks. These contain MIDI data (essentially just note on and off commands) that are then realized as sound by one of Logic's built in synthesizers. The last are also MIDI, but intended to be played on external gear. Select an instrument track (any of the green regions). Click on the Piano Roll button at the bottom of the arrange window. This shows the MIDI events in a piano roll layout. You can click on an event (for example, a sound effect) and move it up or down, left or right (with the video frames reflecting the position for fine adjustment) or crop the beginning or end. Locate the I/O section of a track. In audio tracks the first button is used to select an input source from the interface. On instrument tracks it allows you to select a software instrument (which is a type of input). The current instrument is indicated in blue. This is the synthesizer that is used to create the sounds from the MIDI data. To change it, click and hold on the input. To open the synthesizer window, double click. Instruments are a type of plug-in; a separate module that can be added to the Logic interface. In addition to instruments, which generate sounds, there are processing plug-ins, or … •

Inserts

When you pick a loop, it not only loads the MIDI information that triggers events, but also the instrument and plug-ins required to reproduce that particular character of sound. Plug-ins are placed in inserts at the top of the channel. Signal flows from the top insert to the bottom. The top menu (above the EQ graph) selects a channel insert, which is a group of preset plug-ins (2-Step in this example). The next window shows a miniature diagram of the channel EQ. This example has a compressor, a high-pass filter, distortion (disabled), and a channel EQ. Disable a plug-in with ⌥-click. Move them with ⌘-drag, duplicate to another channel with ⌥-⌘ drag. Double click to open the preset editor. We will spend much more time on plugins in later chapters, but don't hesitate to explore now.

26

•

Bounce

…(under the File menu) commits the entire arrangement to a two channel (mono, or 5.1) file. To export the movie with the new audio, choose File/Export Audio to Movie…. PCM will save to an aiff audio file (16 bit, 44.1k sampling rate is standard for CDs), Mp3 or M4A for compressed audio, and CD to burn a CD. • Exercises 3:1 Continue with the project from this chapter and control click on any track and check Freeze. Click on of the newly created freeze buttons and press play. On an instrument track, click the I/O input and choose the EXS24. Use the Media/Library to locate HeadBanger Kit and record some drums. Press the CapsLock to open the caps lock keyboard and record some drums. With each of these steps examine the session folder where all assets of this session reside. The point of this exercise is not the content of the session (or indeed understanding how a session works), but rather the assets in a session folder. Note what new folders and files are created and how they change the size of the folder. Save the entire session folder to your documents folder on the server. Open the folder from the server and click each reveal triangle to show all assets; the tracks, the folders, the session files, and their contents. Open and run the session file from the server (as I've said not to) to see if you can produce the "disk too slow error." Delete all the assets from the session folder on the Desktop and empty the trash. Launch the session to produce the assets missing error. 3:2 Create a list of 20 sounds that you have never heard in real life, in person, only in recordings or movies. Place an asterisk by those you suspect are different in real life. For example, I've never heard whales sing or a fog horn except in movies and recordings. 3:3 Use Quicktime to create a short segment, with no dialog, of a favorite movie. Import it into Logic, and remove the audio track. Listen to the original sound track, taking note of music and sound effects. Create your own audio track trying to imitate the original mood of the film. With the same excerpt, create an audio track that changes the mood drastically: if intense, try comical, if moody try cheerful. 3:4 Open Logic Pro/Preferences/Key Commands… and explore each section. Choose Options/Copy Key Commands to Clipboard and paste into a text document. Delete those you don't think you would use. 3:5 Open one of the demo sessions that came with the Logic Pro installation. Engage cycle mode, hit playback, and spend 20 minutes exploring: double click on plug-ins, fiddle with the knobs and explore the presets.

27

4. The Nature of Sound Of any material: Change in density. •

Vibration

…of any material is the alternating compression and decompression of atoms; like ripples on a pond, but spherical, in all directions. In digital audio editors this motion is represented as a graph with peaks and valleys. The peaks represent greater density, the valleys lesser density. When the graph of a sound is realized (for example by playing through a speaker) the forward motion of the speaker creates the peaks and the backward motion creates valleys. The first image in Figure 4.1 illustrates smooth changes in air density, dark being more dense, light less. Below it is an audio graph of that wave would be drawn in a digital editor. Fig. 4-1 Waves of Alternating Density

•

Patterns, Periods, and Noise

… make up the character of the sounds we hear. The example above shows a sine wave. It is periodic because the pattern repeats, or goes through cycles. Nature is filled with examples of periodicity; tides, moon phases, days and nights… music of the spheres, so to speak. Aperiodic waves have no apparent pattern—random motion. Few natural sounds are either perfectly periodic, or absolutely random. Rather, there is a continuum of varying degrees of periodicity. Figure 4.2 shows audio graphs of three sound waves with different degrees of periodicity. The first example is clearly periodic, but has some variation. The second looks fairly random, but if you focus on the peaks you see a pattern. There are four peaks that make about three and a half periods. The last has no pattern, and would sound like noise, or static. Our world is filled with patterns. The lines and shapes we see out any window are interpreted as objects such as trees, grass, and people. We decipher sounds in much the same way; searching for patterns. We match the pattern in the waves with real phenomena such as a violin, airplane, rushing water, or a particular breed of bird. If our eyes were as experienced with wave forms as our ears are with sound, we would recognize the first image in 4.2 as an electric piano. Periodic waves have pitch concomitant to the clarity of the pattern. The clearer the pattern, the clearer the character of the pitch. The absence of pattern, to the same degree, approaches noise. I say approach because pure noise, like infinity, is a theoretical abstract, subject to our limits of perception. More on that later.

28

Fig. 4-2 Periodic, Complex, Aperiodic

Each wave can be described and measured using four criteria. These are all important for understanding synthesis and audio engineering. The first is… •

Frequency, Duration, and Tempo

The musical equivalent to frequency is pitch and tempo. A real life example of frequency would be the number of times the reed on a saxophone flaps up and down, or the number of times the string on a guitar sways back and forth when plucked. Slow frequencies are rhythms or tempos.  We measure frequency in both absolute and relative terms. The absolute scale is Hertz (Hz), or cycles per second and bpm (beats per minute) for tempi. The lowest C on the piano is 32.7 Hz, which is close to a frog croak. The next C above that is 65.4 Hz–which is the range of a tuba blat. The next C is 130.8 Hz–that guy who sings bass on Lawrence Welk (Google it). 261.6 Hz (also known as middle C), is the normal speaking range for men. Did you notice that I'm doubling the frequency each time? No? Well go back and check. 523.25 Hz, the next C, is the edge of a good tenor (great title for a short story), but also the normal speaking and singing range for women and the late great Michael Jackson. 1046.5 Hz is the range at which Viking women (or sopranos singing Viking operas) can, using only their voices, sterilize the croaking frogs. There are two Cs above that, covering the highest keys on the piano (2093 Hz to 4186 Hz). This is the range regularly occupied by piccolos and toy pianos. Humans can hear frequencies two Cs higher than that; up to 22,000 Hz. We don't think of this range as pitch, but rather the "brightness" or the edge in sound, such as fingernails on a blackboard or little sister.

29

 The low range of hearing capacity requires a qualification. We can hear sharp edged waves at frequencies as low as one pulse every minute, even one every hour or one a year. We can also feel smoother waves at frequencies as low as 1 per minute. But the point where the pulses begin to blend into a single homogenous pitch is what most texts consider the low range of hearing. So while there is theoretically no lower limit to audible frequency, pitch begins at about 20 Hz.  Periodic pulses lower than 25 Hz are heard as rhythm or tempo. Musicians measure tempo in absolute beats per minute or use tempo markings: Presto — very fast (170 bpm), Allegro — fast and bright or "march tempo" (120 bpm), Moderato — moderately (100 bpm), Adagio — slow and stately (literally, "at ease") (66 bpm), Lento — very slow (40 bpm), etc. By absolute we mean that C will always be 261.6 Hz. Absolute measures of frequency are useful when using a metronome, EQ or filtering. Music (as in melodies and harmonies) is not absolute, but relative. It is defined and described by the relationship between two frequencies. (As an example, a foot, an inch, one mile are all absolute distances. They will always be the same. Relative distances are twice as far, half as long, 1.5 times as tall.) The units for the relationship between pitches are called intervals; a fifth, an octave, a second, a third. We interpret two frequencies (either sounding sequentially or at the same time) as music if they are related mathematically. Since these measures are relative, it is impossible to say how many Hertz there are, for example, in an octave, no more than you could ask how many feet there are in 'half as tall.' The distance in Hertz will change depending on your starting point. An interval, therefore, is not a specific difference in frequency, but rather calculated in relation to a given frequency; 2x higher, 1.5x higher. An octave (2:1 ratio) above 440 is 880 (2 x 440); a difference of 440 Hz. But an octave above 1200 Hz is 2400 Hz, a difference of 1200 Hz. More on ratios later.  Each period of a pitch and each pulse of a tempo has a duration. For example, a single period of a 100 Hz sine wave is .01 seconds (100 periods in a second, each 1/100th of a second). Each beat of a tempo at 120 bpm will be .5 seconds (120 in 60 seconds, each is 1/2 second). In electronic music you need to be able to convert back and forth between the two. To calculate the duration of a single cycle or beat, express it as a fraction then use the reciprocal: 120 beats in 60 seconds is 120/60. Each beat then is the reciprocal—60/120, or 1/2. To calculate 100 bpm (100/60) use 60/100, or 3/5ths of a second. Frequency is even easier because we use a single second rather than 60; 440 Hz is 440 in one second (440/1), so each cycle of a wave is 1/440th of a second.  If you like formulas, given duration, frequency, and tempo: duration = 1/frequency, frequency = 1/duration, tempo = 60/duration, duration = 60/tempo. Two examples of how we might use these calculations: Delays (Chapter 9) or echoes work well when they match a tempo, however, delays are usually expressed in milliseconds, while tempo is beats per minute. To enter the correct value, you need to know, given a tempo, the duration of each beat. Similarly, we may create sounds using a very fast delay. The delay is expressed in milliseconds, but the result will be a pitch. To do the correct pitch, you need to know which millisecond delay will result in which pitch. (A delay time of 0.0022727 will create a pitch at 440 Hz.)

30

•

Amplitude

… is the height of the peaks and depth of the valleys in a wave. The musical equivalent to amplitude, is volume. A physical example of amplitude is the distance a speaker cone travels forward and back. We increase amplitude by driving the vibrating body harder with more energy. Example 4.3 shows a triangle wave at three different amplitudes. Fig. 4-3 Amplitude

 We measure sound in dB, or decibels. Most people speak of volume in absolute terms, as in "a jet engine produces sound levels up to 110 dB." This is incorrect. Like intervals, decibels measure relative amplitude, not absolute; nothing is 100 dB. It is however correct to say a jet engine is 110 dB louder than rustling leaves. But you could also say that late night commercials are 110 dB louder than a jet engine. A sound can be 10 dB louder than what it was before, or you can decrease amplitude by -10 dB. This is why mixer controls often have 0 about two-thirds near the top; because 0 means no change. Increasing or decreasing the volume is + or – dB.  This misunderstanding comes from the practice of setting 0 at the threshold of hearing to compare relative volumes. In this case, 0 might be a light snowfall deep in the forest, without you in it. You arrive on skinny skis, stop and hold your breath to take in 0 dB, but your heartbeat invades at 1 dB. You exhale at 10 dB, whisper to yourself "my skinny skis must think it queer" at 20 dB. Your buddy arrives and you whisper a little louder "listen how quiet" at 30 dB. After a minute of blissful silence you shush off down the trail at 50 dB, saying over your shoulder "miles to go before we sleep" at 60 dB. Your friend shouts "look out" at 75 dB because he sees the Yeti who shrieks at 90 dB (How do I know? Jonny Quest, episode 122 "Expedition to Khumbu"). You shoot your rifle (in the air, just to scare him) at 100 dB. An F-15 clips the treetops from out of the blue at only 10 dB. Yup, you read correctly; 10 dB—louder than the rifle shot.  In general terms, the difference between the quietest sound and the loudest sound you would want to hear is about 120 dB. In musical terms, one level of dynamic (e.g. mezzo forte to forte) is about 7 dB. So in practical terms "A little louder" is 5 to 7 dB. When I'm making adjustments for a client with a sensitive ear, I go 3 dB at a time. •

Shape

… is the fourth dimension of sound. The musical equivalent to shape is tone or timbre. A real-life example of timbre change is simply playing a different instrument, or modifying the instrument for a different quality: a muted trumpet or guitar, the angle of a violin bow, artificial harmonics on string instruments, playing on different parts of a cymbal, and changing the shape of your mouth during speech.  In general, a bright sound will have sharp edges to the peaks and valleys. A dull sound will have a smoother shape. There is no standard measure for timbre, except if you measure the frequency and amplitude of each harmonic. 31

Fig. 4-4 Wave Shape and Timbre

Figure 4.4 shows three sounds that are the same height, same number of cycles, but different shapes. They would be heard as the same pitch, same volume, but one is distinctly a guitar, then a saxophone, the last an electric piano. Fig. 4-5 Sonogram Showing Harmonics

In addition to shape, timbre can be analyzed and displayed on a spectrograph or sonogram. Wave shape is a product of the presence and strength of harmonics, which we will cover in later chapters. Harmonics are pure sine waves at different frequencies (usually multiples) that, when added together, meld into one pitch. Figure 4.5 shows a spectrograph of the three waves above. Each horizontal line represents a single sine wave or pure tone. Higher lines are higher frequencies, 10k near the top, 100 near the bottom. The darker lines are higher amplitudes. Manipulating these harmonics constitutes a large percentage of electro-acoustic composition. There are many sophisticated methods for this, which we will discuss later. The most common tool for adjusting timbre is … •

EQ

There are essentially three types of EQs; low shelf, high shelf, and peak, or band pass. You can combine them, as shown in the top image of Figure 4.6. This shows a low shelf at about 100 Hz, a high shelf at about 5k, and a peak boost at around 700 Hz. Fig. 4-6 EQ Parameters

!

 There are three parameters for adjusting EQ; frequency cut off is the point where increase or reduction begins (174, 500 and 2600 in the second example in Figure 4.6), dB gain or cut is the amount you increase or reduce a given set of frequencies (+14, 0, and 12.5), and Q, which is the width of the frequencies filtered (4.8, 0.71, and 0.23). The last image shows the resulting shape of the parameters above; a narrow band EQ (the sharp peak) and a broad band (the hill). A higher Q has a narrower band. The sharp peak has a 4.8 Q, the hill is 0.23. We see nothing representing the 500 Hz because it's gain is 0. Band or bandwidth describes a frequency range. Octave band refers to the frequencies within an octave. For example, 113 to 226, 1284 to 2568 are octave bands. It's important to note that octave bands grow exponentially. That means the distance between each octave is twice as large as the previous. This is sometimes confusing because the keys on a piano show octaves of equal distance. 32

•

Phase

Fig. 4-7 Phases

… is the next property of sound. Periodic waves go through a 360˚ cycle; like the phases of the moon. A wave begins at 0 degrees. One quarter through the cycle is 90˚. Half way through the cycle is 180˚ and 360˚ is one complete cycle. We don’t hear changes in the phase of a single sine wave, but phase does change how two waves interact, and the overall sound when frequencies blend. •

Velocity and Wave Length

… are dependent on temperature, altitude, and air density, but for this class you can use 1000 feet per second, 100 feet in .1 seconds, 10 feet in .01 seconds.  The length of a wave is calculated by dividing velocity or speed by frequency (v/f). Using 1000 feet per second, a wave with a frequency of 100 Hz (about a G3, where a male voice comfortably sings) is about 10 feet (1000/100). A 1000 Hz wave (about a C6, the upper range of a soprano, and many instruments) is about 1 foot (1000/1000). The pitch A 440 is about 2.2 feet (1000/440). Wavelength and speed are important for understanding echo and reverb, phase cancellation, stereo imaging, as well as proper microphone and speaker placement. What does all this science have to do with music? Most instruments provide the musician with control over these three dimensions. Changing them is what gives an instrument… •

Expression

The more control and potential change over amplitude, frequency, and timbre, the greater the expressive quality of the instrument. For example, the piano was a major advance over keyboard instruments of the period because it allowed the performer to control and change amplitude, and therefore increase its expressive quality. Keyboard synthesizers made a similar leap in musicality when they included weighted and velocity sensitive keys. Less expressive instruments—triangle, snare, woodblock—are missing control over one or more of these elements. An engineer likewise can control the expressive nature of a mix, and electro-acoustic composers draw their expression through control and manipulation of these elements. The human voice has precise control over amplitude and pitch, but it has greater control over timbre than any other instrument. It is so precise and varied that it can communicate not only musical ideas (notes, chords, melodies), but also complex and precise philosophical, religious, emotional, and patriotic ideas. This is because speech is the change in timbre from shaping upper harmonics with your mouth, tongue, and nasal passages. The character of a person's voice (e.g. your mom vs. Mr. Rodgers—a pretty even celebrity death match) is also defined by the presence and strength of upper harmonics. Interesting note: inhaling helium does not raise the pitch of a person's voice, as most people believe, rather, it sharpens the timbre, of the voice. We are very adept at interpreting harmonic structure. That's why (before caller ID) we could answer the phone and immediately say "hi mom, you sound like you have a cold. Are you watching Mr. Rodgers reruns?" 33

Figure 4.8 shows a frequency spectrum graph of a friend saying, "four score and seven years ago." If I played it for you, you would recognize it as Emily, who works in the library. The x axis is time, the y axis is frequency. The "s" sounds are in the 6400 to 12,800 range. The C's and K's are 800 to 1600. The vowel sounds are the fat ribs between 200 and 1600. Note that a vowel is made of a band of related frequencies. The "e" in seven and years has higher bands. Note the changes in upper harmonics when Emily closes her mouth for the "r" sound. Note the slight Boston accent. Fig. 4-8 Four Score

Adding, removing, or changing the amplitude of upper harmonics will change the shape of the wave and therefore the timbre. More harmonics will make a wave sharper, and will result in a brighter sound.

34

•

Exercises 4:1 Of the wave examples on the right (each is exactly 1/100th of a second), which is the loudest? softest? highest pitch? clearest pitch? brightest sound? is aperiodic? is periodic? What is the length of each of these waves: 10 Hz, 250 Hz, 440 Hz, 1000 Hz? What is the lowest pitch humans can hear? What is the highest? What is the lowest frequency humans can hear? Story question! If you clap your hand at one end of a 200' deep performance hall, how long will it take the sound to reflect off of the far wall and back? 4:2 Open Amadeus Pro. Choose Oscilloscope from the Analyze menu. Record your own voice changing amplitude but not wave shape or frequency, then, as you sing, change frequency, but not the shape or amplitude, then shape, but not frequency or amplitude. 4:3 Record ambient background sounds from some random source. Select Effects/Normalize. Select about 1 second and play it back. Can you hear a pattern? Check the loop button in the lower left hand corner and listen again. See if you can identify the meter. Change the length of the selection. What is the largest amount you can copy and still feel a rhythmic pattern? Stop playback and make the selection much smaller—.01 seconds. Turn loop off and playback that segment. Do you hear a pitch? Turn loop on and try again. Change the selection size. What is the pitch of each selection? Choose a loud section and reduce the amplitude by 7 dB. Repeat this process ⌘-T, reducing by 7 dB each time. How many dB until you can barely hear it? How would you characterize 7 dB? 4:4 In Logic, create an instrument track and assign the EXS24 by clicking on the instrument slot below the I/O icon. Click on the instrument pop up menu and select a kit patch. Add an EQ and engage the Analyze feature. Explore and locate instruments from the kit that "fit" into each octave band. Record each in sequence, bounce the recording, open in Amadeus, and generate a sonogram. 4:5 In Logic, import a song or several loops on different tracks with broad frequency spectra (for example, a drum loop, guitar, piano, etc.). Also import a tune that has full frequency representation. Add an EQ to the first track by double clicking the EQ box. Change the gain of the first band from 1 to 14 (you can click and drag the number or the peak in the graphic EQ, or double click the number and type). Change the frequency to 100. Press the solo button on one track, start playback and change the frequency. Note the effect on that instrument. Solo a different track and repeat. Next, click on the automation pull down menu and choose Latch. This will record your changes to the EQ in real time. Press play and double click the 35

frequency during playback and change it about every measure to the next octave; 100, 200, 400, 800, 1600, 3200, 6400, 12800. Leave playback running. When it repeats it will reproduce your changes. As it repeats, make changes to the EQ and note the effect. Use solo to select each track and listen to the results of the change in EQ. Choose two of the instruments and generate a chart showing a short characterization of each band (as shown below). Piano Guitar

100 boomy thump

200 presence body

400 narrow funky

800 boxy thin

1600… crisp bright

4:6 In a Logic session, find a drum loop with a broad frequency spectrum and no effects. (Drag green loops to instrument tracks and blue loops to audio tracks.) Click on an open inserts slot and add a filter plug-in. Select Logic Pro/Preferences/Key Commands… and search for Next Plug-in. Use the shortcut to explore the presets. Add three or four more plug-ins or multiple instances of the same plug-in. Option click on all insert slots to turn them all off ("bypass"). They will turn from blue to grey. (The image to the right shows three plug-ins, two of which are bypassed.) Set Automation to latch and press play. Switch each on and off using option-click. Mix and match. (Optional: modify other settings too). Bounce and open in Amadeus, then generate a sonogram.

36

5. Music Languages I use MIDI every day. Less so, binary. One of my first professors point out that 95% of programming is user interface; writing code that connects the ideas of humans with the capacity of computers. Getting a computer to generate a complex sound is a cinch: it’s just a stream of 0s and 1s. Nothing could be simpler. The problem is, we don’t speak 0s and 1s. What follows is a discussion of languages computer musicians use to pass ideas to and from computers and other musicians, beginning with… •

Binary, Hex, Decimal

Humans use base ten because we have ten fingers. When we’ve reached the end of our fingers, the person next to us holds up one finger to indicate one group of ten so we can start over with out fingers. With each total of ten, our neighbor puts up another finger. When he gets to the end of his fingers, the next guy holds up one finger. So four people holding up a 3, 4, then 5 and 2 means we counted three groups of a thousand, plus four groups of a hundred, five groups of ten, and two ones. In the absence of buddies we use an odometer. Each wheel shows the total number for that multiple. Binary works the same way, except bits have only one hand with one finger. (They are cartoon characters.) The first bit holds up one finger for the first item counted. When you add another item the next bit holds up one while the first shows no fingers. The second bits single finger means they've counted one group of twos. Appropriately, their odometers have two characters; 0 and 1. Hexadecimal also works this way, but since computers themselves have essentially 16 fingers, they use odometers with 16 symbols: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, then A, B, C, D, E, F. When this odometer reaches 9, it doesn’t return to 0, but continues on to A, then B, and so on. After F, it returns to 0. In computer music, you may never have to speak or write in binary or hex, but you may very well encounter them. Here is a comparison chart. Think of each one as an odometer. Each column is a wheel with numbers. The first has 0-9, the Hex is 0-9 then A-F, the binary only has two items: 0 and 1. Total number of items counted: • •• ••• •••• etc. ••••••••• •••••••••• ••••••••••• etc. ••••••••••••••• •••••••••••••••• •••••••••••••••••

Decimal 000 001 002 003 004 etc. 009 010 011 etc. 015 016 017

Hex 000 001 002 003 004 etc. 009 00A 00B etc. 00F 010 011

Binary 00000 00001 00010 00011 00100 etc. 01001 01010 01011 etc. 01111 10000 10001

Here is how they are totaled. Each number represents a total of that position's value. 37

Decimal odometer Totals for each: Binary odometer Totals for each: Hex odometer Totals for each:

3 100s 1 16s F 4096s

4 10s 0 8s 1 256s

1 1s 1 4s A 16s

300 + 40 + 1 = 341 1 2s 5 1s

0 16 + 4 + 2 = 22 (decimal) 1s 61440+256+160+5 = 61861

In decimal, the first column from the right represents 1s, the second 10s, 100s, etc. In binary, the first column indicates how many 1s there are. The second shows the number of 2s, then 4s, 8s, 16s, etc. In hex, the columns are 1s, 16s, 256s, then 4096s. To parse a value you multiply the number in the column by the value it represents. In binary, 10110 is (starting from the right) zero 1s, one group of 2, one of 4, zero 8s, and one group of 16 = 22. Hex is a little harder. F1A5 is 5 1s, 1 'A' 16s (10 16s), 1 256, 'F' 4096s (15 4096s) = 68161. Why use binary? Because that's how bits work. They can only be on or off. Why hex? Because it is a more efficient way to represent high numbers, and bits are combined in such a way that counting in groups of 16 uses all the available bits. Computers speak low level languages such as binary, Hex and MIDI. Humans speak high-level languages such as pitch class, interval, chords, clefs, and staves. When you have a musical idea, you need to interface with machines, networks, and people to get that idea across. The language, or format you choose will make a difference in how easily your target will understand. The more languages you know, the more choices you have. Imagine we’ve been asked to write a catchy leitmotif for a cop show in Bulgaria. How do we communicate that to the producer? If I use my laptop to write it, buried beneath the slick icon interface is this: 0011110000111110001111110011110001000010 Catchy, don’t you think? The problem is, while this is the easiest language for my computer, the producer is old school and can't sight sing binary. So let’s bring it up a level. 0000003C, 0000003E, 0000003F, 0000003C, 00000042 I’d say a handful of you could now hum this familiar theme. But, still not the producer. Up one more level would be… •

Frequency

Here it is: 261.62, 293.66, 311.13, 261.62, 370. It’s unlikely even another degreed musician would understand frequency, but in electronic/computer composition and audio engineering we deal with music on these terms all the time (e.g., EQ, filtering, as in the last chapter or arrays of inharmonic additive sine waves in future chapters). Frequency is not very useful when discussing melodies due to the exponential nature of intervals. The distance between each interval is relative; different depending on where you start. The first three notes of my theme make up a musical scale. The first two are a whole step: 261.62 then 293.66—a difference of roughly 32 Hz. But the next two are a difference of 17.47. That's not 32, but it's also not half of 32. Is it a half step or a whole step? Wouldn't a half step be 16? That’s what I thought until half way through my doctorate, if you can believe that. In fact if my musical theme were two whole steps the next frequency above 293.66 would be 36.7 Hz higher, and that’s if I’m talking to a violinist. If a pianist, it would be, no kidding, 35.9. If I started my melody on an A (440), the distance would be 59.8. Though the distance between these 38

frequencies all differ we hear them as the same distance, the same interval, the same melody. So in this context at least, frequency isn't very useful. To get around the math required to calculate intervals and talk about the mind-boggling phenomena of tuning, we use … •

Cents

It’s simple: 100 cents per half-step (i.e. one piano key). If my TV exec speaks "cents," I could write this: 6000, 6200, 6400, 6000, 6600. Don’t laugh. Take a closer look. I’d wager most of you could sing the tune. An understanding of both frequency and cents is essential for wading through later chapters. Note C3 C4 C5 C6 Distance in Hz ------65.4---------------------130.8--------------------261.6-------------------523.3 Distance in cents -----1200--------------------1200---------------------1200--------------------1200 While hex and binary are rare, and frequency is only useful in broad terms, the digital musician regularly uses … •

MIDI

 Each key on the piano is a MIDI number. The lowest note musicians use is C-1, which has the MIDI number 0. The number 12 = C0, 24 = C1, 36 = C2, 48 = C3, 60 = C4 (which is middle C on the piano, remember this one and you can quickly calculate the others), and so on. Music theorists use C4 for middle C. Unfortunately, this is not standard among synthesizers: some MIDI equipment uses C5 for MIDI pitch 60, which makes MIDI 0 a C0, and Logic uses C3 for middle C. Sorry. Each piano key is 1 number. C#4 would be 61. The pitch A-440 is above middle C so it is 69. Here is our theme in MIDI: 60, 62, 63, 60, 66. You can use decimal precision with MIDI, so 60.00, 62.00, etc. would work. It should look familiar. MIDI and cents are the same. MIDI intervals are just as easy. One piano key (a half step) is one number. A unison is a 0, 2 a whole step, 4 a major third, 7 a fifth, etc. Written in MIDI intervals, our theme would be 0, 2, 1, -3, 6. If you have no musical training this system is actually easier than the major, minor, augmented, diminished intervals we use in theory classes. In fact, theorists rely exclusively on numbers to describe the complex, rich, and abstract harmonic and melodic content of 20th Century styles.  MIDI data files have been a standard for several decades. The main advantages of working with MIDI are file size (they are tiny by any standard), and universal compatibility with all computers and synthesizers since the early 1980s. Other features: Midi files are easily edited, not only the notes, but also instrumentation. Because it is not actual sound, but instructions for creating sound, a single note's length, attack time, release, or pitch can be changed. A common scenario might be: you hire a pianist to record a jazz standard. He makes one pass on a MIDI enabled grand piano. You record the MIDI, not the audio. He makes a few mistakes, but plays a stellar solo. No matter, send him home. Open the MIDI data, correct the mistakes, move the solo to another track and set it to a vibe instrument. Extract the bass line and 39

put that on a sampled upright bass. Then you send everything else back to the live grand piano and record the actual audio. Another feature of MIDI deserves a separate paragraph; online MIDI collections have exploded. I rely on them more than our music library. Example: I want to show a class the final measures of Bach’s WTK book I Prelude in C as a classic example of a pedal tone. A quick Google search, 5 second download and I have books I and II, which I can easily peruse (listen to, but not see) using quick look. These can then be imported into my favorite copy program or Logic. I have literally done this in front of a class for both classical and jazz standards. From there I can improve the orchestration or arrangement, experiment with other voicings or chord progressions, hearing the results immediately, and add note colors or analysis. The quality and arrangements are uneven, you're a little at the mercy of the creator, but I have found several trusted sources.  There are disadvantages to MIDI: 1) MIDI is not sound. The sound is generated by local equipment. What that means is the clever orchestration that sounds great on your equipment will sound cheesy on a cheapskate Bulgarian TV exec’s setup. You have no control over exact instrument choice once you’ve attached it to email. 2) MIDI files contain no information about notation. There are no key signatures, time signatures, measure divisions, or staves. For example, piano music often isn’t written the way it’s played. Bach’s WTK book I prelude is an excellent example. It is written as eighth notes, but played as tied eighth notes. This often complicates my Bach/MIDI trick. 3) MIDI files are written in a very low level, cryptic language. You can't just open a file in any text program and make sense of the data. Writing and reading MIDI files is a task for experienced programmers only. 4) It will soon be replaced. Frequency, cents, and MIDI may seem like ancient Sanskrit, but in computer composition you will often use all three, possibly in the same file, so fluency is a must. •

XML

… is a Music Extensible Markup Language, similar to HTML. It is the second attempt, after NIFF, to standardize a music notation format, and has received wider support.  It has similar disadvantages to MIDI; it contains no sound, it contains no playback instructions, but it does contain information about key, clef, time signature, measures, etc. Another advantage over MIDI is that it is text based. This means that you can open an XML file using any text editor and with some practice make sense of, edit, or even write XML code. •

Pitch Class, Intervals, Notation

… could be used to describe the melody in an email. A pitch class is the single letter that represents a note and all its octave transpositions. A4 is the frequency 440. But 'A' refers to all pitches at multiples of 440: 220, 110, 55, 880, and so on. If I said my melody is C, D, E, C, F , most of you would get the idea. Intervals describe the distance between two notes. For this text, a short list of the intervals in the key of C will do. "M" stands for major, "m" for minor, "P" is perfect, "A" is augmented, "U" unison or 1. A "Step" is two piano keys. Below is a chart showing all the different ways we describe musical intervals.

40

Letter Interval Steps Piano keys MIDI number Cents

C U 0 0 0 0

D m2 ½ 1 1 100

D M2 1 2 2 200

E m3 1½ 3 3 300

E M3 2 4 4 400

F P4 2½ 5 5 500

F A4 3 6 6 600

G P5 3½ 7 7 700

A m6 4 8 8 800

A M6 4½ 9 9 900

B B m7 M7 5 5½ 10 11 10 11 1000 1100

In the language of intervals, the theme would be U, up M2, up m2, down m3, up A4. Pitch class, intervals, and music notation are handy, if my TV exec reads music. If he does, I can send him … •

PDF and Screen Shots

There are two slick methods for getting notation into an email. The first is a screen shot. Select ⇧⌘4, or ⇧⌘⌃4, which places an image of the screen onto the clipboard. I use screen shots to create examples for tests, document unusual errors or dialogs, grab small sections of a web page, but also to send one or two measures of a student’s assignment with my notes and corrections. If you want a nice layout and maybe multiple pages, PDF (portable document format) is the next choice. A PDF can be read on any computer, any operating system. Creating a PDF is simple in OS X. In any application, open the print dialog and choose PDF at the bottom. You can choose Save as PDF, or Mail PDF, and it will not only create the PDF, but also attach it to the bottom of an email message. Here then is a screen shot of the theme, in a higher-level language of pitch class, which I hope you recognize by now. Believe it or not, there are still people who would not recognize any of these digital renditions. My last choice? Call him up and sing it. •

AIFF, Wave, MP3, AAC

… are digital audio formats. Audio is such a universal language that any human, heck, even other species, will get that you’re making music. Though universal, sound has some disadvantages. Audio recordings have no information about note values, durations, chords, theory, notation, or instrument assignment. This means you can’t change the orchestration, you can’t edit single notes from a chord, extract a bass line, or change keys as you can with MIDI. Digital audio, in terms of format, is surprisingly simple to get into and out of a computer. It's just a stream of numbers. There is very little detail (compared to a text or notation document). A wave file (.wav) is an IBM standard. Most Macintosh programs can handle wave files. SDII stands for Sound Designer II, which is the pioneer digital audio editor developed by DigiDesign, the authors of ProTools products. The most reliable and interchangeable standard for digital audio is AIFF, or Audio Interchange File Format. AIFF (Mac) or WAV (Mac or Windows) are the standards for audio CDs. MP3 and AAC are the two most common compression formats. While compression is good, they can degrade each time you open, edit, and save. (This is because compression is a process that interpolates samples. A little graininess in one copy will be interpolated again and again with each compression.) Engineers still do all of their important masters in AIFF or WAV. The quality of these files also depends on… 41

•

Sample Rate and Bit Depth

Most digital editors offer sample rates from 5k to 128k and bit depths of 8, 16, 24, or 32. Logic Pro will record in 16 or 24 depth, but will not read 32. Drag and drop a 32 bit segment and it is converted to noise. Sample rate is analogous to frame rate in video: the number of snapshots taken of a sound as it is digitized. More on this later, but 44.1k is the standard for CDs, 48k for digital video. The reason for this discrepancy is complex and absurd, but we live with it. The story goes something like this: 44.1 allowed 74 minutes, which was enough to accommodate most classical works, in particular Beethoven's 9th. It's less and less of an issue, and CDs are on the way out, so recording in 48k makes the most sense.  We use 48k because you need at least two samples to document, even roughly, a wave: one for its positive value, one for the negative. (See Figure 5.1.) We hear up to 24k, so you double that. This is called the Nyquist Theorem. Granted, you end up with a square wave even if the original was a sine, but at least you capture it. Oversampling is discussed further down. Fig. 5-1 Nyquist Theorem

!

!

Lower frequencies have a clearer contour because there are more samples to represent the wave. Figure 5.2 shows a triangle wave at 1000 Hz, which would be sampled 40 times at 48k. That’s a pretty good representation of a triangle. A lower sample rate, say 10k, would have about 9 samples per wave (three of which are 0, so they don’t show up as a bar) as shown in the next image. You could extrapolate a triangle wave, but that’s because, for illustration, these samples fall at tidy intervals. This could have been a more complex wave with many contour changes between the samples. Though more complex, the wave just happen to fall at the shape of a triangle, but the finer details would not be registered at this sample rate. The other problem is, when the sound is played back, each value is essentially the length of the time between samples, so the shape of the triangle wave will not look like a triangle, but will be distorted and terraced. Fig. 5-2 Sampled Triangle Wave

42

Bit depth is the resolution or size of the numbers used for sampling. CDs use 16 bits. These numbers are used to document the amplitude or size of the wave. Though not technically correct, it is analogous to decimal precision. This means, in practical terms, that numbers are rounded to the nearest precision. Even with a high sampling rate, a low bit depth creates terracing. The first image in Figure 5.3 shows a ramp wave where the sample rate and bit depth are high enough to give a decent contour of the descending wave. There are twenty different bars; each at a different height, so you could say it has a depth of at least 20. In the second, while the sample rate is the same, the bit depth is lower. There are only 5 values, and samples are rounded down to the nearest precision. This is no longer an accurate representation of the wave. Fig. 5-3 Low Bit Depth Terracing

 Think of sample rate as the X axis resolution, bit depth is Y axis resolution. Sample rate represents frequency, bit depth amplitude. We will discuss what this means in terms of quality in the next chapter. In general, higher numbers are higher quality audio.

!

All this information has bearing on our goal; communicating with others. That is because sample rate and bit depth affect … •

File Size

My Dragnet theme is about a minute long. I could send a 128k 32 bit audio file of me singing the theme. It would weigh in at 60M; enough to choke most service providers. Or I could send a 4k MIDI file (that's 1/15000 the size). These days 60M isn't much, but what if it is with a slow connection? And while it's next to nothing on a hard drive, multiply that by two for stereo, then 32 tracks for a large project, then by 360 for an hour of recording—typical for a single cut on a CD. Now we’re talking about 80 gig. That’s for one CD track. A whole CD would be 800, 2.4 terabytes if you’re working on several CDs. It adds up quickly. But storage space is not the only consideration. File size comes into play each time you copy, backup, or open a project. Consider a music school that has five performance and rehearsal spaces. You want to implement automated recordings. The system will turn on and off when the ambient sound 43

crosses a given threshold. On average there are 10 hours of activity in each space every day. That's 50 hours a day, a little less on weekends, so about 300 hours a week. Just to be safe, make it redundant, so 600 hours a week. Archive up to six months in case a performing group forgets to request recordings of their entire season (of course, I'm describing our system, and I have indeed pulled 6 month old recordings). That's 18000 hours. This system, if recorded at 128k, 32 bit, would require 67 terabytes. The more rational 44.1k, 16 bit is a manageable 11 terabytes. That's still pretty daunting on a tight budget. Figure 5.4 shows two charts. The first is the smaller formats only, since they aren't even visible on the second chart, which includes audio formats. Imagine each is a progress bar as you open a file and wait for it to load, or copy to a server from home. Fig. 5-4 Relative File Sizes

Now consider the workflow: faculty, clients, and students want to edit and publish their own material. They will often download the files from home with slow connections. The files are usually an hour long, often several files. I just did an (admittedly unscientific) trial, with a mediocre connection to emulate my Bulgarian producer's internet connection, and I'm getting about 10M a minute. The download times are shown in Figure 5.5. MIDI is 5 seconds. High quality audio is 6 hours. Which should I use? Do I really want this producer to sit for six hours, thinking about other engineers/composers who can get the same idea to him in 5 seconds? Fig. 5-5 Bulgarian Download Speeds

Even if downloads were not an issue, do you really want to have to explain, over and over, why students can't burn a CD directly from your recordings because they are the incorrect, higher quality format? Now consider how the recordings are used. In the rehearsal spaces faculty typically only need to review the recording once, and the quality is not critical. The performance spaces will likely be edited and published, but to CDs at 16 bit 44.1k. You will lose any quality gained from a higher sampling rate. 44

So we've settled on 44.1k 16 bit for performances and AAC for everything else. That gives us a year of acceptable quality archives on 8T. That master class you did at the beginning of the year, for which you didn't request a recording, but now you need for an audition CD? Sure, I'll have it to you in 5 minutes at a reasonable quality. A purist, who insists on recording everything at 128k won't. •

Exercises 5:1 Complete the following charts. The first documents the sizes of digital audio that 5 CD projects might require. In the first, record 1 minute of sound at 48k 16 bit, mono. Use Sound/Characteristics (or just calculate) to change the sample rate and bit depth for other values in the chart. The first column is 1 minute of audio. Multiply it by 60 for an hour in the next column, then 10 for a multi-track session, then 2 for stereo tracks all around, then 5 for all the projects you're working on. In the second chart, check all true answers for each file type 1 min

* 60

* 10

*2

*5

MIDI

AIFF

XML

48k 16 bit 96k 16 bit 96k 24 bit 96k 32 bit 128k 32 bit PDF

Small file size Large file size Accessible to non-musicians … to luddite musicians … to computer savvy musicians … to computer programmers … to computers Contains clefs, keys, tempos Contains actual audio Orchestration can be changed Notes are easily edited/changed 5:2 Give the decimal equivalent to these binary numbers: 00101010, 00000110, 10000001, 10010001, and these hexadecimal numbers: 000000EA, 0000001F, 000000F0, 00000099. What is the next number above each of these binary and hex numbers: 00111011, 00011110, 00111111, 00000099, 0000FFFF, 0000A4CF. Give the letter names, interval name and distance in cents for these MIDI numbers; 60 to 66, 60 to 69, D to E, G to B, D to B. How many octaves between 100 and 6400? 5:3 Search online for and download a MIDI version of a jazz tune. (Try something obscure.) Use Open with… to open it in Logic Pro. Select Screen Set/Rename and type "Instruments." Press 2 to open another screen set. Press ⌘-2 to open the mixer in this screen set. (Optional; close the arrange window in screen set 2.) Rename this screen set "Mix." Solo (in the Mix window) each track to familiarize yourself with 45

the material. Switch back to screen set 1 (press 1) to delete any tracks that don't seem to add to the arrangement or have incorrect notes. In screen set 1 open the Media area then click on Library. Click the channel strip button (Silk Clav in the image) and use the library to browse different channel strips. (Optional, click on plug-ins and browse them.) Choose pleasing instruments for each track. Switch to the Mix screen and double click on any plug-in or instrument to open them in a new window. Explore plug-ins and adjust settings. Open screen set 3 and click on the Lists icon then Events. Also select Piano Roll. Name this set "MIDI." Continue to create a pleasant mix using screen 1 to select channel strips (and plug-ins). Use set 2 to adjust the level and pan position of each track, and also to open plug-in windows to make adjustments. Use 3 to edit MIDI data. For example, delete Control or Program data at the top of the track. It will interfere with your own settings and automation. Click on the filter categories (Notes, Progr. Change) to filter data. Click-drag to select a group of items in the status column. Also in window 3 use the piano roll to select different tools (eraser, pencil, velocity) to edit or adjust velocity of individual or groups of notes. Create a good arrangement. Press 4 to open screen set 4. Name it "Score" and select the Score window. In the local menu (the menu inside the Score window) choose View, Page View. In a single folder save your arrangement as a MIDI file, bounce it as an AIF (PCM, 44.1k, 16 bit), mp3, print the score as a PDF, and take a screen shot of the MIDI screen set. Use the get info dialog to set open with to QuickTime for the MIDI file, AIF, and mp3, set the other two to open in Preview.

46

6. Mics, Cables, Placement Essential for musicians; not just those who sing. In a few chapters we’ll have a foundation on which we can build techniques of electroacoustic and computer composition. The first topic is one of the earliest styles of electronic music: Musique Concrète. While one might argue that any recording, even acoustic orchestras filled with luddites, is set in "concrete" and will never change, Musique Concrète is generally accepted as not just recorded material, but recordings of real sounds that are manipulated in unusual, interesting ways. Before you can manipulate sounds, you have to collect them, which requires mics, cables, and interfaces. •

Connectors and Cables

… can be balanced or unbalanced (the details of which I’ll explain in the noise section), stereo or mono. For now you just need to know how they relate to the various connectors. Unbalanced lines require only one wire for signal, balanced requires two. In addition, each cable needs a ground wire, so a mono unbalanced line requires two wires. One for the signal, and one for the ground. An unbalanced stereo cable has three wires. One for the left signal, one for the right, and one for the ground, which is shared. A balanced mono cable also uses three wires, two for the balanced signal, and one ground. A balanced stereo cable (less common), uses five wires; two for the balanced left signal, two for the balanced right, and one ground. Common connectors are: RCA, TS (tip, sleeve) TRS (tip, ring, sleeve), and XLR, as shown in Fig 6.1. Fig. 6-1 Connectors

RCA connectors are unbalanced mono, but usually come in a stereo pair. XLR cables could be stereo, but are invariably balanced mono. A TS can only be unbalanced mono. TRS is the only cable that is regularly used three different ways: as unbalanced stereo, balanced mono, or as an insert. This connector is what you see at the end of headphones, in this case stereo unbalanced. To be used as an insert, the other end is split into two TS connections. This single patch point saves space on a board because it is both a send and a return. The tip is the send. The ring is the return. At the other end one connector is plugged into the input of the device, the other into the output.

47

The cables I always carry: firewire cable, 400 to 800 converter, a mini to mini TRS, a stereo mini to RCA adapter, an RCA to RCA, and (2) RCA to quarter adapters, a quarter to mini TRS, and a mini to quarter TRS. This collection gets me through most ad hoc recording or playback situations. The two most common… •

Microphone Types

…are condenser and dynamic. Both have a diaphragm (like a small drum head) that moves back and forth in response to sound. The diaphragm of a condenser is made of two thin metal plates. An electrical charge is placed on each disk, one positive, the other negative. The change in electrical voltage between the two is measured when the sound pressure moves one. That voltage is transmitted to the recording device. The diaphragm of a dynamic microphone is connected to magnets inside a coil of wires. The sound waves move the diaphragm and magnets between the wires that pick up the electrical field. Most speakers have the same components as a dynamic mic. You can even use headphones as a low quality microphone in a pinch. Condensers require power either from a battery inside the mic (quieter, but cost a little and if they aren't rechargeable, less green), or from phantom power sent down the cable from a mixer or the recording interface. Phantom is less hassle, more environmentally friendly, less expensive, and won’t damage dynamic mics. (This is important to know because dynamic mics don’t require phantom power—but are often used together with condensers on a mixer with just one phantom power switch for all channels.)  I’m generalizing, but condensers are common in the studio, usually in an isolation booth since they are more sensitive, and will pick up other sounds in the room. The other common application is as a pair to record a large group of instruments, even an entire orchestra. They can be used in live sound reinforcement (PAs), but are more prone to feedback.  Dynamic mics, in general, are good for isolating several instruments in the same room. Because they isolate, are cheap and rugged, they are the standard choice for live sound. The following is a very general comparison of the two. Condenser Expensive Delicate Requires phantom or battery power Wide range, depending on pattern Typically used as studio spot mic or stereo pairs •

Dynamic Cheap Rugged No power required (passive) Effective range of 6 inches Used as spot mic, live sound

Microphone, Cable Care - Keep microphone clips paired and attached to their microphones. They are less likely to break or get lost. - Store mics in a box, closed closet, or bag. This will protect them from dust, which can reduce sensitivity. - Twist the mic stand or boom onto the mic holder (after loosening the anchor thingys) rather than twist the mic to attach it to the stand; less chance of dropping. 48

- To reduce possible damage to the mic and speakers, turn phantom power off before connecting or disconnecting a mic. You should also power down speakers and amps to avoid the mostly annoying, and possibly damaging pop. The rule of thumb for speakers is: last on, first off. - For nearly ten years I coiled cables using the standard alternating backward loop (also used by rock climbers and competitive calf ropers), and that's how most studios will expect you to do them, but now I just "gather" them in arm length sections. They end up in a loose figure eight, which, in my experience, believe it or not, looks messy but is less likely to tangle than the alternating loop method. - Do not roll cables around your arm. Ridicule anyone who does. - I never tape cables, either to label or for security. I prefer to plan the routing more carefully, block off those areas, use rug runners and maybe tape the edges of the rug. Tape is evil. We’re getting closer to your first Concrète composition. You’ve picked out a mic, cable, plugged it into a phantom powered interface (firewire mixer, Mbox, 002, the line input of your Mac), and are ready to record backward satanic messages. Finally, the question you’ve been anticipating. •

Mic Placement

When recording voice, piano, guitar, didgeridoo… where do you place the mic? The frustrating but true answer: anywhere you like. I can’t tell you exactly where to put the mic because I don’t know what you like, or what your performer likes, or even where the sweet spot on that particular guitar is, so I can’t tell you where it sounds best. You have to use your ears, and these techniques. - There are plenty of opinions online, so start there, but don’t let that replace your ears. - Ask the performer. They know their instrument and if they've done some recording will tell you. They might like something different from you. - Listen for a good spot. It’s probably different on every instrument, even of the same type. The spot that works great for one cello may be right over the wolf tone of another. - Experiment—take the time to find the sound you want for this project. As I've watched visiting engineers work, the one thing they do different from amateurs (like me) is take the time, sometimes hours, to try things and get it right. Remember the experiment with the phase cancellation. Sometimes inches make a big difference. - Get it right before it hits the mic. That is, use the mic to document the sound you want. Don’t rely on post-production to repair mistakes made up front. If you're ever lucky enough to stumble an original 16 or 24 track sessions of a great classic, you'll hear that the raw tracks are very close to the final version. - Don’t let poor equipment/halls/circumstances get in the way of the music. Of course use the best you have access to, but do the best with what you have. Contrary to the previous paragraph, the engineering for many of the iconic hits from the 60s and 70s is 49

awful. Some even in mono. (Stereo didn't exist.) They sound great because it’s good music. The poor quality recording just adds character. - Remember there are no definitive methods. I've caught a lot of flack for this opinion, but take distortion as an illustration. We go on and on about the nuanced variations of guitar distortion, which, fifty years ago, in any form was wrong. The earliest examples of distortion came from damaged equipment. You can’t be more wrong than that. The strings in Eleanor Rigby (Beatles) were mic'd so close as to make the musicians uneasy, scooting back between takes. This was contrary to the accepted wisdom of the time and contrary to the express rules of EMI. Paul wanted it to be edgy. It worked, and is now a standard for that kind of intimate, full, studio sound. While there are many opinions and styles, and I encourage you to develop your own opinions on positions and sounds, there are a handful of techniques you should know. •

Proximity … is the distance from the mic to the source.

 A close position will result in a fuller sound, better isolation, hence less feedback when working with a PA. But the closer you are, the more pronounced the change (in dB) if the performer moves. A closer mic will also pick up more mechanical noise; keys clicking, smacking lips and tongue, p’s and b’s from the vocalist, dampers lifting, fingers on strings. Finally, pointing the mic to one part of the instrument will isolate just a narrow part of its overall sound. If you do mic close, it is important to realize (in the case of PA reinforcement) that the performer has as much as 50 dB change at their disposal through proximity. If you want complete control, encourage them not to move, or even touch the mic to the corner of their mouth. Then again, I've always admired groups who use proximity to their advantage. They mix themselves by stepping into a mic for a solo, using the mic and the PA as part of their instrument. Many folk groups have returned to a single condenser instead of spot dynamics on every instrument. This takes the engineer out of the picture, which is often a good thing. If isolation is not an issue, I mic much further back than many engineers. This records less mechanical noise, has a more uniform and natural dynamic change, and allows the instrument to do it's job of blending the overtones into a true piano, guitar, trumpet, etc. On the other hand, a close microphone has allowed musicians, especially vocals, to explore a wider range of expression. The intimacy of Nora Jones, the strength of the strings in Eleanor Rigby, the breadth of Bobby McFerrin, would not be possible without a close mic position. The last consideration regarding proximity is room mix. If you want to capture the natural reverb of a performance hall, the mix (wet vs. dry) can be controlled by proximity. Some engineers capture the hall with another stereo pair near the back, often even facing the back walls to pick up the reverb, which is then mixed later. I try not to fiddle with nature. If it sounds great with me standing 10' away that’s what I record. If the room doesn't provide the reflections you want, then mic close and put reverb in during post-production. In addition to proximity from the source, you have to consider the distance between multiple mics. Two mics at different distances from the source can set two signals out of…

50

•

Phase

Figure 6.2 shows one of my first two track recordings: a violin and piano. One mic was placed near the piano, the other on the violin. They were about six feet apart. You see signal in both tracks even though only the piano is playing. The top track is the mic on the piano; the bottom is the mic for the violin. Since sound travels through the air, it reaches the violin mic a split second after the piano mic. The difference is only milliseconds, so is nearly imperceptible (as a delay, at least). In effect this virtually duplicates and displaces the piano six feet. It is mixed with itself, but out of phase. Fig. 6-2 Waves Out of Phase

 To illustrate, imagine recording a pure sine wave at 100 Hz (the range where you get that solid thump of a kick drum) with two mics 5 feet apart. A 100 Hz wave has a length of 10 feet. Five feet is half the distance of the wave’s length. The second image in figure 6.2 shows what I would see in a digital editor. Both mics pick up exactly the same wave, but the second mic, because of the time it takes for the wave to reach it, will pick up the beginning of the wave when the first mic has already recorded half of it. When mixed down the two cancel each other out. That’s right; no sound. Their energies are mirror images, one telling the speaker to push, the other to pull. So the speaker does nothing. This is called deconstructive interference or phase cancellation. Of course it would be hard to reproduce this precisely because the wave bounces around the room, but it is a theoretical possibility. It can be reproduced using precise (incorrect) editing techniques.  Phasing errors come from poor mic placement, poor speaker placement, poor editing techniques (e.g., deleting a millisecond too much from one track), or incorrectly wired speakers. The solution is to isolate better (which hinders performance), or use a single stereo pair.  You can also rely on the 3:1 rule. The distance between two mics should be three times the distance of the first mic and the source. If mic A is 5 feet away from the piano, then mic B should be 15 feet from mic A. All this said, you should trust your ears. Inches matter. To illustrate, play a sine tone through stereo speakers (but from a mono file, so it's the same in both speakers) and move around the room with one ear plugged. You can find spots where the sine wave more or less disappears. It's an astonishing experiment. Two inches from that spot it will reappear. The moral: you can very easily improve the sound of a mic by moving it an inch. •

Axis

… is the angle of the diaphragm (not just the mic, but the diaphragm inside) in relation to the source. For proper placement you have to know the orientation of the diaphragm. Most dynamic mics have the diaphragm on the end, perpendicular to the barrel, so for on-axis positioning you point it at the source the way you would a flashlight or shotgun. This is also true with small diaphragm (and hence, small barrel) condensers. Most mono, large-diaphragm mics have horizontal diaphragms, so the performer would face them as if using a hand-held mirror. In these mics there is a front and a back. Manufacturers 51

usually put their name on the front of a mic. This is also true for many stereo condensers. The diaphragms are horizontal with the barrel, so you would stand them up like a lighthouse. Mono mics should always be placed on axis. This is the single most common error in PA use. Inexperienced performers hold a microphone so that the diaphragm is pointing toward the ceiling. This places it 90˚ off axis from their voice. The signal will be weaker and thinner. Proximity and axis account for the startling change in volume with multiple announcers: one person can barely be heard, the other is obnoxiously loud. People blame the engineer, but it's the announcer's fault. Don't hesitate to offer suggestions to performers or announcers, and instruct them about proximity and axis. Simply turning it up, which the announcer and everyone in the audience expects, is not the answer. Axis can be used as a strategy for isolation, to control feedback and to achieve stereo patterns (described below). •

Exercises 6:1 Play a pure mono sine tone through speakers in a room. Walk around the room with one ear plugged looking for "dead" spots. Create two tracks (L/R) with identical sine waves, one inverted. Play them through headphones, but hold them on one side of your head at about six inches apart. Move them so that your one ear is close to one, then the other, then exactly in the middle. The sine should disappear. Repeat this experiment after selecting just one channel and inverting the wave. In the same room, clap and listen to the sound of the room. Move around and see if you can find a standing wave (where the reflections bounce back and forth and add onto each other). You will hear a rapid, odd sounding echo, like flipping a spring doorstop. 6:2 In a Logic session import the same audio file (mono or stereo) into two tracks. Zoom in to the sample level. Select just one of the regions in one track and start playback. Use nudge to move the region to the left just slightly. Note the phasing (it will sound more hollow). Repeat this several times so that the sound is increasingly hollow. Notice the frequencies dropping out. Bounce the file, open in Amadeus, generate a sonogram and hand in the sonogram showing the bands of missing frequencies. Do the same with a noise source. 6:3 Record your voice or an instrument using a number of varied mic placements; close, far, on axis, off, different positions around the instrument. Compare the results. Place two microphones 10 and 15 feet from a source. Play a single note and examine the resulting stereo file, confirming that they are out of sync. 6:4 Find a rich source (e.g., play pink noise through quality speakers) and record using two condenser microphones. Begin with them exactly on the same plane and distance, then slowly move one further away. Mix them to a mono file (either in Logic or Amadeus.) Listen to the phasing that results.

52

7. Recording Techniques Ways to improve audio. You'll thank me later. •

Stereo/Mono

Your final electro-acoustic composition should be stereo. While I have no scientific backing, I think our brains can interpret elements of a stereo mix more efficiently. The guitar in the left ear is being processed by our right brain, and the flute in the right ear, by the left side of the brain. Logic Pro and SC can produce very interesting results in 5.1 ("quad"—what we called it back then), but don't let that take over the project. Mediocre music is rarely improved with more channels. On the other hand, SC has extremely powerful spatialization tools, and should be exploited. When laying down tracks in a multi-track session, each track can be mono or stereo. Mono tracks are then placed in a virtual stereo field by adjusting the pan position. If the subject you’re recording includes important stereo placement (such as an entire drum set, street noises, a quartet or entire orchestra), use two mics and a stereo track. But since this takes twice the space, it might not be the best choice with individual instruments. A piano is rich enough and wide enough to make a stereo recording worthwhile, but maybe not a guitar or voice. Some engineers place two mics on a single instrument, and while the sound is not true stereo, the two tracks are different, and when panned left and right thicken the sound. •

Overdubbing

… is ubiquitous in popular music. Rather than place two mics on a single instrument, record an instrument twice, then mix both left and right. This technique is most commonly applied to guitars or vocal tracks. One reason is that overdubbing masks imperfection. Two out of tune lead vocals, when meshed together, magically sound more in tune. In fact, most synthesizers and guitar effects achieve a similar result with a mono signal that is split, then one or the other is detuned for a gentle phase shift. Examples of overdubbing include the guitar riff in La Grange (ZZ Top, studio version), Cocaine (Eric Clapton, vocal, both panned center), Martha My Dear, Beatles, Little Bitty Pretty One, Slippery Saint Paul, China Grove, Black Water, Eyes of Silver, and Long Train Running (Doobie Brothers, guitars, panned left and right). You can go overboard with stereo mics, resulting in a minor engineering error. As an example, consider a session with a stereo pair on the piano, the drums, a stereo feed from the electric guitar, dubbed vocals, and finally stereo mics on a three piece horn section. Since all of these are stereo, you would pan them all center to get the full effect of the two channels, correct? Wrong. This is called… •

Big Mono

… and it compromises the overall image, placing every instrument (or stereo set of instruments) essentially in the center. Even if something was recorded in stereo, consider moving the image to the left or right, giving a place for each instrument or group. 53

Important note: For true stereo panning in Logic you have to use the Imaging/Directional Mixer plug-in. More about this in Chapter 9. Each mono microphone may offer a variety of … •

Patterns, Stereo Pairs

Unidirectional patterns will pick up just one direction; in front. There are also cardioid (shaped like a heart), omni (all directions equally), and figure 8 (front and back) patterns. Two microphones, usually condensers, are often combined in a stereo pair. Stereo mics have two capsules integrated into a single body, but you can also place two mono mics on two stands or a stereo bar in any of these patterns. One popular configuration is an x/y figure 8. The microphone capsules are placed directly over one another, each facing 45˚ off axis to the left and right of the source (90˚ from each other), both in a figure 8 pattern. The figure 8 preserves the stereo image in front and adds a nice hall mix, also in stereo from the back part of the figure 8. The two capsules are placed meticulously over one another to eliminate any distance between the two capsules, which would invite time delays and phase cancellation. A mid-side pattern is very clever, and slightly mysterious. Two capsules are again right on top of one another, but one is on axis with the source, the other faces the sides (90˚ off axis) in a figure 8 pattern. To create an image, the center mic is mixed to the center and the sides are mixed to both the left and right channels, but with one channel's phase inverted. This produces a less convincing stereo image, but the advantage is a clear mono reduction. Since the left and right channels are mirror images, when mixed to mono they will cancel each other out and you are left with the center capsule only. This pattern is used extensively in television programs, where mono and stereo playback are (were, does anyone own a mono TV?) practically equal possibilities. In Logic Pro, the directional mixer has a mid-side option, applying all the wizardry on a single stereo pair. But phase cancelation isn't wrong per se. It’s only wrong when it compromises the sound. The truth is, we encountered phasing in every aspect of music. It is, as we will see, what makes music work; waves interacting in both constructive and destructive interference. It is used to great effect for EQ, phase shifters, flanging, and enhanced stereo. In my first digital recording the sound reached the second mic later, causing delays and phasing. In that case, something I wanted to avoid. But notice our ears are not on the same plane. Sounds on the left will reach the right ear both delayed and phased. That is in fact one of the cues we use to interpret the stereo world around us; a vestige of hunting or being hunted, where audio location was critical to survival. For that reason, my favorite pattern, not only for the results, but because it makes so much sense, is an ORTF (developed by l'Office de Radiodiffusion Télévision Française). There are many variations (Figure 7.1 is a little narrower than 110˚), but the original uses two microphones 17 centimeters apart at a 110˚ angle from each other (each 55˚ off axis from the source), cardioid or figure 8 pattern. In addition to accurately capturing the relative amplitude between left and right sources, it reproduces time delays, phasing, and filtering of frequencies that would occur naturally with our ears, which are also about 17 centimeters apart and at a 55˚ angle from a source. Though I understand that was not the original inspiration for the pattern, it is a good way to remember it; the distance of your ears, and at the same angle as your ears. I love 54

the reaction I get when clients first hear this pattern. It's usually along the lines of "it's so clear" or "how can it sound so good so fast." Fig. 7-1 ORTF

A variation of this type of spacing is a baffled stereo pair. This method places two capsules at roughly an ORTF configuration, but with some type of sound absorbing material in the middle. The Crown SASS is one of my favorite baffled pairs. They are relatively inexpensive (for a pair), but limited to stereo recordings, and a bit bulky, so we use them as hanging mics. •

Sample Rate

… is the number of samples per second used to record a sound. There are four criteria to consider when deciding on a sampling rate: Frequency range, aliasing, concrète, and file size. Recording at lower sample rates will reduce the upper range of frequencies being recorded, essentially filtering higher pitches and harmonics. CDs use 44.1k and video uses 48k. These rates were chosen because they are double the highest frequency we hear. Dog whistles, insects, and most instruments produce frequencies beyond our hearing range. Do we need to worry about frequencies we can't hear? Yes, for four reasons. The first reason is to avoid aliasing. Frequencies that are normally beyond our hearing range can interact with the sampling rate to produce unwanted aliases (described in the section on noise) that are in our hearing range. Recording at a higher sampling rate reduces the risk of aliasing. The second argument for a higher sampling rate is the theory that frequencies beyond human perception interact to produce important artifacts in lower frequencies, the range that we do hear. (In other words, natural aliasing.) In a concert hall those frequencies have a chance to interact and combine. But in a session where the instruments are isolated, they cannot. In this case the instruments can only blend digitally inside the computer during mix down. The theory is that masters should be recorded at 96k or higher to include frequencies in the 30k and above range, which can then blend during the mix. I've seen no solid proof of this theory. On the other hand, while you can always down sample, you can never up sample. Something that was originally recorded in 96 can be reduced to 11 to save space. But if you record in 11k you can never retrieve the information that would have been recorded at 96k. The next reason for higher rates involves sound effects and concrète work. Remember that when you change the speed (and pitch) you essentially change the sample rate. If you want to explore sounds beyond our hearing range by transposing them down, you need to record sounds beyond our hearing. The last reason is metaphysical, philosophical, anthropological and probably flippant. Future technology may reveal some information in those upper frequencies that are not available to our current technology. Record at higher rates so that future generations can extrapolate from 55

those archival documents the existence of beings, possibly from other dimensions, who sing along with our performances, but in frequencies that are beyond our current comprehension.  In review: up sampling avoids unwanted aliasing, captures frequencies that may generate natural aliasing, captures frequencies beyond our hearing for experiments in pitch change, and to capture information we don't yet know how to process. •

Noise

… is any unwanted sound. Noise can be in the room (air conditioning fans, computer fans, resonance that turns into feedback, buzzing lights, cell phones—they are noise booby traps), in the equipment (mixers, fx units or amplifiers), in the digitization or encoding process (low bit depth or aliasing), in adjacent rooms (a candy machine, drum set in the isolation booth, NIN practicing down the hall) and even out there in the ether (radio interference, SETI signals from Elvis). Some of these may have merit on their own, but if they don't add to the sound you want, they are noise. As John Cage discovered in the anechoic chamber (the inspiration for 4’33"), you will never eliminate noise, so the next best thing is to increase the divide between signal (what you want to hear) and noise (what you don't want to hear). The term for this division is signal to noise ratio. For example, an amplifier may list a signal to noise ratio of –80 dB. There are many ways to improve signal to noise ratio. It is, in fact, what an engineer does.  Turn things off, move to another room, isolate the source with partitions, absorb unwanted sounds with insulation. You can also improve signal with proximity: place the mic closer or move yourself closer, being aware of the issues described above (increased mechanical noise, unnatural sound).  To reduce system noise use quality equipment. Some engineers seem obsessed with preamps (the circuitry that brings the low level mic signal up to a useful range), and with good reason. Also, see the section on distortion below. How much should you pay for high end equipment? As a general rule, you get what you pay for, but with a steep diminishing return. That is, there is a difference between a $100 (approximate per channel for a consumer level mixer) preamp and a $1000 per channel preamp, but few people would notice the difference. This is also true for mics. The trick is to find the optimum between your budget and your (client's) ears. When I took a job for the local symphony I envisioned gold plated cables and Genelec speakers in every room. The truth is, they were on a very tight budget, and renewed my contract precisely because I got the most out of what they could afford, which wasn't much. Fig. 7-2 Bit Depth Noise

!

!

One source of noise is the digital conversion itself. The example on the right, believe it or not, is a sine wave that was either recorded at a very low depth, or recorded at a high bit depth, but at a very low amplitude, then amplified to normal levels. The terraces you see are the results 56

of rounding to the nearest bit. When you subtract the sine wave (by adding an inversion of the original wave), you can see (Figure 7.2) the edges are themselves a periodic wave, but in this case unwanted noise. The last resort, for me at least, is digital noise reduction. Signal processors make up for poor engineering and poor quality equipment. It is sometimes difficult to resist fiddling with all the knobs, flashing lights and technical graphs and readouts, but a skilled engineer will never reach that point. Processing always compromises the signal. •

Balanced Cables

… are used to eliminate noise from the ether (RF signals). A balanced cable is a selfcontained noise reduction system. Balanced refers to two mirrored signals. A single signal is duplicated and one is intentionally inverted just before it starts its journey down the cable. If, at the other end, these mirrored signals were just mixed together they would cancel each other out, being exact opposites. To prevent this, they are re-inverted before being mixed together. Fig. 7-3 Balanced Lines

!

!

!

!

The clever part of balancing a line is this; rather than blocking out the noise, it assumes there will be noise acquired, but that it will be present, and exactly the same, in both lines. The first example in Figure 7.3 shows two balanced square waves. The bottom is an exact mirror of the top. The next is not a mirror. It is a noise source captured along the path after the signal was inverted. A short burst of this noise was added to the first wave, and it will be present and exactly the same in both signals (third image). If you mixed the two lines at this point the original square wave, which we want, would vanish, because it’s a mirror image, but the noise would be added together and twice as loud. But if we re-invert the second channel at the receiving component (third image) and mix them down the square wave will add together, but the noise, which is now a mirror image, will vanish. If you prefer grasping this with math, think of the original signal as: [0.5, -0.5, 0.5, -0.5, 0.5, -0.5, 0.5, -0.5]. The inversion would be: [-0.5, 0.5, -0.5, 0.5, -0.5, 0.5, -0.5, 0.5]. Adding these two together would result in: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] We can run the math and get a sneak peak at SuperCollider. Chapter 13 provides a more in depth quick start for SC, and it is used in the chapters after that. For now, type this line in SC, select it, and choose Lang/Evaluate Selection. 7.1:: Balanced Lines [0.5, -0.5, 0.5, -0.5, 0.5, -0.5, 0.5, -0.5] + [-0.5, 0.5, -0.5, 0.5, -0.5, 0.5, -0.5, 0.5] ::

Now add random numbers as noise: [0.14, 0.36, -0.11, 0.102, -0.4, -0. 1, 0.319, 0.1]. We will add it to both the original and the inverted signal. Run each line separately. 57

7.2:: Balanced Lines With Noise [0.5, -0.5, 0.5, -0.5, 0.5, -0.5, 0.5, -0.5] + [0.14, 0.36, -0.11, 0.102, -0.4, -0.1, 0.319, 0.1] [-0.5, 0.5, -0.5, 0.5, -0.5, 0.5, -0.5, 0.5] + [0.14, 0.36, -0.11, 0.102, -0.4, -0.1, 0.319, 0.1] ::

The resulting arrays seem just as random, but watch what happens when we invert them and add them together. The .neg inverts the second array, the /2 reduces it by ½ since we are adding the original signals together. 7.3:: Balanced Lines Re-Inverted ([0.64,-0.14,0.39,-0.398,0.1,-0.6,0.819,-0.4 ] + [-0.36,0.86,-0.61,0.602,-0.9,0.4,-0.181,0.6].neg)/2 ::

•

Distortion

… is any misrepresentation or deformation of a wave. The jagged edges created by low sample rate or bit depth are an example of distortion. Distortion is often used creatively, to transform the source into something new (e.g., Jimi’s screaming guitar). But if your goal is accurate representation then it should be avoided. Fig. 7-4 Saturation, Clipping

The most common source of distortion comes from pushing an audio device beyond its physical or virtual capacity to respond. The medium is then saturated, or overdriven. Anything–mics, preamps, mixers, sub-masters, fx units, instruments, speakers, magnetic tape, digital bit depth, even your ears–can be saturated. The energy that goes beyond that physical barrier is not reproduced and therefore lost, or clipped. The result is a fuzzy or gravely sound (as well as some frequency loss). Figure 7.4 shows two channels of a sine wave increasing in volume. I’ve imposed an artificial limit on the lower channel. When it reaches this limit the shape changes, because the wave is still moving beyond the limit. The wave is clipped, changing the shape and therefore the sound. A mixer channel will usually have an LED meter, showing input level, or a red and/or green single LED to indicate if there is a signal present and if the signal is over the unit’s capacity. Logic Pro channel strips will show red when clipped. These two noise sources, system noise and wave distortion, are on opposite ends of the level spectrum. If your recording levels are too low, you are near the inherent noise in the system. If you record too high, then the signal is distorted. The trick is to record at… •

Optimum Levels

That is, as high above the noise floor as possible without clipping: not too loud, or too soft. Levels are set while the performer plays a loud passage, or even the loudest that the instrument can play. It is good to adjust levels during a rehearsal, playing the actual music to be recorded, if in a session with other tracks playing along. I find that a real 58

performance is more aggressive, and a better representation than a sound check. Keep in mind that any component in the path can be overdriven and should be monitored. You can use a single test tone to set all levels on all equipment to unity. This way a single meter can serve to monitor all levels. Analog tape was much noisier than digital and had a softer distortion; a more forgiving sound. Levels were often pushed daringly hot. Digital distortion is harsh, and the bed of noise is much lower. So if I'm not sure about a signal I err on the low side. It can be optimized later without adding as much noise. •

Aliasing

… is the final source of noise. To illustrate aliasing, imagine filming the second hand of an analog clock at a very low frame rate, say 4 times per second. Though the motion would be jerky, it would still accurately capture the movement of the second hand. Even a snapshot every 15 seconds would give you a pretty good idea. But with a frame rate of one every 60 seconds all the images would be the same; showing the second hand at the same position every 60 seconds. According to these samples the second hand is not moving As the rate of motion approaches the rate of sample, you get something worse than jerky motion. You get incorrect motion. A sample rate of one in 55 seconds would show the second hand moving backward. Fig. 7-5 Aliasing

 It's like the wagon wheel effect in movies, where a wheel appears to slow down or go backward. It’s actually moving faster, but the frame rate, combined with the frequency of the wheel, gives the illusion of backward or slower motion; an alias. The alias is the difference between the sample rate and the rate of the sampled item. This is easy to see with a square wave. Figure 7.5 shows a 1/1000th of a second segment of a recorded square wave. There are 12 cycles in the upper track; 12000 Hz. The lines represent a very low sample rate; 8000 Hz. There are not enough samples to show an accurate picture of the wave, ! but it does indeed return a positive and negative value at each sample. Using only the information returned by the samples, the resulting wave is 4000 Hz, a much lower frequency. The next example shows a triangle wave with samples that are close to the cycle itself, but fall a little further along the wave in the next cycle. According to the samples only, they slowly ascend then descend, documenting a non-existent, and much lower triangle wave in the second track. The final image is the same example zoomed out, showing the original wave, and the alias. (See chapter 20 for a composition technique using aliasing; sample and hold.)  When a frequency is close to a sample rate the alias will be the difference between the two; a sine wave at 44200 Hz, and a sample rate of 44100 Hz will result in an alias at 100 Hz. Even though 44,200 Hz is not within our hearing range, the alias is. For this reason, many engineers record at 96 or 128k. 59

•

Edits

… are the final component to skilled engineering. Years ago we edited reel to reel magnetic tape with razor blades, splicing tape, a splicing block and lots of time and practice. The next generation of recording media (DATs and CDs) could not be edited. For that reason I never owned a DAT or ADAT. I knew that direct to disk was coming, so I skipped a wave of technology. Random access digital audio is revolutionary in its flexibility, ease, and accessible editing tools. It is so easy to piece together complete works with sections of several takes, even with complex audio (e.g., full orchestra), that many students and faculty do it on their own. The graphic display makes it easy to see where to do the edits and can show waves with a resolution down to the sample level. It is easy to be precise and accurate, so it is unforgivable to make this one error: non-zero crossing edits. Fig. 7-6 Non-Zero Crossing Edits

A non-zero edit is when you select, copy, or paste at a point where the wave is not at 0, or the center of the graph. Such cuts result in a sharp change from positive to negative energy in only one sample. Figure 7.6 shows a non-zero cut. (This example is very short. The edits you actually do may be several minutes, but the principle is the same.) The first image shows the selection from an audio file at non-zero points of the wave. The next has the selection deleted, leaving a sharp edge. The result is a click or pop. The error is so common that most editors provide some simple solution such as a cross-fade, interpolation of the wave, snap to zero, or even an automatic "repair clicks." But a skilled engineer should avoid them in the first place. This is accomplished by either carefully choosing edit points at silence (and planning sessions to accommodate such edits) or by zooming into each selection before copying or pasting and choosing non-zero edit points. This goal is often complicated by a stereo file, where the left channel may be zero, but the right is in a positive position. In this case cross fades, or rapid fading in and/or out (e.g. a tenth of a second), or interpolating the wave can be used to correct the edit point. These sharp edged waves can also be caused by a slow disk or a low buffer size. If the system can't keep up with the DA conversion, it simply drops samples out. Fig. 7-7 Descending, Zero-Crossing

Zero-crossing edits are considered good form, but a real pro goes one step further and chooses zero-crossing descending edit points. A descending edit point is where the wave is moving from positive energy to negative energy. If you choose the beginning of an edit at a zero crossing, descending, but at the end select a zero crossing ascending, then the edit will indeed be at a zero point, but will have a quick change of direction. This first example shows a pair of non-zero edits, but one is descending and the other ascending (bad). The next shows a more correct pair where both are descending, non-zero edits. 60

Digital audio workstations use cross fades: millisecond ramping between two audio files, or two sections of audio files. But most of the work I do (concert recordings with stereo pairs) is with two track editors. Choosing edit points carefully is faster, and easier. If the edits are within the work—a measure here and there—I use crossfades. Amadeus Pro uses a system called smart edit. When on, crossfades are automatically applied at every edit. You can set the number of milliseconds in the preferences. For some work (stock concert editing), I find them useful. For other projects (precise experimentation), I turn it off. •

Optimum Engineering You can skip this section if you'd like. It's just my opinion; a bit of a rant.

After all this discussion, what sampling rate, bit depth, pre-amps, mics, placement, reverb or compression (see chapter 9), should you use? One of the best suggestions I can offer is to hone your aural skills and deconstruct the artists you like. When you're invited to work on a project it's unlikely they will hand you written music or an existing session as an example. They will give you a CD. You have to be able to reproduce what you hear, and just about anything you hear. Here is an example of the recording skills used to produce one of my favorite recordings: Find a house close to a road so you can hear the odd bus passing by. Use the cheapest mic you can, preferably used, from a pawnshop. Place it in the middle of a room routed to cheap preamps. Don't spend any time positioning the artists in the room. Let them sit wherever they want. Have the lead vocal sing through a can. Make sure the room has no sound proofing so we can hear the reflections off of flat, boxy walls. Have one of the musicians sit in a creaky rocking chair. Open the windows so we can hear the dog or random rooster crowing outside on the front porch. Set the mic next to an AC unit to get a nice rumbly dark pink noise. Don't do any post-production— no reverb, delays, panning, EQ, or compression. This sounds facetious, but I'm serious. I love this recording, and the ghetto engineering perfectly captures the intention of the artist. He is the only person, and I include the huge collection of classical composers, to whom I can and often will listen all day long: Tom Waits. His work defies and defines both musical and recording conventions as a matter of style. For example, in fact, example one, Chocolate Jesus is genius. The lyrics, the subtext, the gravely voice are all tied together with the recording techniques he used, or should I say, didn't use. I've read on the ever-reliable Internet that the rooster is real, and just happened to crow at the right time—a live chicken for god's sake. Similar ghetto recordings: Jesus Gonna Be Here, In The Colosseum (how does one mic a trash can?) Murder in the Red Barn, Lowside Of The Road, and Johnsburg Illinois, which just wouldn't sound right on a properly tuned piano. Second example: Poorly tuned guitars, everyone playing all the time without any consideration for height, elements of mix or texture—just one big jam, no pop filter on the vocal (where it was clearly needed), no headphones or a click track lest anyone accidentally play on the same beat. These recording techniques gave us the profound, inspiring, so emotionally draining that it's not even on my playlist, Like A Rolling Stone (Dylan).

61

Admittedly, these are cult artists, a bit left of main stream. So I offer this more familiar example: encase a microphone in a condom and immerse it in a bottle of water. Oh, make sure it's a powered condenser to increase the probability of electrocution. The engineer who should and might have been fired if John Lennon hadn't hidden the bottle/water/condom/mic technique behind his back? Geoff Emerick—three Grammies. The effect they wanted didn't work, but they tried, and they continued to experiment, adding to an impressive list of firsts, which we now take for granted: compression, close mics on drums, close mics on strings (Eleanor Rigby), looping, Concrète, sampling synthesis (the Mellotron that opens Strawberry Fields), music videos, reversal (Revolver), generative orchestration (A Day in the Life), dubbing to mask weak vocals, controlled feedback (years before Hendrix), moving the tape machines to the control room, phased vocals for richness (Tomorrow Never Knows), and synchronizing two tape decks with a pulse tone. I have a folder dedicated to flawed recording techniques. Taken out of the context of each artist's standing in the music world they would disgust any engineer with any training. That is, an engineer who is more afraid of liking what is bad than embracing what is good. I suspect that strongly and narrowly opinioned engineers are clinging to something they've read or heard rather than forming their own opinions with their own ears; knee jerk reactions (as with political demagogues) in lieu of rational thought. I know Waits didn't use a $2000 mic and I love his recordings. Most of the Rolling Stones' recordings are noisy—both hiss and hum. Hendrix wouldn't be Hendrix without a healthy 60 cycle buzz. This is even more apt for electro-acoustic music. I discovered early on that the spaces between the samples, the mistakes, the noise from handling the mic, often contain the most interesting material. A good engineer is one who has lots of equipment and knows lots of techniques, but doesn't feel obligated to use all of it/them all the time. A good engineer does what works … … best with the limited time, storage space, budget, and a wide variety of patrons that keep calling because he does what works. Deciding on a mic, preamp, and sample rate is similar to setting levels on a mixer. You don’t want to saturate resources, but you want to stay above the bed of noise. I can make a similar observation about live PA. For several years I played with a full sized jazz band that never used any kind of reinforcement. You heard correctly, we never used a PA, even on tour. At competitions we insisted they take it off the stage. Oh, sure, we were loud, and we played some big halls, but we were soft too. We knew how to play our instruments and we knew how to listen to one another. I’ve adopted this director's philosophy. I've also received plenty of calls from panicked stage managers because, for whatever reason, the single announcer's mic isn't working. Sure, I try to trouble shoot, but sometimes I offer this miraculous solution: "could you try just talking louder?" or "could you have people stop talking and listen carefully." Turns out that works really well. We listened to, and even danced to, acoustic music for thousands of years. Audio reinforcement has become a crutch, and all audio equipment compromises and colors the sound. Of course Bobby McFerrin's mic is not about reinforcement. He plays it like an instrument. Also, early blues singers discovered that a mic allowed a greater level of intimacy, which is standard procedure on female vocals such as Nora Jones or Tracy Chapman. 62

I live about 20 miles from the greatest snow on earth. I'm up there often, and it's not always perfect; too wet, too crusty, or too old. On days like that I try to remember; there is no such thing as bad snow, just bad skiers. Many platinum selling classics were recorded on (especially by our standards) horrible equipment. It doesn't get in the way of the music. In summary; use your skills as best you can with the equipment you can afford and focus on the art. If someone criticizes your technique, suggest they bring a chicken to their next session. Waits can have any mic he wants. He records with chickens. •

Exercises 7:1 Experiment with noise cancelation headphones. 7:2 Record an ensemble or other stereo rich source using ORTF, x/y, Mid-side, and mono. Compare the results. 7:3 Balanced Lines: In Amadeus, generate a sine wave at 40% amplitude in a stereo file. Split the tracks. The playback (upper right corner of the track) will show front right and front left. Change them both to front. On the second track, invert the phase and playback. There should be no sound, as the two cancel each other out. In a separate file, generate pink noise in mono. Select a section, switch to the sine wave drop a marker anywhere. (This will ensure a precise paste position in both channels.) Use option-arrow to position the cursor on the marker and choose Edit/Paste over. Repeat for the other track. This will reproduce the effect of a balanced line; the sine waves are inverted, and the exact noise has been introduced to both channels. Now re-invert the second track. Note the missing noise, and the return of the sine wave. Repeat this exercise using noise as the audio source and a sine as "noise." (Make sure it's the same in both tracks; generate it as mono or choose Tracks/Convert Track to Stereo.) 7:4 Noise Floor: In Amadeus, generate a sine wave at optimum levels in two tracks (or files); same amplitude and frequency. Invert one and mix them to confirm that they cancel each other out, and are therefore precise inversions. Reduce one by 60 dB then increase it by 60 dB to its original volume. (This reproduces the effect of recording one of them at a low bit depth, or at a high bit depth with too low levels.) Listen to the effect. Finally, mix the inversion of the original wave, leaving only the bit depth noise. 7:5 Signal to Noise: Realize Alvin Lucier's I Am Sitting in a Room. 7:6 Aliasing: be sure the sample rate is set to 44100. Generate a wave at 44000 Hz to create aliasing. What pitch do you actually hear? Repeat at different sample rates and frequencies. 7:7 Start with a high quality (44.1k, 16 bit, stereo). Generate separate files in the qualities listed below. Use these files to copy and paste non-contiguous sections (so that the edits are obvious), of musically accurate measures (i.e. maintaining the flow of each measure) into one file with musically accurate, cleanly spliced edits. The result should be a rhythmically accurate, but abruptly shifted collection of gradually degraded qualities, illustrating the difference of each rate, depth, etc. Here are the resulting qualities: 44k, 16 bit, stereo 63

44k, 16 bit, mono 44k, 8 bit, stereo 22k, 16 bit, stereo 11k, 16 bit, stereo 5k, 16 bit, stereo 5k, 8 bit, mono 7:8 Import into a logic session a stereo song that has very clear separation between left and right channels (e.g., early Beatles). Move the default pan dial left and right to confirm that you lose signal on the right and left. Insert a directional mixer control by clicking on an open insert and confirm that adjusting it moves both signals left or right. 7:9 Run these examples in SuperCollider. (Maybe wait until you have a few SC chapters under your belt.) Boot the server (Language/Boot Server), then run each set of lines inside the parentheses to see how bit depth and sample rate change a sine wave, white noise, and an audio example you can choose. (Open an example that has plenty of highs, like crisp and sizzly cymbals.) Move the mouse to change the quality. The left top of the screen is the highest quality. Move right to reduce the sample rate, down to reduce bit depth. Note the aliasing when the sample rate is near the sine wave. Note that with the audio example, even with a terribly low quality, you still recognize the example. I find this slightly amazing. You can resize the scope for a better view of the wave. ( {Latch.ar(SinOsc.ar([400, 600]), Impulse.ar(MouseX.kr(22000, 5000))).trunc(MouseY.kr(0.001, 1.0))}.play ) ( {Latch.ar(WhiteNoise.ar, Impulse.ar(MouseX.kr(22000, 5000))).trunc(MouseY.kr(0.001, 1.0))}.play ) ( ({ b = Buffer.loadDialog; a = PlayBuf.ar(1, b.bufnum, BufRateScale.kr(1)); Latch.ar(a, Impulse.ar(MouseX.kr(22000, 5000))).trunc(MouseY.kr(0.01, 1.0, 1)) }).play; )

64

8. The Way Music Works When does sound become music? When it's repeated. My definition includes the silence of Cage, the imaginary music of Johnston, conceptual music of Ives, bio-music of crickets, and metaphysical music of Oliveros, but for this text, we’ll narrow it down to my neighbors. If I hear a thump from the apartment upstairs I grab my first aid kit, worried that someone has fainted. If I hear a second thump, I grab my WWE costume. If there are three thumps, all evenly spaced, I grab my guitar; it’s a jam session. One thump; someone fainting. Two thumps; pretend wrestling. Three or more thumps; … •

Meter

Music is repeated sound. Varied repetition is interesting music. Varying the volume or accents of a pulse creates rhythms, which can be parsed into two musical ideas: groups and divisions. All metered music boils down to groups and divisions of threes and twos. For example, 3/4 time is a group of 3, division of 2; while 4/4 is a group of 4 (but essentially two groups of 2), division of 2; and 9/8 is a group of three, but a division of 3. Additionally, 6/8 is grouped in 2, divided by 3. Even more complex meters, like 7/8 or 5/4, are conceived and performed in groups of 3 and 2, where 7/8 can be 2 + 2 + 3, and 5/4 is usually 3 + 3 + 2 + 2. We count 1, 2, 3 to identify groups, and for the division insert "and" or "and a." (There are many variations to counting divisions.) Below is an example meter with HH playing the subdivision, the kick on the beat, and the guitar syncopated on the sub-division. Meter HH Kick Guitar

1 & 2 & 1 & 2 &1 & 2 &1 & 2 & 1 | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |

The simplicity or complexity of any style comes from how and where the beats sync up. In this example, every other high hat matches the downbeat of the kick, but the other high hat beats are in sync with the guitar. The complexity is often layered: lower instruments playing very basic beats, middle ground instruments (such as rhythm guitar) playing interesting, but repetitive beats, and the lead instruments as abstract or complex as the style merits. But it is important to understand that (in metered music) the beats eventually come together in one way or another. These relationships can be expressed as ratios. The high hat and kick are in a 2:1 ratio. The guitar is a 1:1 ratio with the kick, but out of phase. You can pair up more complicated divisions such as 3:2 or 4:3. Even though the beats don't line up as often, it's still music because the important beats match. The 2:1 rhythms are the bread and butter of the music industry, while 3:1 are less common. A 4:3 ratio is pretty rare, but 3:2 can be heard in many popular songs including This Will Be by Natalie Cole, Minute By Minute by the Doobie Brothers, and Kashmir by Led Zeppelin, to mention just a few. Below are illustrations of a simple 3/2 then 4/3. The next example illustrates how the beats of a popular song might fall together. The lower instruments lay a foundation. The vocal part can be very complex, sometimes in sync, sometimes free of the underlying meter. 65

HH Kick

1 2 3 1 2 3 1 | | | | | | | | | | | | | 3 against | | | | | | | | | 2 1 2 1 2 1

HH Kick

1 | | 1

Vocal

2 |

3 |

4 |

| 2

| 3

1 | | 1

2 |

3 |

4 |

| 2

| 3

1 | | 1

4/ 3

P/G

(complex variations) 1 2 3 4 5 6 7 8 9 A B C 1 2 3 4 5 6 7 8 9 A B C 1 | | | | | | | | | | | | | | | | | | | | | | | | | 12:4 (3:1) 1 2 3 1 2 3 1 2 3 1 2 3 1 | | | | | | | | | | | | | 3:2 (with pulse)

K/S

1 |

HH

2 | 1 |

Snare 1 Kick | 1 Cymbal |

2 |

3 |

1 |

4 |

1 | 1 |

2 | 1 |

2 |

3 |

1 | 2:4 (1:2)

4 |

1:4 1 | 1:4 1 | 1:1

Everyone knows about groupings of 9/8 and 6/8, but here is something you might not know. It turns out that pitch works the same way. Pitch is just very fast rhythm. Earlier we saw that non-periodic (noise) waves could produce pitch if a small section of this noise source, any section, was repeated. Even though the wave is very complex, the fact that it is repeated as few as three times creates pitch. The duration of the section that is copied will determine the pitch. If the selection is 1/100th of a second, 100 of them will fit into one second, which becomes 100 Hz. The pitch is the reciprocal of the duration of the loop. Figure 8.1 is an illustration. First, a millisecond of noise, then the noise repeated four times, and finally, a complex wave of an actual sound (piano?) for comparison. Fig. 8-1 Repeated Noise = Piano

Rhythms are in sync when the beats match. Likewise, pitches are in tune when their frequencies match cycle for cycle. If one guitar string is vibrating at 440, the other at 441, they sound (slightly) out of tune. To understand why, consider rhythms that don’t quite match, for example two cars at a stop light with their turn signals going. They may begin in sync, but since one is usually slightly faster it creeps ahead of the other. Eventually they will be completely out of sync, like the kick and snare. Eventually the faster light will catch up and again they will be back in sync. Frequencies that differ slightly are the same. They may begin together, but slowly (actually, quickly) they move completely out of phase, then back into phase. Below is a graph of 66

two frequencies moving in and out of phase. I’ve increased the difference (200 and 210) so that you can see the phasing in a few hundredths of a second. Fig. 8-2 Phase and Constructive Interference

At the beginning of the example the two waves are added together (shown in the bottom track). This is called constructive interference. At about 0.05" they are out of phase and cancel each other out; destructive interference. This alternation between constructive and destructive interference, in and out of phase, creates changes in amplitude. The change in amplitude produce pulses, which are understood by performers as out of tune. An interesting note: you can use these pulses for incredibly precise tuning. The number of pulses per second is equal to the difference in Hz. If two pitches produce 3 beats per a second, they differ by 3 Hz. If one is 500 Hz, then the other has to be 503. This is because they go in and out of phase 3 times in a second.  Since detuning a little seems to improve and thicken a sound, two questions emerge: what is the threshold? When does in tune but phased become out of tune? And what is the threshold between the same pitch, but out of tune, and two different pitches altogether? Our (unscientific) experiments show that 1% is still in tune. 1% to 10% is the same pitch but out of tune. At 10% and greater you hear two different notes. Different can be used to create … •

Harmony and Melody

When any sound, even noise, is repeated it becomes music. When several frequencies are played at the same time (harmony) or linked together in a series (melody) they become musical ideas; melody or chord progressions. If you're keeping score at home: thump = faint, two thumps = wrestling, three thumps = rhythm, fast thumps = pitch, two pitches within 1% of each other = same note, two pitches greater than 10% = harmony and melody. Fig. 8-3 1:1 Ratio

If two frequencies match exactly, we hear them as the same pitch. Using the same system of tics (conveniently set up by the author during the discussion on rhythm) to mark the peak of each wave, we see the two frequencies in Figure 8.3 match in a 1:1 relationship. We hear this as the same note: 67

1: 1

| |

| |

| |

| |

Rhythms that are different but connected by matching beats are considered music. The same is true for melodic and harmonic intervals. Take for example 440 and 880, a 2:1 ratio. Again, the tics mark the peaks of the waves. 2: 1

| |

|

| |

|

| |

|

| |

|

Fig. 8-4 2:1 Ratio

 A 1:1 ratio is the same pitch and a 2:1 ratio is heard as, drum roll, the same pitch! Any musician, even those with abbreviations in front of their names, would say without hesitation that these two frequencies are the same pitch. This is called octave equivalence. That is: two frequencies that differ by any multiple of two are considered equivalent. 261.6 is a C, but so is 523.2, as well as 1046.4. They all belong to the pitch class C. In music discussions it is often irrelevant which C I'm referring to. For example, C, E, G is a chord that could be played anywhere on the piano and we would still call it a C chord. To identify a specific pitch of any class, add a number to indicate the octave, C4 being middle C. All notes between that C and the next C are in the fourth octave: D4, F4, etc. So C4 is a pitch with a specific frequency: 261.6, and C is a pitch class that includes all multiples of 261.6. Fig. 8-5 3:2, 4:3 Ratios

While 2:1 and 1:2 result in the same pitch, different octave, other rhythmic ratios become different pitches, for example, 3:2 and 4:3. Two frequencies that are related by these ratios make up the intervals of a fifth and fourth, respectively. 150 Hz 100 Hz

3: 2

133.3 Hz 100 Hz

4: 3

| |

|

|

| |

| | |

|

| |

| |

|

|

| |

| | |

|

| |

| |

| |

When they are mixed together, and this happens either on the instrument or in the air, their constructive and destructive interference creates another overlaying pattern, the cycle of which is the point where they come back into sync. Figure 8.6 shows the two ratios again. The first two tracks are the individual frequencies; the last channel shows how they add together to create a larger pattern of their combined interference patterns. The first

68

Fig. 8-6 Constructive, Destructive Waves

Our music experience is made up of the patterns that result from the interference of waves. We recognize them as intervals, chords and melodies, use them to tune instruments and as clues for space and environment (reverb and echo). In casual conversation I'm often asked about the nature of music; why do we use the major and minor scales, why are there 12 notes in a scale, how do we account for the black and white key arrangement? Do we accept the system used in Western music because that's what we're used to (nurture), or is there a fundamental principle at work (nature)? What is the universal field theory of music? •

Music is Math

 All the intervals in Western music can be expressed as ratios. Figure 8.7 shows the white keys of the piano. Intervals down are the reciprocal: An octave up is 2:1, an octave down is 1:2. A fifth up is 3:2, down is 2:3. (And the math pans out. A fifth down, 2:3, then octave up 2:1 is 3/2 times 2/1, or 4/3.) The chart below describes the white keys on the piano beginning with the traditional theory terminology of intervals and steps, then the MIDI interval (one piano key per number), a pitch class example, then the ratio. Fig. 8-7 Interval Ratios

Interval P1 M2 M3 P4 +4 P5 M6 M7 P8

Steps unison 1 step 2 steps 2 ½ steps 3 3 ½ steps 4 ½ steps 5 ½ steps 6 steps

MIDI 0 2 4 5 6 7 9 11 12

Example up C to C C to D C to E C to F C to F# C to G C to A C to B C to C

Ratio 1:1 9:8 5:4 4:3 45:32 3:2 5:3 15:8 2:1

Example down C to C C to Bb C to Ab C to G C to Gb C to F C to Eb C to Db C to C

Ratio 1:1 8:9 4:5 3:4 32:45 2:3 3:5 8:15 1:2

 And finally, an easy way to remember 8ve, 5th, 4th, M3rd, m3rd is that the numbers are super particular (first number is one higher than second): 2:1, 3:2, 4:3, 5:4, 6:5. Just remember the jump in the series for a 9:8 major second (what happened to 7:6, and 8:7?). These are the ones I remember. You can easily calculate the other intervals using combinations or inversions. For example, a m6th is M3rd down, then moved up one octave (4:5 * 2:1 = 8:5). A tritone is a M3rd + M2nd. It is more useful, and more logical to order them from lowest ratios to highest ratios: 69

P1 P8 P5 P4 M6 M3 m3 m6 M2 m7 M7 m2 +4

unison 6 steps 3 ½ steps 2 ½ steps 4 ½ steps 2 steps 1 ½ steps 4 steps 1 step 5 steps 5 ½ steps ½ step 3 steps

0 12 7 5 9 4 3 8 2 10 11 1 6

C to C C to C C to G C to F C to A C to E C to Eb C to Ab C to D C to Bb C to B C to Db C to F#

1:1 2:1 3:2 4:3 5:3 5:4 6:5 8:5 9:8 9:5 15:8 16:15 45:32

And finally, theorists, especially 20th C theorists, simplify this system by grouping intervals into classes. That is, a major third up (C up to E) has the same character as major third down (C down to A ), but also the same as a minor sixth up or down. This is not the case in tonal music, since the third up implies the key of C, down it implies A-flat or E-flat. Since most 20th C (art) music is post-tonal, key does not come into play, and only the fundamental character of the interval matters. Also, the major, minor, perfect nomenclature has no bearing, so simple numbers are used (which you already know; they are MIDI intervals). The theory is they all have the same character. Just as all Cs are the same pitch class, all 7s and 5s, up or down, belong to the same interval class. The equivalency intervals up and down obviates values above a tritone. Intervals larger can be identified by the inversion, since they are equivalent: 7 is a 5, 8 is a 4, 9 is 3, 10 is 2, and 11 is 1. Interval classes may seem like a complicated convolution, but they explain more than just the scientific foundation of melodies, chords and intervals. They define the very nature, character, and … •

Quality of Music

The second party question that inevitably follows a discussion of white keys on the piano is this: Why does some music sound happy, bright, cheerful, and others somber, sad, confusing, abstract, dark, even frightening. The answer is the same. (Spoiler alert.) Hard math is frightening, easy math is cheery. Seems too simple, but it's true. The character of a melody and the accompanying chords come from the type of scale used. The clearest examples are major and minor. Music in the minor mode sounds somber, serious, intense (Beethoven’s Fifth Symphony), while major modes sound bright, hopeful, and cheery (Sh’Boom by the Light Brigade). But there are many more modes than just major and minor. Consider these (simplified) examples, which in my opinion, show a clear progression from bright and cheery to dark and menacing, and finally abstract and confusing. In this table I use the 20th C system, or MIDI values. The white keys on the piano are: 0, 2, 4, 5, 7, 9, and 11. Black keys are bold.

70

Scale

Example

Steps

Pentatonic Major Myxolydian/Blues Dorian Minor Phrygian Diminished? Chromatic

Eastern music styles Sh’Boom Theme from Shaft Are you Going to Scarborough Fair Beethoven’s Fifth What to Do Before I’m Dead (Kidney Thieves) The New S**t (Manson) Webern

02479 024579E 024579T 023579T 0 2 3 5 6 8 T (E) 013578T 0235689E 0123456789TE

You will notice that the progression to darker modes includes more black keys (1, 3, 6, 8, and T). In very general terms, more black keys = more dissonant music. Why? Because black keys are harder math. Major and minor thirds are the best example. Minor scales sound darker, more somber, because these two intervals have higher ratios, which mean fewer waves match, which means more destructive interference. The reason This is the New S**t seems menacing right out of the box is the first three notes use ratios of 8:1, then 45:32 and 6:5; a very dissonant, unstable diminished chord. Compare the ratios for the first three chords (from tonic) in Bach’s WTK Book I prelude—a very simple, accessible work—with the opening intervals (the first two are E and 6, the two most dissonant intervals) of Webern's Op 27, interval to interval, since it's not tonal. Bach: 1:1, 5:4, 3:2, 2:1, 5:2; 1:1, 9:8, 6:5, 8:4, 4:3; 8:15, 9:8, 3:2, 9:8, 4:3 Webern: 15:8, 45:32, 4:3, 16:15, 45:32, 3:5, 32:15, 4:3, 6:5, 9:8 With major through Phrygian the number of whole and half steps are the same. The character relies on establishing tonic through repetition, phrase structure and cadence. With more exotic collections, found in the music of Bartok, Crumb, Webern, or Babbitt, one can precisely identify the character of any collection of pitches using … •

Set Theory

Take, for example, C, E, and G (0, 4, and 7). These three pitches make up three intervals: 0 to 4, 0 to 7, and 4 to 7. Each of these intervals has a character, and adds to the character of the whole. To quantify that character, theorists tabulate all intervals into a vector showing 1 to 6. The example above has an interval of 4, and an interval of 7 (but we use its inversion: 5), and an interval of 3 (4 to 7). Interval class: 1s 2s 3s 4s 5s 6s Total: 0 0 1 1 1 0 Except for the 6, these are conveniently ordered by dissonance. An interval 1 is the most dissonant and 5 the most consonant. I think the 6 is more dissonant than the 1, but many students disagree. It is certainly more dissonant than a second. So if you place the 6 between a 1 and 2, they are in order of dissonance. This chord, with the 6 moved to position 2, is 000111. One might say it has a dissonance of 111 units. Compare this to a diminished chord: C, E , G , or 0, 3, 6, where 0 to 3 is a 3, 0 to 6 is a 6, 3 to 6 is a 3. Tallied up, with the 6 to the left, this equals 010200. (One 6 and two 3s.) One might say this has a dissonance level of 10,200 units. More dissonant than the major triad. One of the most dissonant chords in tonal music is a French Augmented sixth chord: C, D F and A , or 0, 2, 6, 8. The intervals from 0 are 2, 6, 8 (becomes 4). From the 2 they are 4 and 6, then from 6 another 2; 2, 6, 4, 2, 4, 6, or 022020. 71

This system can precisely compare the relative dissonance and consonance of any style of music, but in particular, non-tonal styles. What is the most dissonant collection (on a piano)? The chromatic cluster; CCCCCC (in hex—see, I told you we'd use it). The notes are 0 1 2 3 4 5 6 7 8 9 T E. Intervals from 0 are 1, 2, 3, 4, 5, 6, 5, 4, 3, 2, 1. From 1 they are 1, 2, 3, 4, 5, 6, 5, 4, 3, 2. From 2 - 1, 2, 3, 4, 5, 6, 5, 4, 3. From 3 - 1, 2, 3, 4, 5, 6, 5, 4. From 4 - 1, 2, 3, 4, 5, 6, 5. From 5 1, 2, 3, 4, 5, 6. From 6 - 1, 2, 3, 4, 5. From 7 - 1, 2, 3, 4. From 8 - 1, 2, 3. From 9 - 1, 2. And from T - 1. Twelve 1s, 2s, 3s, 4s, etc. We have to use hexadecimal notation to indicate 12 in each column. Figure 8.8 shows a chart with works from three periods. Bach, where dissonance is (to our ears) mild, Chopin's more chromatic works—rich, but still normal to most concert patron's ears—then Webern, who most describe as a cat walking on a piano. These are just the opening chords or in the case of Webern, note groups. Fig. 8-8 Comparative Dissonance

Bach, WTK I, prelude in C C0, E4, G7 000111 2 5 9 0 D , F , A , C 001131 G7, BE, D2 000111 8 0 4 A ,C ,E 000111 2 6 9 0 D , F , A , C 011211 G7, BE, D2 000111 0 4 7 E C , E , G , B 100122 Totals: 112908

Chopin, Op. 28, no 4 E4, G7, BE 000111 4 7 E 0 E ,G ,B ,C 100122 6 9 E 4 F , A , B , E 002103 F 6, A9, BE, E 3 011211 F 6, A9, C0, E 3 020400 F5, A9, BE, E 3 022020 F5, A9, BE, D2 011211 167178

Webern, Op. 25, Pierrot Lunair F 6 , F5 , D2 , G7 , E 3 302111 9 8 1 0 A ,G ,C ,C 200111 E T 7 B ,B ,G 100110 7 4 3 6 1 G ,E ,E ,F ,C 212345 5 2 E T 3 6 F ,D ,B ,B ,E ,F 321333 0 9 8 1 6 5 C ,A ,G ,C ,F ,F 113433 3 2 5 4 E ,D ,F ,E 302100 571543

And it makes sense. We are just counting up all the intervals. (I'm sure my example is riddled with errors.) Webern uses 1s and 6s more often than Bach or Chopin. Chopin uses more dissonant intervals than Bach. It's just that simple: music is math, hard math is unsettling. •

Consonance and Dissonance

… are subjective and relative. Many books define consonance as pitches that sound "good" together and dissonance as pitches that don't sound good together. This is wrong for two reasons. First, it implies intervals must fall into one or the other category. They don't. Secondly, it implies that you should avoid bad sounding dissonance. You shouldn't and we don't. Rather, all intervals fall somewhere along a continuing scale of relative consonance and dissonance, and musicians use intervals all along this continuum to achieve a particular mood. Music students spend four full semesters learning to understand and carefully control the balance of consonance and dissonance in common practice styles. At the end of the first year, a final exam can drop an entire grade if they don't approach and resolve an augmented 4th correctly. A more appropriate definition is suggested by the Latin root consonare: sounding together. Intervals with low ratios have more constructive interference, and the waves coincide more often; they—like rhythms of 1:1 or 2:1—sound together. The waves of dissonant intervals coincide less often, and thus have more waves that sound apart.

72

Fig. 8-9 Patterns of White Key Intervals

Using our ticks once again you can see the patterns (Figure 8.9) for each of these ratios. Note that higher numbered ratios have a longer combined cycle, therefore more time between coincident beats, which is harder to see on the page, and more difficult to hear. Higher ratios are more dissonant and lower ratios are more consonant. The goal of the 12-tone school was precisely to avoid simple relationships, thereby avoiding tonality. It is abstract and complex for a reason; it’s harder math. If the collection of ratios strike a balance that keeps our interest or matches our mood at the time (which changes depending on the context; church, movie theatre, concert hall), the music is effective. If a composer’s goal is to sooth he will choose simple relationships. If she wants to unnerve us, or engage our intellect, then she may try more complex relationships. Even if a composition is not what we expect, it can still be effective, if it achieves the composer's goals through the use of relative dissonance and consonance. The same is true for rhythm; complex rhythms either intrigue or confuse, depending on our ability to focus and the context of the work. Simple rhythms are more visceral. They work well for, say, The Andy Griffith Show theme song; bright, happy, comfortable. Complex or abstract rhythms and intervals are and were used for more intellectual, mysterious moods, such as the Twilight Zone incidental music. When offered a collection of ratios we intuitively look for the simplest combinations, the 1:1s the 2:1s, and 3:2s. Western tonal music works with motion away from stable, or consonant structures, to dissonant structures and back to consonant again. That motion, and the relative consonance/dissonance in infinite variation, mixed with a similar balance in rhythmic simplicity or complexity, is what we find satisfying in the works of Bach, Wagner, Schoenberg, Chick Corea, and Alison Kraus. All styles can be boiled down to motion to and from low ratios and high ratios, consonance and dissonance, structure and prolongation, stability and instability, low on the harmonic scale to high on the harmonic scale. Or as Schenker might put it; tonic (1:1), departure to dominant (3:2) return to tonic (1:1). Complicated? Note compared to… 73

•

Tuning

My party guests are reassured to know that the intervals found on our modern piano are based on such rational scientific principles. So I usually follow up with this unsettling truth; for the majority of musicians it's a lie. I lied in Figure 8.7. We don't use those ratios for the keys on the piano. In Figure 8.9 the white key intervals do not line up so tidily. I recently visited the studios of a world-renowned luthier. He was an eccentric artisan, filled with surprising insights. When we got to the fretted instruments, he made an off-hand comment about how terrible modern guitars are. "What do you mean?" I asked. "Well," he said dismissively, as though the whole world knew why, "you can’t tune them." I already knew this about pianos, but guitar was my instrument. How could I have gotten to a college music program still ignorant of such a fundamental flaw? Turns out all keyed and fretted instruments by their nature—the way they are designed—cannot be tuned correctly, that is, to the ratios we just described. There are not enough keys or frets to represent all the pitches we use today. In addition, while non-fretted or keyed instruments such as the voice or violin can tune to pure ratios they often don't, either because they need to match a keyed or fretted instrument in the ensemble or because over years of acclimation (nurture) they have come to accept the piano as "in tune," which it's not. The best example of this contradiction is choir singing to piano accompaniment. The singers, if correctly trained, tune their notes to the piano, which is incorrect. When singing A Capella, if correctly trained, they will sing slightly different intervals.  I can see you shaking your heads, so let's do the math. We'll start with the frequencies for the black key 8 (used for both an A-flat or G-sharp) using the lowest possible ratios (most consonant or in tune). Starting from C4 (about 261.6) we would go up a minor sixth (8:5) to A . The result is 418.56. A G on the other hand is two major thirds above C (261.6 * 5/4 * 5/4), which comes out to 408.75; not even close to the A . As a matter of fact, these two "enharmonic" pitches, A and G , are nearly a quarter step (41 cents) from each other! Students are always a little suspicious of these results (or they suggest some kind of rounding error), so I'll wait while you dig through the kitchen drawer for a slide rule. Try major seconds down, then an octave up; 261.6 * 8/9 * 8/9, then * 2/1 = 413.4 Hmm… that’s right in the middle. How about a minor third and a fourth: 418.56. How about just whole tones up: 419.03. Yikes! What about half-steps? (438.43) Half steps down? (404.19) Which do we use for the black key 8; 418.56, 408.75, 413.4, 419.03, 438.43, or 404.19? This curious, almost unbelievable warp in music, where logically equivalent but mathematically different intervals will result in different frequencies is called the … •

Pythagorean Comma

Your stunned look is justifiable. I freely admit to relishing the labored emails I exchange with engineers who dabble in music, or musicians who dabble in science, as they struggle with the math. But the explanation is simple; it is because while we use one black key for both A and G , they are different pitches. A cellist or violinist will play them differently. In a recent recording session an internationally renowned producer stopped a well degreed performer and said (paraphrased). "I believe your C was high." The performer wasn’t offended, but he didn’t believe he was wrong. He matter-of-factly said "you are hearing it as a pianist, I’m playing it as a 74

violinist," and then played three slow long notes asking "which C do you want me to play?" That was followed by 15 minutes of very professional, reasonable, polite disagreement. It wasn’t an easy answer. Most violinists spend half their lives getting just one C to sound right. He could do three. But despite experimental pianos with typewriter style keyboards ("Janko"—just Google microtonal keyboard) or split black keys to accommodate each pitch, our modern piano only has one key for both. What do you do? Tune to one or the other? If you tune to the A , then the key of C minor, which uses A , will be in tune. But the key of E major, which uses a G , will be out of tune. Fretted instruments have a similar flaw. The precise position of each note changes depending on the context. Our luthier guide explained that early guitars had movable frets to accommodate such adjustments, but were difficult to manage. Fixed frets began to be used, which raises a similar dilemma. Which note do you set the fret or tune the guitar to? If you tune it to one, the other may sound off. This explains why you can tune your guitar to sound great on a G chord, but an E sounds wrong; it is wrong. Fig. 8-10 Movable Frets

Three centuries ago the music community picked one. The practical result was you could only play in certain keys; those that included the pitch you tuned to (e.g. A , not G ). In this system the farther you stray from those easy keys the more out of tune it sounded. Many musicians, notably the renegade progressive Bach, weren’t satisfied with such a restriction. The compromise they proposed evolved into our modern equal tempered tuning, where the strings are adjusted (the fifths are flat, thirds are high) to be out of tune enough to represent all enharmonic pitches on the keyboard, but in tune enough to sound ok. It took nearly 100 years to be accepted, and the debate continues, but in rarified circles. Bach's masterwork The WellTempered Clavier includes fugues in all 12 major and minor keys to prove the value and to encourage the migration to this tuning. The accordion was most instrumental (pun intended) in worldwide acceptance, because it was cheap and couldn't be tuned. Another practical conclusion is that you cannot trust your ear when tuning a guitar. The frets are set for equal temperament, so that's what you have to use. The fifths have to be slightly flatter than what your ears tell you is correct. How can you make such fine adjustments if … •

Tuning by Ear

For nearly twenty years I included this discussion in my theory and technology classes before a smarty pants student raised this intriguing question: Tuners use stroboscopes to adjust pianos to precise "wrong" frequencies. Modern guitarists use tuners. "What," a student asked, "did they use in Bach’s time, hundreds of years before electricity?" "That's 100 years," I corrected before admitting I had no clue. I should have read the first part of this chapter, where I point out that the difference between 440 Hz and 441 Hz is easily detected by counting the beats as they move in and out of phase. Early tuners calculated the correct number of beats per second for each fine adjustment and tuned so that the beats matched a metronome. 75

So the next time you are at a friend's house, play a few keys on their grand piano and tell them it’s out of tune. When they counter that they just paid someone to tune it you can respond: "I know, and it’s out of tune. The G is way off." But there is a payoff: The acceptance of equal tempered tuning and the simplicity of the keyboard "interface" made chromatic harmonies and enharmonic interpretations more appealing. This allowed Beethoven a greater freedom in harmonic choices, which was expanded by Chopin, taken to tonal extremes by Wagner and Brahms, laying ground for the inevitable leap of logic by the second Viennese school to non-tonal systems. One could argue, therefore, that Bach, the pinnacle of tonal thought, helped plant the seeds of atonality. (Furtive chords.) Should our compositions reflect a distinction between A and G ? Are there renegade, progressive composers who are not satisfied by the constraints of 12 notes per octave? Will it take us 100 years to give in? Is the next step in synthesis a piano that can play both? Yes, yes, probably, and yes: Hermode tuning, which is available in Logic Pro. Computer music is a phenomenal break through in the field of tuning. We can try ideas that we only talked about twenty years ago. The piano, a machine, taught us to play and sing to the equal tempered scale. There’s no telling what these computerized machines can teach us, and where they can take us, as we'll see in future chapters. •

Music is

… the mathematical relationships of constructive and destructive interference resulting from the ratios of sound waves. Intervals, chords, melodies, and even large-scale formal structures, are mathematically related. The musical experience boils down to our search for related patterns in sound, in time. We compare and match the patterns we hear with those in our memory; from the micro level of 1000th of a second ago to the macro level of days and years. When we hear two or three frequencies in a row we not only recognize each cycle of a wave as a periodic pattern, and together their relationship with each other as a pattern (a set of ratios that become a melody), but we match that sequence of patterns with memories of sets of frequencies (melodies) from as far back as our childhood.

76

• Exercises 8:1 Calculate the dissonance vectors for these segments of Wagner's Tristan und Isolde, Steely Dan's Josie and Marius Constant's Twilight Zone. The notes are: [F, B, D , G ], [F, B, D , A], [E, G , D, A ], [E, G , D, B] for Tristan [F, E, G, C, E], [F , C , A , E, A], [B, D, A, D, F ], [G , C, F, G]—Josie [F , G , C ], [D, G, C], [A, C, G], [G , A, B , E ]—Twilight Zone

8:2 Using Amadeus, generate sine tones (in left and right channels) at the ratios for an octave, 5th, 4th, and third. Convert the two tracks to mono to see the resulting interference waves. 8:3 Using Amadeus, generate sine tones in left and right channels at the consecutive intervals: an equal tempered fifth, then just fifth, equal tempered major third, then just major third. Can you hear the difference? 8:4 Open a MIDI version of a fugue from the Well Tempered Clavier in Logic Pro. Choose a "dark" major key, one with lots of accidentals. Open Project Settings/Tuning. Explore different tunings. Find a tuning that was used in 1722. Switch between it and equal temperament. Which do you prefer? Try Hermode tuning. Can you hear a difference? Split them into separate tracks, each with a tuner plug-in. Note the differences. 8:5 With the same fugue, open the piano roll window and select the first four bars. Choose Functions/ Transform/ Random Pitch. (We will use this as a template, but be sure to try the random pitch algorithm too.) Uncheck Hide Unused Parameters. Select All under Pitch then select the seventh step of the scale for each octave in use, (if the work is in B major, choose A 3, A 4, A 5), and select fix to change them to flatted sevenths. Repeat for each four bars, changing each step of the scale so that the original major mode transforms into five or six of the other modes. Bounce the results. 8:6 Bounce the exercise above in two of the tunings. Open them both in Amadeus and copy/paste over so that you hear both tunings together. 8:7 Beginning at A 440, calculate each pitch in Webern's Op. 25, no. 1 using pure ratios. For chords, start with the bottom pitch and move up the chord. Compare the results with equal tempered pitches. Here are the pitches: F , F, D, G, E , D, G , C , C, B, B

77

8:8 Run the following example in SuperCollider, which realizes a free-just Frere Jacques. Replace the array r with the ratios for the Webern above. ( f = 440; r = [1, 9/8, 9/8, 4/5]; Task({50.do({arg i; i.round(0.1).postln; f = f * r.wrapAt(i); f.round(0.1).postln; (freq: f, tempo: 2, legato: 0.3).play; 1.wait;}) }); )

78

9. Mix: Style, Dimension The elements of a mix. Popular styles first. Programs such as Logic Pro and Garage Band have made digital recording more accessible than ever. Some argue that while earlier recording artists passed through a natural filter of inaccessibility, the best rising to the top, today's crop of musicians can merely noodle around on an electronic keyboard, bounce the file and print the manuscript then post it online. Removing the struggle, the argument goes, has to lower the quality of music produced. •

Music, and Machines

But remember the piano is a machine. It made music more accessible by simplifying the representation (such as a single key for A and G ) and the tuning (which is such a struggle for string players). It is such a successful interface that any self respecting 3 year old can hammer out Mozart. (Even other species—you've seen the videos—can plunk at the keys.) I believe it's true that this simplification has produced billions of mediocre musicians, but it has also produced millions of excellent pianists, thousands of superb pianists, and certainly more true masters than in any other period of history. Furthermore, this machine has inspired composers to take music to new horizons. Subject and Reflection, by Bartok, was written expressly for the piano keyboard interface. In fact, the 12-tone school grew out of the enharmonic tuning, which inspired composers to think outside the box of one key and closely related modulations. Logic Pro, Max/MSP, SuperCollider, Cubase and Reson are rife with interesting sounds and clever loops. It's easier than ever to sound good quickly. This rise of common experience and mediocrity makes it all the more important to distinguish yourself. Noodling is not enough. When performers of traditional instruments audition for a position you can assume they will play correct notes. That's not enough either, and it’s not what committees look for. Performers distinguish themselves and make their music memorable through technique, treatment, and interpretation. The same is true for electronic and popular styles. For example, the engineering on Bone Machine (Waits) is very distinct. It sounds like it was recorded with a single cheap microphone set in the middle of a living room using instruments from a junk yard or pawn shop. But so is the engineering on Steely Dan's Aja; layered with carefully crafted reverbs, delays, panning, and EQs. Norah Jones’ (anything) is natural, transparent, simply beautiful. The engineering on Marc Broussard's work is tight, even, punchy, perfect for cutting through a noisy bar or the chaos of a road trip. Though very different, I love all of these styles, because they capture the artists so well. Like a skilled photographer, these engineers frame, filter, mat, light, focus, and compose their subjects with a wide range of techniques. Electronic composition benefits from the same attention to detail, beginning with… •

Mix Styles

 There are four styles, each named for the region where they evolved; LA (natural, transparent), New York (aggressive, tight, even), London (layered with fx and digital sophistry), and finally Garage Band (raw). You can certainly do an LA style in London or Garage style in downtown NY or any of them in your basement. 79

Examples of New York style include ZZ Top, Dixie Chicks, (as well as most modern rock/country), Kidney Thieves, Daft Punk, Reel Big Fish (and most Ska), and The Pretenders. LA Style artists include Doobie Brothers, Norah Jones, Steely Dan, Willey Nelson, Eagles, Ricky Lee Jones, and pretty much everything I do (in particular, classical recordings). London includes, of course, most of the post Revolver Beatles, many of Kidney Thieves tracks (which is a blend of NY and London), Hendrix (Experienced), Kidjo (Voodoo Child Return), and many cuts from Zeppelin. Garage style: Waits, Dylan, and a lot of the Stones.  We will narrow in on each style while discussing the techniques below, but one of the best ways to distinguish each style is by examining the role of the engineer. In LA style the engineer is transparent. In New York style he is a craftsman. For London style the engineer is a collaborative artist, working along with the musicians. (The first, and quintessential example of London style is George Emerick, who engineered Sgt. Pepper.) In Garage, the engineer is absent. •

Mix Dimensions and Elements

… describe the position and role each of the instruments has in the whole. Dimensions of a mix are height, width, and depth. These describe the virtual space created with frequency's ranges (height), pan position (width), and reverb and delay (depth). Absence of one or more of these three is the single most common error in a student's portfolio submission. The elements of a mix have a parallel in orchestration, where we talk about background, middle ground, and foreground. There are similar components to a good mix. The background or foundation is made up of instruments in the lower register, but more precisely repetitive simple components: the drums, bass, possibly guitar or piano. They lay a foundation on which other musicians build. The goal of a foundation is not to draw attention, but to provide a solid bed of simple, clear rhythms out of which the more interesting elements can grow. In the middle ground there are two elements with diametrically opposed characters. The first is a rhythm. This is typically a keyboard or guitar playing an interesting, but repetitive idea. It is interesting enough to draw your attention, but repetitive enough not to demand complete focus. It is often the first catchy hook with which the song is identified. The guitar parts in The Doobie Brother's Listen to the Music, China Grove, Long Train Running, and Dependin’ on You, as well as ZZ Top's Tush, Eric Clapton's Cocaine, and Atlanta Rhythm Section's Spooky, She Knows All My Tricks, and Voodoo are all good examples of a rhythm. It doesn't have to be the guitar. In Stevie Wonder's work it is, of course, the keyboards, the tambourine in the Beatle's We Can Work it Out. Sometimes it's a subtle offbeat snap of a guitar chord, as in Patsy Cline's Crazy, or Elvin Bishop's Fooled Around and Fell in Love, or the guitar harmonics and the riff in Buffalo Springfield's For What it's Worth. To summarize, a rhythm is interesting but repetitious. The other middle-ground element is the pad; a cohesive, smooth sustained pitch, chord, or series of chords that hold together and blend other components. The most obvious example, almost cliché in blues, is an organ (vocals or strings) playing quiet sustained chords. The organ in Norah Jones’ Turn Me On, second verse, Bob Marley’s I Shot the Sheriff, second verse, Tracy Chapman's Give Me One Reason, and Willie Nelson’s Stardust, are good examples. The foreground elements are the lead (usually pretty obvious), and the fill. These are both interesting, melodic, varied (not repetitive), and share the center of focus. The fill is as interesting as the lead, but fills the gaps where the lead is not active. 80

 To summarize: Lead and Fill are interesting, melodic, inventive, and share the foreground. Rhythm and Pad share the middle ground. Rhythm is interesting, but repetitive. Pad is sustained and cohesive. Everything else is background and foundation. Students seem to struggle most with the rhythm. I offer these suggestions: first, not all examples are so cut and dried. An instrument may be a blend of several elements, or switch roles in the middle of the tune, beginning as a rhythm, then evolving into a fill. Second, the key is repetition and interest. No matter how interesting, if it repeats, it is a rhythm. We can relegate it to our sub-conscious and focus on the lead and fills. Life in the Fast Lane, Eagles; Don't Stop to Watch the Wheels, Doobies; Walk This Way, Aerosmith; Nearly Everything, Police; all contain intricate guitar lines, but because of the repetition, they are rhythms. Likewise, elements of a mix, though taught in the context of pop engineering, are often what distinguish quality electro-acoustic compositions. Young composers tend to err to one extreme or the other; either all middle-ground, repetitive elements with no clearly focused lead, or all foreground material, fighting with each other for focus. You don’t have to have all five elements. Elements can also shift between one role to another. To better familiarize yourself with these components, listen to your favorite music; try to place each instrument in one of the categories. You will soon be surprised how easy they are to identify, how effective they are in their roles, and how common they are. Less with elements, but often with style, there are gradients. Students always disagree, for example, about Kidney Thieves, Blood Sweat, and Tears, and Steely Dan. They have strong elements of at least two styles. Bob Marley is a convincing mix of all three. We can further characterize these styles by how they use each of the following. •

EQ and Filtering (Height)

Fig. 9-1 Channel EQ

… is built into every channel strip because it is assumed that every channel will require EQ. The EQ window, when double clicked, will add an EQ plug-in. You can add more, or different EQs by clicking on an empty insert slot.  The three parameters of an EQ are frequency cutoff, gain, and Q. Cutoff is where in the range of frequencies the EQ is applied. With roll offs and shelves, all frequencies above or below that cutoff are reduced or increased. With notch or peak filters, frequencies above and below the cutoff are filtered. Gain is how much the frequency is increased or decreased. Q is the width of the band of frequencies filtered. (Other apps such as Logic Pro instruments or SC patches use the terms rq and resonance.) We will use these terms in later chapters on synthesis. The Q is a little counterintuitive in that a higher number is a narrow band of frequencies, a lower number results in a wider band. The EQ in Figure 9.1 shows a low roll off at 57 Hz, a broad-band (0.37 Q) boost at 186 Hz, a narrow band (2.10 Q) notch cut at 2000Hz, and a high shelf boost. The EQ interface is intuitive. Click inside the shaded area of a band and drag left or right, up or down to change frequency and gain. To adjust the Q, click on the small circle that 81

appears when the mouse hovers over the cutoff frequency. You can also click on the numbers below the graph and drag them, or double-click to type in a precise value. I typically use the gain to boost the band, which makes it easier to grab in the graphic interface. 

Some general suggestions about EQ: • Bypass the EQ often to remind yourself what you started with. • Remember that when you are boosting frequencies you are increasing the gain, the same as you would with the volume slider. Adjust the levels to avoid saturation. • Start with the presets. Many engineers dismiss presets, arguing that every instrument is different, and any preset will miss nuance. But the presets aren't random. Each is a private EQ workshop with a professional. They are certainly worth trying, and a good starting point. • Compare with recordings you know and try to match the sound. • Listening fatigue can skew your judgment. Take breaks. Trust your first impression after a break or the next day. • Make small adjustments to improve the quality of a sound (e.g., for an LA style). Large adjustments for effect (a London style). Cut deep and narrow bands to eliminate noise. • Always be aware that when you are removing unwanted sounds (hiss or hum), you are also removing sounds you probably want (depth, crispness). EQ should be a last resort for fixing signal to noise. Get it right in the first place. • When adjusting the EQ of two instruments with similar ranges (two guitars), consider complimentary adjustment; boost in one the frequencies you cut in the other, and visa versa, so they can occupy different ranges of the spectrum.



Here are descriptions, reasons to boost, and reasons to cut each octave band.

0 to 25 Hz: "Rumble" frequencies provide special effects such as thunder, helicopter thrash, explosions, and flutter. In music mixes there's not much reason to boost them. Rather, they are a potential surreptitious source of noise. Most monitoring systems won't reproduce these frequencies, but many home systems do, so you should mix with that in mind. There could be a 5 Hz rumble from an equipment room or AC that you'll never hear in a pair of headphones, but your client with a 5.1 system will. Cut or "roll off" the final mix at about 20 Hz as a standard mastering procedure. Fig. 9-2 Clean Sine Wave From Low Roll Off

25 to 50 Hz: "Chest thump" frequencies are what you feel in your body with a solid NY style rock mix. Boost them only in your bottom instruments; kick, toms, and other drums. Cut them in low melodic instruments (such as a bass) to bring out the higher overtones for better recognition. Bleed from other tracks (through an isolation booth wall or in the same room) will interfere with the tracks you do want down there. For that reason, roll off all tracks that you don't want in this range, even if they are incapable of producing frequencies this low.

82

A solid low end can be seen in the waveform in the audio graph. Figure 9.2 shows a pop tune zoomed in far enough to see individual waves. The low frequency sine wave is very clear; evidence that there are no phasing issues. 50 to 100: "Foundation" frequencies come from the drum set, bass guitar and piano. Boost to fatten or warm up the sound. Cut if it sounds boomy, or to remove ground hum (60 Hz). This will increase the clarity, even for instruments that can play below that. 100 to 200: "Warmth" frequencies come from the lower end of rhythm instruments such as the guitar and piano, but also toms and snares. Boost to add fullness to guitars but also vocals. Cut to reduce muddiness and bring out the lead instruments. 200 to 400: "Rhythm" frequencies make up the natural range of most rhythm instruments. I try to keep this octave as flat as possible. Boost for a punchier bass or lower drums. Cut to eliminate overbearing frequencies or feedback issues in live PA using the boost/sweep/cut procedure covered below. 400 to 800: "Lead" frequencies make up the natural range of most melodic instruments. Boost to bring out the lead but also to sharpen the attack of lower instruments. Cut to eliminate overbearing frequencies or feedback issues. 800 to 1600: "Harmonic" frequencies are above most melodic lines, but provide the harmonics of those instruments. They are what define the timbre of the lead. Vowels and diphthongs are shaped in this range. Manipulate this octave to change the character of an instrument, either adjust for a more natural sound, or increase for a (London style) megaphone or cheap radio effect. Sweep through with a narrow Q for a wa-wa effect. Boost for unusual effects. Cut if an instrument sounds cheap or boxy. Fig. 9-3 Sonogram Of R, K

1.6k to 3.2k: "Consonant" frequencies are the k's, t's, ch's in speech, but also the attack of most instruments. Boost for clearer lyrics and crisper guitar strumming. Cut to reduce sharpness or edge and listener fatigue. This is also a common range for feedback. 3.2k to 6.4k: "Edge" frequencies provide s and sh sounds as well as cymbals and shakers. Boost for clarity and brilliance. Cut also to reduce listener fatigue or to soften a mix. This is the range of hiss from older tape machines, but cutting it will also take the edge off the mix. 6.4k to 12.8k: "Sizzle" is air and light cymbals, but also can be piercing and painful. Boost for metallic sounds but also to liven up a dull vocal. Increasing slightly will lighten the strum of a guitar. It will also liven up a reverb. Cut to reduce noise (hiss), but only in spoken recordings, or to take off the edge and avoid listener fatigue.  25 Thump

To review: 50 100 Foundation Warmth

200 Rhythm

400 Lead

800 Harmonics

1.6k 3.2k Consonants Edge

6.4k Sizzle

A common application of EQ is to reduce feedback problems in live PA. There is a law in California requiring feedback on every PA. I know this because whenever a Hollywood hero 83

steps up to a mic to make an embarrassing, but pivotal speech, the PA squeals. Directors can't seem to resist this cliché to make the moment more uncomfortable. These movie engineers are timid, afraid of feedback, and simply hope it won't happen. But problem frequencies need to be confronted–flushed out and eradicated long before the event—using … •

Boost, Sweep, Cut

Boost the gain of a single mic until it is on the verge of feedback. When the resonance begins pull it back just a little. Next, use a narrow band EQ and sweep it slowly through the general range of the offensive tone. When you hit it, it will pop out like bad tooth being tapped by a dentist. Cut by dragging the band down. Rinse and repeat until you've found all the resonant frequencies in the hall. Not only will this reduce the chance of feedback, but also you've compensated for the harmonic character of the hall, resulting in a more even frequency response. If you don't have a nice graphic interface (for example the selectable EQ on a mixer), and if you have a musical instrument handy, you can use it to fish around for the pitch that emerges from the sweep, then (assuming you've read Chapter 5 and can calculate frequency from pitch class) reduce that frequency. We work with a few pianist who, at the first hint of a feedback ring, will quickly find it on the piano and yell back the pitch so we can calculate the frequency. Use the same technique to find annoying frequencies in a mix; a boomy guitar, nasally vocal, honky saxophone. •

Noise Reduction Algorithms

… are the holy grail of plug-ins; something that will remove the hum and hiss of a low quality recording. These work by first reading a sample of what you consider noise, analyzing the frequencies and amplitudes, then applying gates on selective bands. They do well with speech, but I've never had success with music. They create too many artifacts. I remain hopeful, but dubious. •

Panning (width)

… can be adjusted with the knob just above the volume slider. It's name—Pan/Balance Control—reveals a flaw. Pan and balance are different, and it is confusing to use both terms for one knob. This is because on a mono track pan and balance are essentially the same. But with a stereo track they are very different. When a mono signal is routed to a stereo output, it is split equally in both channels. (You can replicate the effect manually by duplicating a mono track and sending one to output 1, the other to output 2.) This effectively places it in the center of the field because we hear it at the same level in both ears. Turning the balance knob to the left turns down the volume in the right channel. The aural effect however is a pan; moving the instrument to the left or right in the stereo field.  But on a stereo track the information in the left and right channels may be and probably is different. In the case of a stereo drum set, the toms to the left and the hi-hat to the right. The balance control in this situation will turn down the right or left channel. The effect is not a pan, but a balance. Moving the knob all the way to the right will turn the left channel off. You will lose the toms.

84

Fig. 9-4 Pan, Balance, Directional Mixer

For true panning on a stereo track or sub-master you have to add a directional mixer as an insert. Turning the knob to the left will not simply reduce the amplitude of the hi-hat (for example) in the right; it will increase it in the left, effectively moving it in the stereo image.  To summarize: the default knob effectively functions as a pan on a mono track, but as a balance mixer on a stereo track. A directional mixer must be inserted on a stereo track for true stereo panning. Most of the engineering I do is live classical performances. The stereo image is a reproduction of the actual positions of the performers. Pan never comes into play. Even so, I've gathered some guidelines about generating an image through panning in multi-track recordings. Again, I encourage you to listen and analyze your own examples. • Extreme image placement and real-time change of placement is common in London style mixes. For both New York and California the panning should be more evenly distributed and never change. Spatial effects, especially dynamic effects, should always be a part of electronic composition. It will set your work apart. • Your individual tracks may be either mono or stereo, depending on the source. Even if stereo, consider panning (with a directional mixer) to one side or the other to avoid big mono. • In California style you would place the instruments where you might expect them on stage. For a jazz group the piano on the right, the bass on the left, the drums a little off center, horns behind the piano, and so on. For a New York mix, consider panning based on the roll they play in the mix. Place the lead and fill a little off center opposite each other. Place the guitar and piano opposite, as well as the bass guitar and bass drum. For a horn section, (though back and to the right in an LA mix), you might pyramid with the lead in the center, horn II and III slightly left and right, IV and V harder left and right, as they often provide support. Fig. 9-5 Gain Control, Mono

• Pan position changes the blend of a channel and therefore can improve or worsen phasing. Phasing issues are more evident in mono. For that reason, consider fine-tuning the pan position in mono. Yes, you read correctly; pan in mono. Place a gain control on the master track, select mono, then make small adjustments to improve the balance, listening to see if something drops out or pops out. • Pan overdubs either both center (vocal), or both hard left and right (rhythms). •

Compression

… controls volume. The goal of most compression is uniformity. Two typical applications are vocal and bass tracks. Basses, especially upright, are notoriously inconsistent in attack and level. Even the best performers struggle with dynamic control because of the resonant nature of the body of the instrument. This is also the case with inexperienced vocals. Compression can be used to correct both.

85

It was first used in radio to protect equipment and avoid distortion by reducing amplitudes that would exceed the available threshold. Artists heard their music over AM radios, post compression, and liked the sound. The uniform amplitude cut through the small speakers and noisy environments. Geoff Emerick was the first engineer to intentionally apply compression to drums to alter the character of the set (A Day in the Life). It is now standard. Fig. 9-6 Compression and makeup gain

 The four main parameters are threshold (the level, in dB, where compression is applied), ratio (the amount the amplitude is reduced), attack (how quickly compression is engaged) and release (the time it takes to return to normal). Figure 9.6 shows three audio waves. The first track has three soft notes, one loud note, then three medium range notes. The first three pitches are at about -9 dB. The fourth note is -3; a difference of 6 dB. The goal is to first bring the louder amplitudes down to the same level as the lower volumes, then raise all levels. To accomplish this you would set the compression threshold at -9 dB, so as not to change the first three notes and only one or two of the last four in this example. The second image shows the results of a -12 dB threshold with a light compression ratio (3:1). Notice the first three notes have not changed, but the loudest note has been reduced. The last four have been modified slightly. With the additional headroom you can raise the level of the entire track, shown in the third example. Most compressors also have a control for make up gain. Compression is used extensively in New York and creatively in London style, sparingly in LA style. It is common to compress every channel, maybe slightly, and often the sub-masters, and masters in a NY mix. A snare drum, for example, through a handful of sub-mixes, may have compression applied eight times before it reaches your ears. Heavily compressed styles are great for the car, dance venues, and sports events, because the amplitude needs to be consistent to better cut through competing noise. But dynamic change is an important expressive tool in LA style, and even more important in classical programs. Rampant, knee-jerk compression is often criminal and painful to listen to. Keep in mind that dynamic range gives art music expressive power. Compression flattens that dimension. A century of technological progress and yet popular recordings have less dynamic range than an Edison phonograph cylinder of 1909. There are two cases where I will use compression on classical performances; to squash Mr. enthusiastic concert goer, eager to demonstrate his erudition by bolting into an obnoxious applause before that last beautiful note fades in the hall (did that sound sarcastic?), and to lower ever so slightly operatic vocalists, who are notorious for singing an entire concert around -12 dB, then nailing one note at -2 dB, defining the limits of the entire recording. Figure 9.7 shows three examples of differing levels of compression. The first is a NY style (She's Got a Girlfriend Now, Reel Big Fish), then LA (Stardust, Willie Nelson), then a classical program (Kantate Nr. 2, op. 31, Webern). The first is in your face all the time. The 86

second is consistent, but expressive, the last certainly has loud spots, but very, very soft spots too, which is the whole point of most classical work. Fig. 9-7 NY, LA, Classical Styles

Other common applications: • To increase the sustain of a guitar or snare. By compressing the first and loudest part of the attack you are then able to raise the decay substantially. In effect you are turning the volume down just as the note is played, then raising it as it dies away. (For this effect you must have a fast release.) This is one of the techniques (including riding the faders) used for the final "infinite chord" on A Day in the Life, Beatles. (Listen carefully for Ringo's shoes creaking in the right channel near the end.) • In PA work, especially political speeches, compression is apparently mandatory. It usually raises the noise floor. • The commentary in sporting events is compressed with a fast attack and slow release. This isolates nicely the color commentary while maintaining the sports ambiance. When the announcers stop speaking, you hear the crowd behind them, then they disappear at the start of the commentators next sentence. • "Ducking" is a type of compression where the volume of a background musical track is lowered whenever the voice-over is present. (Covered in more depth under side-chains.) •

Reverbs and Delays

… account for the depth of a mix. It is the second most common omission (after image) by inexperienced composers and engineers.

87

Fig. 9-8 Reverb

 There are three key parameters in a delay; delay time, feedback, and mix. The delay time is set in milliseconds. Feedback controls how many repetitions there are before the echo dies away. (You can increase this until the delays perpetually build on themselves rather than decrease. Combine this with an EQ to carefully control the evolving echoes.) The mix controls the blend between the original ("dry") signal and the delay ("wet"). It is either a single control labeled mix, or two controls labeled dry and wet. Delays are more effective in stereo. Selecting the plug-in Stereo Delay-mono to stereo, will split the source into two channels so that they can be treated differently, for example slightly different delay times in each channel. •

Doubling Delay

Fig. 9-9 Slap Echo, Doubling

… creates an effect similar to overdubbing. This is intended for a mono track. Leave it panned center. Add a stereo delay (mono → stereo). Turn Beat Sync off. Turn the right output mix to 0% (no delay) and the left mix to 100% (all delay). Set the left delay very short, say 80 ms, which produces a doubling of the original track. Increase the feedback for a similar early Beatles treatment. (Lovely Rita Meter Maid, A Day in the Life, Being for the Benefit of Mr. Kite, Beatles, but also Still Rock & Roll to Me, Billy Joel.) Alternate method: leave both output mix levels at 50% and set left delay to 50 ms, right to 80 ms. This places the original in the center and the two doubling delays left and right. Once you've settled on the effect you like, adjust the location within the stereo field using a Directional Mixer as described above. •

Synced Delays

… match the tempo of the song. For example, there is a sixteenth-note triplet delay (on the "sh-k") at the beginning of Come Together.  Since tempo is expressed in beats per minute, but delay time is set in milliseconds, getting the correct delay requires some math. To calculate the duration of each beat, divide the tempo by 60: 60/bpm. Multiply that by 1000 to convert to milliseconds. Each beat of a tempo at 120 bpm (60/120 * 1000) is 500 millisecond. This doesn't mean you have to set the delay to 500 ms, but good choices would be multiples of that; 250 (for an eighth note delay), 125 (sixteenth). 88

The tempo of Come Together is 82 bpm. Eighth notes would be at 164. Sixteenth triplets (multiply by 3) would be 492 bpm. 60/492 * 1000 = 122. So that delay was set to 122ms. The formula simplifies to 60000/bpm = ms delay. You would expect Apple to include a method to sync a delay to a tempo automatically, and of course they have. To use it, the session needs to have the correct tempo. If you recorded the session using Logic's metronome, then the tempo is already correct. Most musicians prefer to work without a metronome. In this case, the first step for a synched delay is to set the tempo of the session. This is done in the transport bar by clicking the default 120 and dragging up or down, or double clicking and entering a value. Next, in the delay select the Beat Sync button. You can now select quarter, half, eighth, or sixteenth. You can also adjust the Groove to deviate from the exact tempo. I never have two delays exactly the same, adjusting one groove slightly off from the other (52.3% in the example). Note there is no option for a sixteenth triplet, so to replicate the Beatles effect you're just going to have to do the math. Synced delays are used in NY and LA style, to add a pulse that breaths with the beat. In London Style use… •

Asynchronous Delays

… to create complex effects that stand out. Crazy and Arsenal by Kidney Thieves are both good examples. Delay Designer has so many modulation options it transcends delay (as a tool to add depth) into the realm of modulation. The Warped collection (where I always start) should keep you busy for hours. •

Chorus

… performs a similar function to a doubling or slap delay. The signal is duplicated then detuned or delayed by a small enough amount that the two blend as if doubled. The sound is very homogeneous, maybe too much so, since it is a precise duplication. •

Serial Routing

… describes an effect that is inserted directly into the signal flow. A good example of serial routing is an electric guitar with pedals. The guitar plugs into one unit, then out of it into another, then another, then another. If any item in the chain lost power you would lose all signal. Everything we've done so far has been serial. Since all of the signal passes through serial routing, all the signal is processed. That is why some effects give you the option to mix dry signal with wet. EQs and compression process the entire signal, and cannot be blended with dry signal in a serial path. Fig. 9-10 Channel Strip, Serial Signal Flow

 Since the signal is processed in series, order matters. Should you apply EQ after the reverb? Or apply reverb to the EQ? If you place the reverb after the compressor, then you are reverberating a compressed signal. If you place it after, then you are compressing the reverb, and it will change the decay. In logic, signal flow is from top to bottom. The channel strip in Figure 9.10 has a Tuner first (which is disabled), then a Noise Gate, Flanger, BitCrusher, EQ, then compressor. After the inserts the signal passes through the level control and pan. When a new plug-in is inserted, Logic virtually disconnects the cable between 89

two units, places a new one between them, then plugs the output of the one above into the one below, then that out into the one below it. Figure 9.11 shows what the channel would look like if it were laid out on stage like a guitarists set of pedals: Fig. 9-11 Signal Flow

>

>

>

>

>

>

>

>

 Many engineers place an EQ before and after a set of DSP inserts to adjust the original signal and adjust after DSP. Compressing after a reverb will exaggerate the length of the decay. As a general rule: Amp models, flangers, phasers, complex filters and modulation (that are intended to create a sound rather than adjust it) first, then EQ, compression, delays, reverbs. •

Exercises 9:1 Begin with a source that is rich in upper harmonics (live or recorded). Add a delay and increase the feedback until it perpetuates, then pull it down slightly. Add an EQ and boost a narrow band, sweeping for frequencies. Adjust both the feedback and EQ to "play" the evolving patterns. 9:2 Start with a four-track session (stereo drum set, bass, and two rhythm or lead instruments). On one rhythm instrument, pan slightly right; add a slap delay with dry signal in one channel and the delay in the other. On the other rhythm instrument, pan left with a sync'd delay, dry in one channel, wet in the other. Apply a complimentary EQ to these two. Add a reverb to the drums. Adjust the EQ of the drums and bass. Add a compressor to the bass and drum set. Add a compressor, gain, and multi-meter to the master channel. Check the mix in mono and in reversed phase. Demonstrate "before" and "after" with automation; begin with all your modifications off (pan center, level at 0, fx bypassed), then turn them all on at once. 9:3 Begin with a simple five or six track dry mix. Mix in both LA and London styles. 9:4 Using examples in Chapter 12 as a template, provide an analysis of your 5 favorite mixes. 9:5 From your music collection, list work, composer, and minutes and seconds (if appropriate) of five examples each: NY style, LA style, London Style, Fill, Rhythm, Pad, and overdubbing. Find one example each of a slap delay, synced delay, asynchronous delay, and dynamic panning.

90

10.

Auxiliaries Sub-mixes, sends, and headphones. Post and pre-fader.

When preparing for a remote recording I go through a mental checklist that links outputs with inputs, tracing the audio all the way from the instrument to my ears. This helps me pack all the correct gear: Piano → microphone (clip, stand) → cable from mic to house patch bay (do I need a gender adapter?) → patch bay in the control room → patch bay cables to the preamp or mixer (power for the mixer) → cable from the mixer to the laptop (400 to 800 adapter?) → power for the laptop → headphones from the laptop to my ears (ears to brain, brain to frontal cortex, etc). This is a very common… •

Signal Chain

When there is a problem, say a hum in the left channel (which I had just last night), you need to locate the weak link in the chain using a swap and follow strategy. At any point in the path you can swap another signal chain (if stereo, left w/ right). If the noise moves with the swap (i.e., it was in your left ear, now it's in your right), it is somewhere between that point and the mic. If it does not move, it's between that point and your ears. For example, you have a stereo pair connected to a stereo cable, white for left, red for right (red is always right) plugged into the mixer. You hear noise in the left ear (the white cable). At the mixer, switch the white cable to channel 2 and switch the red cable to channel 1. If it's now in your right ear, the noise is the cable, mic, or something in the piano. If it is still in your left ear, it’s the preamp, channel, a dirty pot, your headphone, in the room, or your ear (no kidding). With this particular hum I first tried turning around the headphones, that is, swapping left and right. (Rule of thumb: the headphone cable is always the left signal.) It stayed in the left ear. That meant it was after the headphones—either my ears or in the room. It was from a candy machine through the wall. If you're working with a lot of mono channels, replacing components or choosing different patch-points will give you the same information. Digital audio workstations also have a signal flow, though more illusive since you can't physically trace it down cables. Each audio channel strip (and plug-in for that matter) has in input and output. The top slot of the I/O for audio tracks allows you to choose an input source from the interface, which then connects to a real mic on the floor. The default output for each track is stereo out. Since the output channel is the final destination before it is sent to your interface and headphones, it has no choices for input or output. Instrument tracks also have an input, but the connection is to one of Logic's built-in synthesizers—just as you would connect a keyboard into the line input of a mixer. It is often useful to arrange tracks in… •

Groups

A drum set may have 5 to 7 mics. You can adjust the level of each one, but what if you want to bring them all down proportionally? It would also be useful to mute them all at once, or solo them all. One method of modifying a group of tracks is to first select them using click drag 91

or shift click. All changes will be applied across the selection. This is temporary. They are grouped only as long as they are selected. The group feature creates a permanent group. The group select window is between the track name (e.g., Audio 1) and the automation (default set to off). Select all the tracks you want grouped and click in this window to define a group and choose what actions are applied as a group. ⌘-G toggles the group on and off ("clutch") if you want to modify only one setting in a group. Truth is, I rarely use groups because … •

Busses and Auxiliary Routing

 … give the same functionality plus allow more efficient use of serial effects. Busses are virtual patch cables behind the user interface. They are not assigned to any input or output by default and are meaningless until assigned. There are three applications of auxiliaries; monitoring, parallel effects, and… •

Sub-masters

… where you send a group of channels to an intermediary mix between the original track and the final Stereo Output. In a single session you may have 7 mics on a single drum set. You can adjust the gain on each track, and use groups or select to adjust several. This is also true for background elements; drums, bass, guitar, maybe piano. But how would you add an effect (compression, reverb, delay) to just those tracks? The solution is routing these tracks to a sub-mix. To insert a sub-mix you have to disconnect the output from Stereo Out and reroute it to a bus. Which bus? It doesn't matter. It's just a cable. Select all the tracks you want mixed together and click on the output (Stereo Out) of any selected track and use the pull down menu to select bus, then any bus number. Choose 47 to reinforce that it doesn't matter. (Of course you should eventually order them logically. I do auxiliaries subs on 1-16, reverbs and delays 17-32, and above that monitoring. But it really doesn't matter.) This single action has several results. It will connect the output of all the selected channels to bus 47. It will create a new auxiliary channel, which will appear near the right of the mix window. The input of that auxiliary will automatically be set to 47, essentially connecting the output of all selected tracks to the input of this new auxiliary track. The output of the auxiliary track also defaults to Stereo Out. You now have a sub-mix. More channels can be added to this sub-mix at any time, and you can change the bus it uses or output to another bus. You might do this in order to stack sub-mixes, sending all the drums to a "drum" sub, then that to a "foundation" sub, and that (along with horns sub, strings sub, voices sub) to an "orchestra" sub. Figure 10.1 shows the I/O section of a sample session with sub-mastering. The outputs and corresponding inputs are color coded. The output for Audio 1-3 are no longer Stereo Output, but rather Bus 47. Auxiliary 1 has bus 47 as it's input. This is the drum sub-master. The output of the drum track is not Stereo Out, but bus 50, which is routed to Aux 2; my foundation sub-master. The bass track is also routed to bus 50. The guitars and keys (49) go to a rhythm sub (Aux3), which in turn is routed to Aux 2. So Aux 2 is a sub-mix of sub-mixes (and bass), but it is 92

still not routed to the main output, but rather Aux 5 via bus 53. Track 8-9 have a sub-mix, then they join the foundation sub-mix at Aux 5. Audio 10 has no sub-master. It is routed directly to Stereo Out. Follow the output (bottom color) to the matching color input in the upper slot. Fig. 10-1 Sub-Masters

I may have gone overboard with this example to demonstrate the flexibility of grouping this way. At any time I can listen to just the guitars, or all the foundation, the voices and the rhythm, the rhythm and the drums, and so on. The beauty of DAWs is that you are not limited to the number of sub-mixes, where in real life you'll be lucky to have one or two (stereo) subs. A single effect (compression, reverb, EQ) can be applied to the sub-master. This is more efficient in terms of DSP, but also your workflow. One change to that single plug-in effects the entire sub-mix (as opposed to placing a compressor on each channel, then having to change the settings on them all if you want to tighten up the entire set). You can adjust the level, mute, or solo all the instruments at once. The balance you spent hours achieving between each individual element of the drum set will be preserved. If you want to adjust a single item, you can do that on its own track. I like to have my sub-mixes next to the channels they are receiving. Unfortunately, you can only rearrange tracks in the arrange window, not the mixer. To do that, you have to place them in the arrange window. To do that, select all auxiliaries, control click and choose create/select arrange track. Finally, add colors for clarity, festivity, or added confusion.

Other tricks. • Double click on any output (say, Bus 47) and the auxiliary channel that receives that bus will become highlighted. • In the inspector there are two channels shown; the selected channel, and the next channel in the signal chain. • In the upper right hand corner of the mixer click on Single to show the entire path of any single selected channel. • When soloing a track, logic automatically solos all sub-mixes along that chain.

93

•

Parallel Effects

… are the second application of auxiliary routing. Consider a mix with six vocals; four backups and two leads. You will probably want the backup vocals deeper in the field (more reverb) than the leads. You also may feel a female vocal benefits more from the reverb than the male vocal. Placing a reverb on each channel is inefficient, yet using a single reverb on a master does not allow the flexibility of a different wet/dry mix for each instrument. The problem is this: you want independent control over the dry and wet mix, but you want to use a single reverb. You need two mixes, the dry mix, and the reverb mix. Fig. 10-2 Reverb Mix vs Dry Mix

The solution; split them into two mixes. One main mix, the other an auxiliary mix for reverb. Figure 10.2 shows a possible dry/wet mix. Blue is the dry mix, green the reverb mix. Note that all the backups are at the same level, but the reverb for these same channels is greater, and all the reverb levels for the backing vocalists are greater than the leads. Similarly, the male vocal is louder than the female in the dry mix, but the female has more reverb. The only way to achieve this is with an auxiliary send. Of course, Logic, and any mixer, digital or steel and plastic, will auxiliary sends.  All the plug-ins we've used so far were placed as an insert in the channel strip. This is called serial routing. The effects on a channel strip are part of the main routing, in series. When you use an auxiliary mix it is called parallel routing. The signal is syphoned off to create a second auxiliary signal flow which is treated with DSP then re-inserted into the final mix. It is parallel because it runs parallel to the serial flow. Imagine a park with water pipes. You have a main pipe carrying water to the fountains, but you want to add treatment in precise amounts, so precise that you can't add it to the main stream, which varies in size. So you attach a small pipe that takes some of the water along a parallel route for treatment. It is then added back to the main flow before exiting at the destination. Note that with serial reverbs and delays it's important to set the balance between dry and wet signal on each plug-in. In a parallel system the dry signal is supplied through the main routing. For this reason the parallel reverbs don't need and shouldn't have any dry signal. The amount of the effect heard in the mix is controlled by the level of the dry signal being sent into the auxiliary channel.  In short, serial reverbs: blend of dry and wet mix. Parallel reverbs and delays: 100% wet (no dry signal). To add an auxiliary send, select all the tracks you want to send (and there is really no reason not to select them all, so select them all), then click on an open Sends slot and choose a bus. Which bus number? It doesn't matter. A new auxiliary track will be created with that bus as its input. On this new auxiliary track you can place a reverb, delay, or even compressor. This is a parallel effect. How is this different from a sub-master auxiliary track? The sub-master is serial; in series. It is part of your main mix, and you adjust the levels using the gain sliders. The auxiliary track is parallel; independent of the main signal flow. You adjust their levels with the small knobs to the right of each send. These levels can and should be different from the main. 94

Fig. 10-3 Auxiliary Sends vs. Sub-masters

Figure 10.3 shows the three different types of auxiliaries. Tracks 11 and 12 (BU Voc) are being routed to Bus 51 using the I/O section, which runs to Aux 4 (labeled Voices). Auxiliary send 2 (the second row of blue boxes) are connected to Bus 17, which is routed to the input of Aux 6, labeled Reverb. To the right of each Bus 17 is a small knob. This is the level being sent to Auxiliary 6. While the main mix has all the channels at the same level except Voices and Backup (-7.2 and -1.4), the auxiliary mix 1, routed to Bus 17, therefore Aux 6, which has a Space Dsn plug-in, is the reverb mix. It's levels are very different; some of track 13, more of track 14, none of 15, 11, or 12. The reverb (Aux 6) is sent to the final mix with these relative amounts of reverb. So the Voices will have some reverb, and Backup will have a lot. The third type is send 1, Bus 11. This is being used as a monitor (covered next). As you raise the level with the knob to the right of each Bus 17 you send more signal to the reverb. It is not uncommon to have four or five auxiliary fx sends. They are handy for trying different types of reverbs. Real life example: a marching band had recorded in a very dry room and wanted to match another recording done in a very live stadium. We did 10 auxiliary sends. On each auxiliary channel placed a Space Designer with a different stadium convolution reverb. The directors could easily solo one stadium, adjust the level, even mix and match. They found a matching sound within a half hour. As a matter of fact, most of your reverb and delays will be auxiliary. The only reason for a serial reverb is if you're doing a London style effect to set that instrument off with a delay or character of its own.  There are now five critical mix points you need to monitor and adjust: The dry signal, controlled by main sliders of each track; the wet signal, controlled by the small knobs next to send 17 on each track; the sub-master, controlled by slider on the sub-master track; the master reverb mix, controlled by the slider on the reverb track; and the master, controlled by the slider on Out 1-2 (was that five?). 95

Auxiliary routing should be standard procedure in any session. A project can be filled with creative, well performed material or sophisticated concrete treatment with stacks of plug-ins on every track, rife with automation and invention, but missing auxiliary mixes or sends is telling. Auxiliary routing is, for me at least, a touch stone of an engineer's ability. Auxiliaries add efficiency. There are two additional strategies for efficiency: •

Freezing and Distributed Processing

Reverbs, delays, and (covered later) physical models tax the CPU because they generate signal in real-time. You can monitor both CPU (real-time calculations to generate audio) and disk I/O (reading and writing audio from the disk) by choosing Options/Audio/System Performance. You can also control click on the transport and choose Customize Transport Bar… then check Load Meters. Freezing a track can transfer CPU load to I/O load. It commits the real time effects on that track to an audio file; essentially bouncing it to disk. In the mixer area or window, ⌃click on any track and choose Freeze. This will add a snowflake icon to each track. The next time you press play Logic will take a few seconds to bounce that track to the hard drive. Once it has been frozen, pressing play will read the track from disk. When unchecked, the freeze file will be discarded and audio will be generated from scratch. Freezing cost very little, so it makes sense to freeze all tracks that are more or less finished, or even that are not the current focus. The second CPU management technique requires several computers. Distributed processing can turn a 15 station lab into a parallel processing supercomputer. There is a small app called Logic Node that will accept audio processing load over a (high speed) network. On the master station you can select, say, tracks 1 through 10 and send those to be processed on station 2, 11 through 20 on station 3, and so on. It takes a little time to set up, but once in place your CPU overhead can be increased dramatically. •

Auxiliaries for Monitors

In addition to subs and effect sends, auxiliaries are used for monitoring (or any mix that differs from your main mix). Consider a live mix scenario. An engineer may spend hours getting the house to sound exactly the way he wants. But the main house sound is only one several mixes that need attention. In the middle of the performance you often see the lead vocal point to her monitor and mime "more piano." The engineer is certainly not going to turn up the piano in the house. It's right where he wants it. Then the bassist will mime less piano in his monitor. How can the engineer increase the piano only in the monitor for the vocalist and decrease it in the bass, but not change either in the house? A similar conversation may occur in a studio session. In this scenario you are laying down additional solo tracks to an existing mix. The vocalist, in one isolation booth, will want more piano in her headphones, but the guitarist only wants drums, bass, and himself in his headphones. But the producer, sitting next to you, only wants to hear the vocalist. How can you satisfy all three? The only solution is three separate mixes; separate from the mains but also from each other. This can only be done with auxiliary sends. You create them in the same way as an effects send, with one added step. Select the tracks that you want sent to the auxiliary mix (and you just as well select them all, including the effect masters—performers may want to hear some reverb), then click on an open auxiliary slot and choose a bus. Any bus will do. A new auxiliary track will be created. Name each one 96

monitor 1, monitor 2, etc. The second step is to make them pre-fader. Click the send once more and select pre-fader. With a parallel reverb you want the level of the reverb to match that of the main mix. If you turn down the guitar in the main mix, you will want the reverb to come down too, to maintain balance. For that reason, effect reverbs are post-fader. That is to say, they inherit the level set by the main channel strip fader. Monitors are the opposite. You may spend hours getting two headphone mixes just the way the artists want, then you decide the guitar is too loud in your main mix and turn it down. You don't want that to change the mix in the headphones. Similarly, you may mute several channels in your main mix to focus on the foundation instruments. If your headphone sends are post-fader, they will be muted too, and the artist can no longer play along with the tracks. If they are pre-fader, in both cases they will be unchanged by either changes in the fader or muting/soloing. 

As a general rule: effect sends are post-fader, monitor sends are pre-fader.

Fig. 10-4 Routing for Headphone Sends

One final step: Signal is sent to these mixes by increasing the gain using the knob to the right of the bus, just like the reverb. The difference is the output. With a reverb auxiliary, you want the reverb to be mixed with the main signal, so the output on parallel effect sends will be Stereo Out. But monitors need to be routed to a different output; 3 and 4, or 5 and 6, etc. Figure 10.4 shows the possible routing choices for an 18 channel interface. (You were wondering what those other outputs were for, weren't you?) The send can be stereo for headphones, but for stage monitors, or if you're doing a lot of them, you can choose mono. If, for example, you choose 5 and 6, then you would connect the outputs 5 and 6 of your interface to the headphone amp. Take another look at Figure 10.3. Auxiliary send 1 is routed to Bus 11, which is connected to the input of Aux 7, with an output to 3-4. It may take a few sessions or live mixes for auxiliary mixes to sink in. It's one of those things that you don't really get until you're faced with the problem. You'll be in the studio, laying down tracks with the performer who is listening (via a splitter which is how everyone does it until you get it) to the same main output as you, then the artist will ask if they can just hear the foundation. You'll solo the foundation and sit there thinking, "but I still want to hear the other parts to work on the mix—I wish there were a way to do that." Then you'll come back and read this chapter. The same is true for the post/pre-fader options. When your soloist yells at you for muting one of the tracks in his mix, you'll think "how can I mute it in my mix, but not in his?" Pre-fades are the answer.

97

• Exercises 10:1 Start with a session provided by the instructor. Add serial effects, compression, EQ, where necessary. Add sub-masters, two effects sends, and two monitoring sends. 10:2 Route all your drum tracks to an auxiliary, parallel send. On the resulting auxiliary track, place a compressor and apply extreme compression: very low threshold and high ratio. Blend the two so that the heavily compressed auxiliary is just below the natural track. That is, treat the compression as you would a reverb. 10:3 Use the image below to practice your skills with the following scenario: You are in the studio prepared to record two additional tracks; guitar and bass, each in a separate isolation booth with headphones. They can only hear themselves through their headphones. Though no output is shown in this image, assume the bass is hearing Aux 4, the guitar Aux 5. Plus; the producer is looking over your shoulder and thinks his distracting comments are helpful. Match each of the requests using the correct component in the mix window for the producer (P) in the main mix or one of your two performers (G and B). The rows are labeled A through M, the tracks P through Z. Include in your answer the correct term for each item used. For example, if the producers asks you to pan the Bass guitar to the left, your answer would be: K-S, pan control. D through F refers to the gain knob to the right of the Bus 1, 27, 28. M is mute. N is solo. You should be able to answer in less than two seconds. During the third and fourth seconds the producer or performers are thinking "does he know what he's doing?" 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

(P) More reverb in just in the snare. (P) Compress all the drums. (P) Less snare in the main mix. (P) More of all the drums. (B) I don't hear the click track. (B) The balance is good, but it's all a little too loud. (G) My click is too loud. (G) Turn the organ up as high as I am. (P) Group the guitar and bass. (P) Send the click to the drum sub-master. (P) Pan the organ to the left. (P) Duplicate the roll-off to the overheads

13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

98

(G) My balance is good, but could be higher overall. (P) Adjust the OH balance to the right. (G) I'm switching my guitar to input 3. (??!!?) (B) Could you mix a little reverb in my headphone? (P) Solo the guitar and drums. Toggle them. (B) Can you take the organ out of the reverb? (P) I want to hear it. Duplicate it as a serial effect. (G) Add that funky delay we tried earlier to my track. (P) Toggle the reverb on and off. (P) Increase the master reverb. (P) Boost the highs in the reverb. (P) Compress the bass.

Answers 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

D Q, turn the dial up C W, click, choose compressor L Q, move down L W, move up E V, turn the dial up L Y, turn it down F V, turn the dial down F U, turn dial to 12:00 Select S and T, click J I V, click and choose bus/17 B U (! not K U), double click, pan left A T, option-click-drag to R L Z, move up K U, dial right H T, click, choose input 3 E X, dial up Shift click T and W, N D U, dial down B X, option, command, drag to U, adjust dry/wet C T, click, select delay designer M X, click L X, up A X, adjust B or C S, click, choose compressor

99

11.

Adding Sound Effects which often depends on style. Don't over do it.

Auxiliary reverbs and delays are more efficient, but only if you want to apply the same style reverb to all instruments. In electro-acoustic composition and a London style mix you may very well want two or three independent delays or reverbs on each channel. •

London Effects

Fig. 11-1 Loop Browser

Logic is littered with sophisticated DSP modules. To try them out, first find a dry drum loop with a broad frequency spectrum from the Media area, Loops. Clicking on a category narrows your search to items in that category. Drag a handful of similar loops for experimenting. Insert a plug-in and browse through the presets to familiarize yourself with the effect. Once you find something interesting, adjust the controls to see how the sound changes. For now just explore. (I always gravitate to the presets in the warped folder.) Later chapters will cover many of these processes in greater detail. This chapter contains some of my favorites, with a few hints for each one. •

Media/Library, Shortcuts, User Settings

Fig. 11-2 Channel Strip Focus

We are about to rifle through presets for these plug-ins, so I'll draw your attention to the Library tab of the Media area. The content of this window changes to reflect the focused item on each track. Click once on any plug-in, including the main channel strip or an instrument in the I/O input slot, and the Media Library shows all the presets for that plug-in. From here you can use the up and down arrows to select items and the left and right to move up to a higher folder. You can also select new plug-in settings and presets using a shortcut. Logic's shortcut structure is flexible and extensive. Choose Logic/Preferences/Key Commands to learn existing shortcuts or to create your own. (It's worth reading through them all once, but you'll never remember or understand them all. Review them periodically.) In the search field type "Plug-in" to find the shortcut for Next and Previous Plug-in Setting or EXS instrument. Start playback, click on the plug-in window to bring it to focus, then use this shortcut to quickly scroll through the presets. You can collect your favorite presets using the Save As menu. This will move that setting, whether renamed or not, to the top of the menu. An alternate method is to open Hard Drive/Library/Application Support/Logic/Plug-In Settings/ and option drag your favorites to your own library: ~/Library/Application Support/Logic/Plug-In Settings/. •

Delay Designer

Try the complex and warped categories. This effect allows you to precisely control the character of each reflection. Find an interesting 100

preset and click each of the tabs; cutoff, reso, transp, pan, and level. Adjust a single reflection and note the effect. •

Evoc 20 Filterbank

Adjust the LFOs at the bottom. The column in the middle allows you to choose the shape of the wave that is used. (LFOs are covered in depth later.) The buttons on the left are for the left LFO, the right for the right. The Rate is how fast each LFO repeats its cycle. If you choose either of the ramps, with a rate of 1 Hz, the aspect of the sound being controlled by that LFO will rise 1 time per second. The right side of the knob sets a free rate independent of the project metronome. The left side syncs the rate to the tempo of the current song. This is a powerful feature. •

Evoc 20 TrackOscillator

Likewise, experiment with the LFO settings and wave form, as well as the Formant stretch, shift, and resonance. There is a tiny piano keyboard which sets the scale being used to modify the pitch. Try deselecting most of the keys for a more exaggerated effect. •

Spectral Gate

This effect is very similar to noise reduction systems (where you sample a section of noise then suppress it), which, though ineffective with music, creates interesting effects; frequencies pop up above a given threshold. The spectral gate works on the same principle. Change the center frequency, band width, sub and super energy. Change the threshold to blend the dry and processed signal. •

Flanger, Modulation Delay, Phaser

These are all variations of phase shifting (the thing I told you to avoid in Chapter 6). The input signal is duplicated, delayed, and detuned in various ways. This thickens the signal, similar to over-dubbing. It also increases spatial effects as did the old Leslie system with a speaker that physically rotated, sending signal cascading around the performance space. Try the presets, fiddle with the knobs. •

Ring Shifter

In brief, a ring shifter blends two signals together, one modulating the other. This process generates improbable and complex overtones (sidebands). The details of this process are covered in the FM chapter. •

Side Chains

A side chain links a plug-in on one channel with the content of another (usually audio) track. The best example of a side-chain is the Ducker. The Ducker is only available on submasters (which doesn't make sense to me). Once in place, turn on the Ducker, click on the side chain pull down menu and select the track you want the sub-master to duck. When there is signal present in the ducked track the ducking track will duck…the track being ducked. Other typical 101

side chains are compressors (which then have the same function as a ducker) and the vocoder, which uses the side-chain track as input for synthesis. •

EVOC 20 PolySynth

Though this is an instrument, not a plug-in, it is a good illustration of a side-chain and would be very well suited for a London effect. It can be used as a stand alone synthesizer, but it's most impressive settings are in the vocoder presets. It requires another source of audio which is used as an input to be modulated. For best results, first try speech. Record or import an audio file with spoken dialog (make sure it is optimized) into an audio track. Create a new software instrument track and select the EVOC 20 PolySynth in the input slot. First choose one of the Synthesizer presets just to make sure you're getting sound. Press the Caps Lock to bring up the caps lock keyboard (or use an external keyboard) and play a few notes. This is the stand alone synthesizer. Next, choose one of the Vintage or Warped vocoder settings. These settings require input from another audio track. To link the two, click on the Side Chain pull down menu and select the track that contains the spoken audio. Start playback and while the audio track is playing, press notes on keyboard to play the vocoder synth. You should hear (and should be astonished by) the speech being "sung" at the pitches played by the keyboard. This is the same effect as Imogen Heap's Hide and Seek, though I can't say what program or instrument she used. Mute the original audio for the Evoc effect only. Enjoy. •

Spatialization

Reverbs and delays can serve two roles. In a popular mix or electro-acoustic composition they thicken, enrich, or change the character of the source. The effect helps define the sound. That is why a post-fader send makes sense. The effect is integral to the material and should increase or decrease when increasing or decreasing the dry signal. Otherwise the carefully balanced effect will be compromised. But they can also be used for spatial effects; the sense that a sound is moving around in a space, or the listener is moving around a space filled with sounds. This is an interesting (though all too often omitted) technique in electro-acoustic composition. In film sound tracks it is absolutely necessary. The complexity of spatialized audio in film goes unnoticed until pointed out.  Anything moving on screen, or any change of perspective of the lens (for example cutting from outside a cafe to the inside where a couple is having a conversation) changes the nature of the sound in five ways: 1) the overall volume of the source 2) the relative volume in the left and right ears 3) the ratio of reverb to source 4) the timbre and 5) Doppler. Of course the volume of an object will change relative to its distance and position left and right, but unlike a post-fade send, the reverb doesn't change in relation to the dry signal. When someone walks away from you in a hall their footsteps get quieter, but the actual volume of the steps stays the same, so it will generate the same amount of reverb in the hall. In other words, to create the sense of motion in a space, the reverb must be pre-fader. High frequencies travel further than low frequencies, so the timbre will also change as the source position changes. Finally, motion creates a Doppler effect. As a sound approaches you the wave forms are compressed relative to the speed of the object. As it passes and moves away it becomes decompressed. In other words, when sounds are approaching the pitch is slightly 102

higher, and slightly lower when moving away. This effect is most evident, even cliché in emergency vehicle sirens. Fig. 11-3 Binaural Panner

To automate these effects with the tools you currently understand you would first need to change your auxiliary reverb send to pre-fade (or with a serial reverb automate the mix), then automate the volume, but also the pan—closely coordinated with the volume, and even more difficult, the EQ, also carefully aligned to the volume and pan. Oh, and don't forget Doppler. Of course, Logic provides a tool that combines all of these. Click on the output slot and choose Binaural. Double click on the resulting pan knob to create spatial effects with a single control point. (A stereo binaural control will include a left and right spread. I'll cover just the mono parameters.) Adjusting the blue dot in the center will adjust the volume of the source without changing the level of the reverb (i.e., it is a post auxiliary send), the timbre, and if Doppler is selected, the pitch. Change the size for a more exaggerated effect (greater change in volume). With a large size, the edges of the circle will result in nearly all reverb and no dry signal. Small footnote: you might be tempted to use this in place of a directional mixer. But be aware that this does more than change the pan location. It also modifies the character of the sound. It's best used only for spatialisation. •

Automation

Just a little perspective so you can appreciate the power of automation. Thirty years after the seminal Forbidden Planet (the first major motion picture to take electronic "tonalities" seriously), composers mixed from an 8 or 16-track analog tape. Creating spatial effects in four channels required two hands, sometimes more, per track. One had to assign the output to determine the two panning points, then adjust the volume and pan at the same time, and sometimes the auxiliary send. After preparing all the "parts" on the tape deck, one might write a musical score on staff paper, for four engineers, each controlling two tracks. They (ok, it's me and my three comrades) then perform the piece by playing the mixer. With Logic Pro, it's a click away. Specifically the automation icon in the toolbar. This changes each track into a timeline. (Check the shortcut for your system; probably "A".) These timelines represent a parameter on the track, effect, or instrument, and the luxury of carefully planned and executed dynamic change. It would be criminal not to exploit such power. Fig. 11-4 Automation Menu

Any parameter can be automated. If you can change it with the interface, it can be automated. The default timeline is volume. Click on the pull down menu of another track to change it to pan. Automating these two parameters should be routine in a London style mix. Add an instrument track and choose the ES1 as the input. Change the automation menu to ES1 and notice the complication of nine sub-menus, one of which has 70 items (duplication, but still). With these complex machines I never discourage blind 103

exploration. You're not going to break anything. (As always, use headsets and turn them down before trying something unusual.) Click and dial away. That's how we all start out. (In an interview I can no longer find, Louis and Bebe Baron were asked about their methods when working on Forbidden Planet. They admitted that they often had no plan, but rather just tried patches, adjusted knobs, and stood back in awe, tape deck always running to capture the serendipity of their poke and rummage method. During my first electronic music course, which used vintage patch and dial modular synthesizers, our instructor put a grad student in charge when he was out of town. Our group of five quickly hi-jacked his agenda and proceeded with this single goal: use up every patch cable and every module. It was one of my favorite classes.) The problem is, you will stumble onto an interesting parameter and not be able to identify it in the automation list. Turns out, even when you know the name and understanding how the parameter works they are difficult to find. Solution: •

Latch

Just below the group field is the automation menu (default "off"). Click and choose Latch. When engaged, during playback (not record) Logic will automatically register any change made on any parameter on that track, plug-in or instrument. Fig. 11-5 Multiple Timelines

This may seem handy, but in order to smoothly replicate your movements it splatters the timeline with unnecessary break points. In most cases you will want a smooth line, not precise duplication of your imperfect motion. I prefer to work with the timeline itself, creating and moving breakpoints with smooth lines between. You can also view and edit automation using the event list in the Lists field. • Clicking anywhere in the track will generate a break point and move the timeline to that spot. You don't need to create a break point then drag it. Just click where you want the next one. • Use the Marquee tool (the crosshair) to select a section. Click and hold on the selection to create four break points at the ends of the selection. You can now drag an entire section down. • A click with a quick release will create a break point. But click-hold will grab the line and move that entire section. • Shift click-drag to select a section of automation. You can then click and drag to move, option drag to duplicate, or copy and paste. • Click the reveal triangle to open up another timeline (Figure 11.5). Each click will bring up a new timeline for that track. Choose a new parameter for each. This way you can closely coordinate several parameters. But latch does another very useful thing: it brings to focus (i.e., selects in the pull down menu) the parameter timeline of anything you change. Just touching (clicking) a parameter during playback will bring up that timeline. So I use latch to find parameters. Try this: click on the automation track reveal triangle in the lower left corner of the track to open a new 104

automation timeline. Click on its triangle to open another. Continue until you have six or seven timelines. They may all show the same automation at this point. Open a plug-in or instrument with plenty of dials, sliders, and buttons. Start playback and click away on five or six parameters. They each pop up in a new track timeline. Automate at will. •

Exercises 11:1 Drag a short video excerpt from a science fiction movie (say, Forbidden Planet) into the arrange track of a Logic session. (Choose something with lots of special effects.) Populate the tracks with sound effects. Focus on spatialization, using Doppler for moving items and automation to adjust levels with scene changes. 11:2 Add a voice-over in the style of MST3K. Route all the incidental music and sound effects to a sub-master. Place a ducker on the sub-master, and duck your voice over track. 11:3 Pick four of your favorite DSP effects. For each one find three interesting presets. Explore the parameters without reading any explanations. When you arrive at an interesting sound, save it to a folder.

105

12.

Review, Examples First a session start to end. Then analysis.

•

Review Session • Create a session or open a template with instrument or audio tracks. You can add more later, or drag content (either MIDI or audio loops) into a blank space in the Arrange window, which creates a new track. • Check the preferences for input interface, sampling rate, buffer size, etc. • Set the audio input by clicking on the upper slot in the I/O section. Choose an input source that matches the interface input (that corresponds with the physical mic that is sitting in front of the real musician). Note that the input source may not and often does not match the track number. You may be using mic 1 (input 1 of the interface) on Logic's track 13. This is true especially when limited with a 2-channel interface. • Choose an instrument on software instrument tracks by clicking on an input slot in the I/O section. • Add a channel EQ and compressor, at the least, to each track. If you're going for a particular sound, give in to the channel strip settings, but with this caveat: Try to get the sound you want before it hits the mic. Adding EQ, compression, or even reverb at this point will weaken the final product. Exposed tracks inspire better musicianship. • Record content, either live audio or MIDI on the instrument tracks. Click the R button to arm the audio tracks for recording. MIDI tracks are automatically armed when selected. • Alternate: Add content by dragging loops from the Media area or dragging items from the desktop into the arrange window, either in an existing track or a blank area to create a new track. • Familiarize yourself with the material and plan the rest of the session. Get it to sound the way you like before moving on. Adding sub-mixes, sends, and plug-ins too early risks obscuring the original intent and quality of the work. • Foundation: identify elements of the foundation, group them or route them to a sub-mix by clicking on the output of the I/O and selecting a bus. Still in mono, adjust the balance of each using the volume sliders. • Create an image for the foundation by panning. Be sure to avoid big mono by panning stereo tracks with a directional mixer. 106

• Next work on rhythms, sub-mixing when appropriate, setting levels, then image using pan and directional mixers. • It's likely that you haven't yet recorded the leads, fills, or pads. So you will need to add headphone mixes using the sends (select Pre Fader). Choose any bus. On the resulting auxiliary track select the output under I/O and choose an output other than 1-2. Connect those outputs to your headphone amp. • Finally, record lead, fills, and pads. Still, you are mixing dry. Engage the cycle to loop record. This will create comps, or multiple takes joined into one region. You then can select very short segments of a given take. • Use the comping feature to select which sections of which takes you want to keep. • Adjust EQ for each track. Start with the presets. Add roll-offs on everything that has no low end. Adjust the drum tracks or guitar tracks for crispness and sizzle. Consider unusual EQs if a London mix. Use boost/sweep/cut to find and cut annoying spectra. You may end up leaving it flat, but it never hurts to explore. • Compressors, gates, limiters next. Likewise, there's not much reason not to compress everything, even just slightly. Compression and EQ take very little CPU. (The exception would be synthesizer tracks or apple loops, which typically have compression built in, or as discussed above, classical works.) Solo individual or groups of instruments (bass and guitar, then drums and bass, guitars and drums, then all three), adjusting as you go, slowly adding more and more instruments. • Now turn your focus back to individual instruments and improve them with slap delays, synced delays, exotic filters, reverbs and overdubbing. As a general rule the order of inserts should be EQ, compression, filters, delays, noise gates, reverbs. (Signal flow on a channel strip is top to bottom.) You may also want to put a compressor or limiter and another EQ at the bottom of the list of processors. • Automation next, since you are now working on the character of the individual instruments, and especially if this is a London mix. First automate image (panning). Even in LA mixes, a subtle pan can be effective. Automate main levels next, but mostly to correct small flaws in the mix, or to fade elements in and out. ins.

• If a London style, consider automating the parameters of software instruments or plug-

• Depth: select one of the sends slots and choose a bus. On the newly created auxiliary add a reverb (100% wet). Dial each instrument into the "space" using the level control to the right of the bus. Solo, mute, engage groups or sub-masters to listen how instruments interact with the other. Consider at least three levels of depth; far, middle, and close. To hear the quality of the space, raise the level much higher than it should be and mute the main track or pause playback to hear 107

the reverb tails. When happy with the sound, dial the reverb down until it is no longer apparent (LA mix). Then A/B the dry/wet by muting the reverb master to see if it improves the sound without sacrificing clarity, adjusting accordingly. • Listen to the mix and review style (NY, London, LA, Garage) and ask what's missing or can be improved. Consider the elements of the mix; are there two rhythms? Do they compliment each other? Are they panned to compliment each other? Should one be removed? Consider panning according to function and frequency spectrum. Do both left and right channels have plenty of highs? plenty of punch? crispness? Is there a frequency spectrum that needs to be filled? A position in the image that is empty? You may very well decide not to fill a particular spectrum or not pan hard left and right, or even exclude an element (e.g., a pad). You may choose not to do three levels of depth, or no depth (for an intimate, LA style). But you need to consciously choose not to. It would be a mistake to omit something because you haven't considered it. Plus, it doesn't hurt to try. You might discover something interesting. • Try a parallel compression for a more natural sound. That is, use a send to route signal to an auxiliary. Send a group of instruments (drums, typically) to that auxiliary. Apply aggressive compression, then blend that signal just below the level of the natural signal (as you would a reverb). • At this point, you might want to listen to some of your favorite mixes, in particular those you want to replicate. Compare and adjust. • Approach automation again. For a fresh sound in a chorus or middle verse, consider changing the position, EQ, or depth of an element. Remove instruments entirely for a variation in texture. • Fine tune volume levels using a ducker or a compressor side-chained to a lead track. (Not all plug-ins have side-chains, and those that do apply them differently. Read the manual for specific functions.) For example, add a compressor to a rhythm guitar track, adjusted to reduce the volume slightly. Use bypass to test the compressed and uncompressed version. Side-chain the compressor to the vocal track so that it kicks in only when signal is present. If you want to do this with a group of tracks, send them to a sub-master with a ducker. • Add to the master track (and possibly sub-masters) an EQ, (yet again) compressor, gain control and a multimeter. Make any final adjustments to the EQ, but be subtle. Likewise, compress discreetly. On the gain control select mono, flip the phase of one channel. Check the multimeter for phasing (correlation). Blue is good, red is bad. If you see or hear problems, EQ, pan in mono, low roll-offs. • Now you can listen—with headphones, through speakers, through a ghetto blaster, ear buds, cranked, barely audible, through the crack of a door, with the vacuum cleaner running, in the car, in the space where it will be performed (do 108

three mixes; for a dry hall, medium hall, and wet hall), in the morning when you're fresh, in the afternoon after listening to other things all day, after two weeks—and make adjustments. •

Electro-acoustic mix

It is a common fault to electro-acoustic works should be treated differently, or that outrageous or even clever and inventive material exempts one from following elements of mix, composition, organization, and style. This error leads to indulgent works filled with all the material developed during the semester in the foreground, as lead, without discrimination. Foundation, rhythm, pad, lead, fill, depth, height, width, EQ, compression, repetition and variation work just as well with more artistic or intellectually challenging material. •

Examples and Analyses

Below is a collection of the topics we've covered with examples from the popular literature. The next section provides a synopsis of analysis in style, elements, and image from my own collection. LA Style: Doobie Brothers; Norah Jones; Steely Dan; Eagles; Ricky Lee Jones; Tracy Chapman; God Bless the Child, Blood, Sweat, and Tears (panned as if on stage); Girl Talk, Chaise Lounge NY Style: We Built This City, Starship; After The Boys Of Summer Have Gone, Don Henley; Life in the Fast Lane, Those Shoes, Eagles; The Bitch Is Back, Elton John; Gypsy, Fleetwood Mac; Everything, Dixie Chicks, Heart, Marc Broussard, Pretenders, ZZ Top London Style: Most of, Because They Invented It, Beatles; Crazy, Zero Space, What to Do Before I'm Dead, Kidney Thieves; Voodoo Child, Angélique Kidjo; Sunshine Superman, Chaise Lounge; Everything, Queen Garage Style: Tom Waits, Rolling Stones, Bob Dylan, Grateful Dead, Hendrix Fill: Give Me One Reason, Nora Jones, Guitar (R); Stardust, Willy Neilson, Guitar and Harmonica; Knockin' on Heaven's Door, Clapton, Wah Guitar; Fooled Around and Fell in Love, Elvin Bishop, Piano; Roly Poly, Asleep at the Wheel, Guitar Rhythm: Give Me One Reason, Nora Jones, Guitar (L); Fooled Around and Fell in Love, Elvin Bishop, Offbeat Guitar pops; Everything, Doobies, Guitars; Oh Girl, Chi-Lites (what a great mix), Guitars; Every Breath You Take, Police, guitar Pad: Give Me One Reason, Organ; Stardust, Willy Neilson, Organ; Knockin' on Heaven's Door, Clapton, Organ; Fooled Around and Fell in Love, Elvin Bishop, Organ; We Can Work It Out, Beatles, Accordion; Every Breath You Take, Police, strings? Height, Width, Depth, Elements (exemplary mixes): Steely Dan; Most of Toto; Owner of a Lonely Heart, Yes; Crazy, What to Do Before I'm Dead, Arsenal, Zero Space, Kidney Thieves; Voodoo Child, Angélique Kidjo; Reanon, Fleetwood Mac; Every Little Thing, Seven Days, Brand New Day, Sting/The Police; Nikita, Elton John; Low Down, Boz Skaggs; One Love, Bob Marley; Home, Marc Broussard; Have a Good Time, 50 Ways to Leave Your Lover, Paul Simon; Englishman in New York, Sting Concrète treatment (covered in chapter 14): We Have Heaven, Yes; Owner of a Lonely Heart, Yes; Revolution 9, Beatles (though not a great example—too obscure), but how about Strawberry Fields Forever, For the Benefit of Mr. Kite and the gobbledygook at the end of A 109

Day in the Life; 2000 Light Years From Home, Rolling Stones; Magic Man, Heart, Opening guitar solo; Brand New Day, Sting; Another One Bites the Dust, Queen (reversal); Castles Made of Sand, Hendrix, left channel guitar solo is reversed Overdub: A Venture, Yes; Fast Car, Chapman; Red Red Wine, (vocal) Marley; Sunshine Superman, (vocal) Chaise Lounge; Eyes of Silver, Dependin' on You, Listen to the Music, Black Water, Long Train, Little Bitty Pretty, Pretty Much Everything? At least pre-McDonald, (guitars) Doobie Brothers; And Most of the Eagles For That Matter, Eagles; Every Single Word He Sang, Sometimes Even Different Lyrics, Except Maybe the Live and Unplugged Stuff, But Who Cares, He's a Genius, Clapton; The Bitch is Back, guitars, Elton John; Gold Dust Woman, guitars, lead vocal, Fleetwood Mac; Dog and Butterfly, Heart, Guitars; Brand New Day, Sting, Chorus; Another One Bites the Dust, Queen Filtering, Creative EQ: Voodoo Child, Angélique Kidjo; Voodoo Child, Hendrix; Stir It Up, Marley; Gold Dust Woman, (drone) Fleetwood Mac; Magic Man, Heart, at 4'00"; Let My Love Open the Door, Pete Townshend; Higher Ground, Stevie Wonder; Who Are You, The Who Creative imaging: Crazy, Kidney Thieves; Beatles Doubling or slap delay: Owner of a Lonely Heart (1:17), all three guitars, Every Breath You Take, The Police; She's So Cold (guitar left), Rolling Stones; Shine Star, (guitar, vocals), Earth, Wind, and Fire; Got to Get You Into My Life, (horns) Beatles; D'yer Mak'er, Zeppelin, vocal; Every Breath You Take, Police, guitar; It's Raining Again, Supertramp, everything, I think Synced delay: The Battle of Evermore, Whole Lotta Love, (mandolin) Led Zeppelin; Is This Love (organ), One Love, Red Red Wine, (snare) Marley; Crazy, Kidney Thieves; Under Cover, Rolling Stones; Lido Shuffle, (snare) Scaggs; Drive, (tambourine?) Cars; Sunshine Superman, (0'37"), Chaise Lounge; Tell it Like it Is, (vocal) Heart; Keep Comin' Back, Marc Broussard; Every Breath You Take, Brand New Day, Police London style delays, panning: Kidney Thieves; Little Wing, Hendrix Compression: I'm Yours, Jason Mraz, Sunshine Superman, Chaise Lounge; Dixie Chicks Chorus/Flange: Owner of a Lonely Heart (1:04), Yes; What to Do Before I'm Dead (1'50"), Kidney Thieves; Life In the Fast Lane, 3'40", Eagles; Little Wing, Hendrix Notes about the analyses: I omit obvious things such as foundation and lead. The chart details items that are the most difficult to distinguish, and what you might expect on the exams; style, fill, rhythm, pad, and (since it's so common) overdubbing. The third column maps out the image: hard left, left, center, right, and hard right.

110

Title, Artists, section in m:ss. Comments. Style Fill Hard left Left

Rhythm Center

Pad Right

Dubbing Hard Right

Give Me One Reason, Chapman, 1:00-2:30. There are two fills, one is mixed near, the other quite deep. LA Guitar on right, Deep Gtr Guitar on left Organ Gtrs Rhym Gtr Voice, organ, gtr fill Gtr fill Bu vocals Set Set Set Is This Love, Marley, 0-1:00. Synced, panned delays in drums, organ. Nice example of height. Slap delay on guitar fill hard left. London Organ on far left Bass, guitars Organ, Various percussion Voice, Organ, Bass, Guitars Percussion Percussion bright lead delays gtr Rocky Raccoon, Beatles, 1:30-3:30. Why not LA? It's too raw. Notice the extreme image. Many of their early works were 4 track, so things were grouped at extremes. Garage Harmonica, Accordion Guitar Accordion? Set! Vox Acoustic Guitar Harm Bass! Honky-Tonk Accord Janie's Got a Gun, Aerosmith, 4:25-5:30. Excellent depth, width, and height. Vocals in background: imitating reverse Concrète? Dynamic change in image (voice and synth). NY (Ln?) Synth Strings Popped Guitar Guitar harmonics? HL Harmonics BU Vocals Vox Popped guitar Synth Strings BU Vocals Set Bass Set BU Vocals Pseudo Reverse BU Vocals Set Voodoo Child, Angelique Kidjo, 1:15-1:40. Excellent depth, width, and height. Several examples of filtering. The didgerido-esque instrument is a filtered sharp wave with random LFO cutoff. Unusual EQ of lead voice. Envelope follower on wa-wa fill. Sweeping sound dynamically panned. I like this example because of the varied treatment of vocals. Every voice has a different EQ, pan position, and depth. UK BU Vocals, Wa-Wa Didgerido-ish synth Vocals Vocals Lead Vocals Vocals, wa-wa Vocals BU Vocals Didgerido-ish BU Vocals You Ain't Seen Nothing Yet, BTO, 0-0:45. I'm saying NY because of the wave form: very consistent. Great example of dubbing (even playing different chords) and synced delays (lead guitar). The lead guitar is not a fill; never really exchanges with the lead. When it's there, it is the lead. NY None! All other guitars None! Guitars Guitar Lead Guitar Voice Lead Guitar delay Guitar Cymbals Bu vocals Set Set Cymbals Set Louis Loon, Bare Naked Ladies 0-1:00. I love this example. Piano and banjo are both rhythm, but panned hard left and right. Note that they pull a little Zeppelin trick; the drums and bass are playing perfectly normal, standard 1 and 3, but offset by one beat (4 and 2). Very clever. Very dry, just a little reverb on the lead. BU vocals are slightly deep in the field. NY Vocals and bells Piano, Banjo Piano BU vocals Lead BU Vocals Banjo Set Set Like a Rolling Stone, Dylan. Entire track. Garage Sometimes the guitar Detuned Everything else piano

Everything Bob

Everything else

Set!

Sunshine Superman, Chaise Lounge. Absurd amounts of reverb, delays, and compression. Distorted guitars, clean blues guitars, overdubbing a flawless, silky voice that has absolutely no reason to be dubbed. Genius. London Bluesy and distorted guitar None None Voice Horns Set (snare, kick, ratchet), Bass Horns, Distorted guitar Voice Set Set The New Kid in Town, Eagles. 1:30-2:00; Great example of elements. Set is panned mostly right. Not much on the left. LA Rhodes Piano Off beat guitar, right Organ

111

Rhodes and lead guitars

Strummed guitar

Organ BU Vocals

Lead Vocal

Organ delay Set Rhythm guitar BU Vocals Set Life in the Fast Lane, Eagles. 0:30-1:15; Lots of doubling; voice (which would sound much worse undoubled), guitars, lead and rhythm. The riff during the chorus is not a fill or lead, it's too repetitive. It's a rhythm. NY Guitars trade off rhythm and fill Guitars (opening riff) Voice Guitars Rhythm Rhythm guitars Lead (and doubled takes) Rhythm guitars Rhythm guitar Set Guitars Set Right Down the Line, Gerry Rafferty. 0:30-1:30; Textbook foundation, rhythm, pad, lead, sparse fill. LA? Very sparse, guitar Guitar, far left Organ, close right Vocal Guitar Set Lead (all overdubs) Organ Guitar delay Set, Bass Set How Sweet it is To Be Loved By You, James Taylor. Choruses LA Piano, Sax Guitar (very subtle) Set Set Pretty much everything else

Strings Set

None Set

A Day in the Life, Beatles. complete; Unusual delay (un-synced, probably tape delay), kit is close mic'd with lower heads removed, heavy compression. Dynamic panning (voice moves to center, then right). Phone ringing was a joke, happy accident. Generative "chaotic" transition, Paul's instructions were to start low and play high. Middle segment mic is intentionally muffled. Contrasting depth between lead of second section and "ahs" in the background. Image reverses for third section. Concrète loop after pause at the end. First example of time code? Good grief, just read Emerick's book. London Drums Shaker (not the guitar) None None Shaker Set Bass Set Vocal piano Set Anything from Come Away With Me, but let's do Cold, Cold Heart, Norah Jones. 1:00; Perfect example of an intimate, close mic position on all instruments, very little reverb LA Piano Guitar (left) Steel Guitar None Guitar Piano Voice Steel Guitar ? Piano Bass Oh What the Heck, Let's Do Turn Me On, Norah Jones. 0:44; It's such a great example. LA Nothing? Piano/Guitar Organ None Guitar Bass Organ Piano Voice Set Set Englishman in New York, Sting. 2:34; Height, depth, width. Seven Days is very similar. LA/London? Saxophone, Piano Pizz. Strings The reverb? None Percussion Saxophone Voice Most of the set Piano Strings Strings Set Set Give Me , Sting. 2:34; Height, depth, width. Seven Days is very similar. LA/London? Saxophone, Piano Pizz. Strings The reverb? None Percussion Saxophone Voice Most of the set Piano Strings Strings Set Set

• Exercises 12:1 Reproduce this chapter using examples from your own collection.

112

13.

SC 3 Quick Start Just a quick introduction. Skip it if you want.

When you first launch SC a window appears with three sections. On the left is an editor. This is the window where you type and edit lines of code. The right side of the window is split into a help browser and a post window that prints information about the program during startup. This is called the IDE (integrated development environment). It's new to 3.6. You can detach, close, or rearrange these three windows. •

SC Intro

A second program, sclang (the interpreter), will launch at the same time. You won't interact with this element except when it presents a window (such as a signal plot, open file dialog, or scope). This collection of windows is called the client, where you type and develop code. Those commands and algorithms are sent to the server, which actually makes the sound. The text editing functions (select, copy, replace, etc.) are similar to any basic editor. But there are editing features that are especially useful for (and unique to) editing code; toggling comments, moving a line down or up, or moving the cursor by block, region, or empty line. Figure 13.1 shows the defaults. An asterisk shows those that I had to change in the preferences dialog to be more OS X friendly. If you've never written code their value is not immediately apparent. Review these once you have a few patches under your belt. Fig. 13-1 SuperCollider IDE Shortcuts

⇧⌘ P ⌘= ⌘E ⇧⌘ R ⇧↩ ⌘↩ ⌘. ⌥⌘ ↓ (↑) ⌃⌘ [ ( ] ) ⌘ F, G, R ⇧⌘ G ⌘L ⌘ ]([) ⌥⌘ ] ( [ ) ⌥⌘ 0 ( 9 ) ⇧⌘ [space] ⌘/ ⇧⌘ D ⌘D ⌘M

Clear Post Window Split Window at Bottom Command Line for Quick Code Evaluation Select Region Evaluate Current Selection or Region Evaluate Current Region Stop All Playback Duplicate Line Down (Up) *Move Line Down (Up) Find, Find Next, Replace Find Previous Go To Line *Go To Next (Previous) Region *Go To Next (Previous) Empty Line *Go To Next (Previous) Block Engage Auto-Prompt (Shows Arguments) Toggle Comment Enter Text To Look Up In Documentation Look Up Documentation for Text Under Cursor Show Server Meter

113

•

Hello World

International geek protocol binds all programming authors to use "hello world" as the first coding example. Le voici. Type this single line precisely into the left window, position the cursor anywhere in the line (or select it), and press ⌘[enter]. Check the post window. You should see the program echo your first line of code by simply printing it to the screen. 13.1::

Hello World

"Hello World"; ::

You are now a programmer. This next example is a bit more pretentious; demonstrating how far we've come since I first saw "Hello World" flash on a computer screen. They use Apple's speech algorithms. It may take a second to engage. 13.2::

Hello Hal

"Hello World".speak; "I've just picked up a fault in the AE35 unit. It's going to go a hundred percent failure within seventy-two hours.".speak; ::

You are now a 20th C programmer, 2001, to be precise. Note that each line is terminated with a semi-colon. In computerese this indicates the end of an expression. Example 13.3 incorporates some music math. Run each of the following lines separately, but in sequence. As you type you will notice some contextual menus popping up offering auto-completions, matching enclosures, and information about the items you're typing. These will make sense as you become more familiar with the program. 13.3::

Music Math

a = 5 + 4; b = (440 * 3/2); c = 60.midicps.round(0.1).asString; a.asString.speak; speak("The frequency for E5 is " ++ b.asString); ("The frequency for middle C is " ++ 60.midicps.round(0.1).asString).speak; ::

There is a good reason for calling this code. It's cryptic, a new language, just like French or, maybe more apt, Greek. Once you understand it, it can be read just as you would a sentence 114

in English, although sometimes you read from the inside out. Here is a translation of the last line in the example above: Convert the MIDI (midi) number 60 into cycles per second (cps), then round that value to 0.1, convert it to a string. Concatenate (++) with "The frequency…" and speak the results. The examples above don't make use of the Server because they don't generate sound. The next will. Turn it on ("boot") using ⌘B. You only need to start (i.e. boot, or power on) the server once for each session and leave it running. In the lower right hand corner you should now see green numbers in the server monitor. The first two indicate how much processor is being used. This is important information, as we will most likely be stretching the limits of your hardware. Before we start a sound, it's a good idea to know how to stop (true for many things). ⌘-period stops all playback (but leaves the server running). This text assumes you will stop playback after each example. Type and execute the following line. 13.4::

First Patch

// Start the server first: Language/Boot Server {SinOsc.ar(LFNoise0.kr([10, 15], 400, 800), 0, 0.3)}.play; ::

Translation: play the stream of values generated with the function ( {…} ) that uses a control rate (kr) low frequency noise generator (LFNoise0) with a range of 400 above and below 800 at a frequency of 10 times per second in the left channel, 15 in the right, to give values for the frequency of a sine oscillator (SinOsc) at an audio rate (ar) at a volume of 0.3. Example 13.5 is more complicated, with several lines of code. This time, as you type, see if you can make sense of the auto-completion features. Note that indents are applied automatically (according to code formatting conventions), and that enclosures (parentheses, brackets, and braces) turn red when the cursor is next to its matching enclosure. Try pressing return when the auto-completion offers the correct word. Finally, the program automatically colors words based on syntax. All of these features are extremely helpful when developing code. This time position the cursor anywhere in the text (on any line) and press ⌘ [return] 13.5::

Second Patch

( { RLPF.ar( LFSaw.ar([8, 12], 0, 0.2), LFNoise1.ar([2, 3].choose, 1500, 1600), 0.05, mul: 0.4 ) }.play ) :: 115

•

Error messages

By now, you may have encountered an error message. (The examples in this chapter will be omitted from the code file—available by request to those who prove that DRMs are unnecessary and buy the text—so that you can learn how to resolve error messages.) You will quickly notice that the compiler is unforgiving when it comes to syntax, spelling, and punctuation. Every comma, period, upper and lower case, and semicolon needs to be correct. If not, an error message will appear in the data window. Figure 13.2 shows a common error. Fig. 13-2 Error Message

ERROR: syntax error, unexpected INTEGER, expecting ')' in file 'selected text' line 1 char 40: {SinOsc.ar(LFNoise0.kr([10, 15], 400 800), 0, 0.3)}.play; ^^^ ----------------------------------ERROR: Command line parse failed nil

"Parse error" usually is the result of a typo. The third line of the error message tells you the line (1) and character (char 40) of the mistake. The fourth line places carets (^) at the position of the error. But the error is often several characters or lines before that. Compare the line in the error message to the original to see if you can spot the problem. There is a missing comma. Don't be discouraged with error messages. They are normal. While I was writing one of the examples below (about 12 lines of code) I had to run and correct errors six times (not just six errors, but running it, correcting it and running it again six times). I should probably be more careful, but my point is, it's normal, like spell check. •

Elements of Code



It is essential that you learn to identify these elements code.

Enclosures

Parentheses define argument lists, braces define functions, and brackets define arrays: (0, 100, 555), {arg next; next = 100}, [100, 200, 500]

Objects

Objects do something. While technically everything in SC is an object, for the next few chapters we will focus on objects that are paired with a message. These include upper case words, numbers, anything in enclosures, and quoted text: Synth, SinOsc, LFNoise, EnvGen, [1, 2, 4], 5, 6.7, "string", {rand(100, 200)}, (4 + 25), etc.

Messages

Messages tell the objects what to do. These are lower case words, sometimes on their own, but usually connected to objects by a dot (.): play, scope, ar, max, rand, midicps.

Arguments

Arguments tell the object how to do the message. They are a list of items separated by commas, enclosed in parentheses: (1, 2, 4), ("string", 45, myVar), ({function}, 0.1, [1, 2, 3])

UGens

Unit generators are the combination of an object, message, and argument list that result in some type of output. UGens are analogous to antique synthesizer modules which are connected together to create a patch. Examples of UGens are LFNoise0.ar(0.5), SinOsc.ar(440, 0, 0.4), Sequencer.ar(s, r, j) 116

Arrays

Similar to argument lists, but used very differently, they are a collection of items that are used as a group. Each item is separated by commas, enclosed in brackets: [1, 2, 3], ["C", "D"], [a, b, c]

Functions

Anything enclosed in braces: {arg next; next = 100}, {possibly many pages of code}, {SinOsc.ar(60.midicps)}

Variables

User defined names of memory locations, beginning with lower case letters (not a number), contiguous (no spaces), and often words strung together with caps in the middle: pitchClass, nextCount, freqArray, myVar, a1, start4, etc. Local variables are declared using var myVar1, myVar2;. Global variables a through z can be used at any time and work throughout all patches. You can also declare global variables by simply preceding the name with the tilde character: ~myGlobalVar.

Floats

Short for a floating point number: 1.235, 167.24, 78.127489273. The "point," or decimal, floats around to different positions. Values below 1 must be expressed with a leading 0: 0.123, 0.15, 0.5.

Integers

Integers are whole numbers: 1, 56, 109. They have no decimal.

Expression

A line of code punctuated by a semicolon. "Hello World".speak;

Comments

Comments are identified with two slashes; "//". Everything from that point to the end of the line is ignored by the compiler. They are used to clarify sections of code. They can also be offset with /* (start) and */ (end).

•

Enclosures (parentheses, braces, brackets)

 Enclosures are used to group things together and indicate the order of execution. Each opening parenthesis, bracket, and brace has a matched closing parenthesis, bracket, and brace. This is why the editor has a built in aid for seeing the matches. The compiler runs or evaluates the code from the inner most matching enclosures, then works outward. When items are nested you often will see six or seven open enclosures as you read left to right, then all of the closing enclosures later in that line, as illustrated below. (This is a fabricated example for demonstration. It won't run.) Fig. 13-3 Matched Enclosures

{Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play [ ] {Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play ( ) etc. {Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play {Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play {Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play {Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play {Saw.ar(Pulse.ar(max(rrand(rrand([12,15],5))),400,800),0,0.3)}.play 1 2 3 4 5 67 7 654 3 21

The outer most enclosures of any section of code define a region. That is why most of the examples are encased in opening and closing parenthesis. ⇧⌘R will select the region. Choosing ⌘-return will select that entire section (block) and evaluate it. 117

White space (spaces, returns, tabs) are often used to clarify matching enclosures. SC ignores all white space, so it can be used to visually line up enclosures and argument lists. In practice, one should strike a balance of concise code and open clarity. Variables, discussed later, will also clarify coding. 13.6::

Balanced Formatting

// too dense? {SinOsc.ar(SinOsc.ar(3,0,500,1000),mul:max(0, LFNoise1.kr([12, 14])))}.play // clearer, but too spread out? ( { SinOsc.ar ( SinOsc.ar ( 3, 0, 500, 1000 ), mul: max ( 0, LFNoise1.kr ( [ 12, 14 ] ) ) ) }.play ) // clear ( { SinOsc.ar( SinOsc.ar(3, 0, 500, 1000), mul: max(0, LFNoise1.kr([12, 14])) ) }.play ) // clarity using variables. ( 118

{ a = SinOsc.ar(3, 0, 500, 1000); b = LFNoise.kr([12, 14]); c = max(0, b); SinOsc.ar(a, mul: c) }.play ) ::

 Look again at all the examples so far and practice identifying each of the components we've discussed. I'll do the first patch: {SinOsc.ar(LFNoise0.kr([10, 15], 400, 800), 0, 0.3)}.play; The objects are SinOsc, LFNoise0, the function ({…}), the array ([10, 15]), and all the numbers. The messages are play, kr and ar. Functions and arguments are a little harder to spot. All of the text {SinOsc … 0.3)} is a function (everything between { and }). Arguments can be confusing if they are nested. The arguments for LFNoise0.kr are ([10, 15], 400, 800). The array [10, 15] is the first argument, 400 the second, 800 the third. For the SinOsc.ar the arguments are (LFNoise0.kr([10, 15], 400, 800), 0, 0.3). The first is LFNoise0.kr([10, 15], 400, 800) (yes, the entire UGen), the second is 0, the third is 0.3. •

Arguments

Arguments are control parameters. They are similar in function to the knobs on analog synthesizer that change frequency range, filter cutoff, amplitude, sequencer speed, etc. In example 13.7 the arguments for LFNoise1.ar are (10, 400, 800). Change these three values and run the code to see how they change the sound. Try 15, 25, 30, or 40 for the first argument. Try 100 and 700, 30 and 100, 1600 and 1700 for the second two arguments. 13.7::

Arguments

{SinOsc.ar(abs(LFNoise1.ar(10, 400, 800)), 0, 0.3)}.play ::

The knobs on synthesizers are labeled. These controls typically are not. How do you know what the arguments represent, or how they change the sound? The first step is to look up the help file; highlight the item you want help with and press ⌘D. •

Object.Message

The examples above are difficult to understand because of the cryptic syntax but also the foreign words (SinOsc, LFNoise1). It would be helpful to use more familiar, though fictitious, objects and messages to explain how they work. (These examples won't work in SC!) If you are comfortable with these terms, skip this section. If SC were designed to test recipes for sandwiches it might have the fictitious objects Sandwich and Tofu. Every object understands a collection of messages. The messages tell the object what to do. Likewise, there are many objects that understand any given message. For example, let's assume that Sandwich understands three messages: make, cut, and bake, and that Tofu understands three messages: bake, fry, and marinate. The power of object-oriented languages lies in the way you can mix and match messages and objects. The syntax for sending the make message to the Sandwich might be this: 119

Sandwich.make;

If you wanted the Tofu to be baked you might write: Tofu.bake;

Notice there are no argument lists to describe what the sandwich is made of or how the tofu is baked. Most messages have default values. Try running the lines in example 13.8. The first uses no arguments in the .ar or .play messages. Next, run the example with an argument list for both .ar and .play. 13.8::

SinOsc defaults

{SinOsc.ar}.play {SinOsc.ar(800, 0, 0.1)}.play(s, 0, 10) ::

The first result is a sine tone at 440 Hz, 0 phase, 1.0 amplitude, played on the default server, output bus 0, no fade. Those are the defaults for SinOsc and play. Often you are able to use one or two of the defaults, but rarely will you use a message with defaults only. The second example overrides the defaults and plays a sine tone at 800 Hz, 0 phase, 0.1 amplitude, on server s, bus 0, with a 10 second fade in. To see the help file for SinOsc, position the cursor inside it and press ⌘D. In each of the help files are prototypes of all the messages (with argument lists) understood by that object. Sandwich and Tofu might be documented this way: Sandwich *make(vegArray, bread, meat) *cut(angle, number) *bake(temp, rackLevel) Tofu *bake(temp, baste, rackLevel) *fry(temp, length, pan, oil) *marinate(sauce, time)

It is important to understand that argument lists are not interchangeable between objects. The bake message used with Sandwich has two arguments, while the Tofu bake argument has three. Not understanding this, and using the same arguments with a message to different objects is a common beginner error and will have unexpected (and possibly interesting, but don't do it) results. When bake is used with Sandwich, as in Sandwich.bake(20, 2) the 2 is rack level, while in Tofu.bake(20, 2) the 2 is baste time. Now that we understand what the arguments for Sandwich.make are, we could put together a Sandwich with this mock code. 120

Sandwich.make([lettuce, tomato, pickle], rye, chicken) or Sandwich.cut(90, 1) and Tofu.marinate(peanut, 160)

The first line will make the Sandwich using an array (list) of vegetables, bread, and chicken. The second line will make one cut of the Sandwich at an angle of 90 degrees. The Tofu will be marinated with peanut sauce for 160 minutes. Another powerful aspect (the whole point, really) of SC and object oriented languages is that everything is an object, and you can substitute them freely. Instead of chicken as the third argument, the entire section of Tofu code could be used (nested). Sandwich.make([lettuce, tomato, pickle], rye, Tofu.marinate(peanut, 160)))

When a program evaluates the code it begins with the inner most parts, and uses the results of those values to run the subsequent outer layers. In English, the example above might read: Marinate tofu in peanut sauce for 160 (minutes?). After marinating the tofu, use it as the meat (third argument) for a sandwich with lettuce, tomato, and pickle, on rye bread. It is possible to link messages. For example Sandwich.make.bake.cut would first make the sandwich (in this case using defaults), then bake it (using defaults), then cut it (with defaults). One object can be used as an argument for another instance of the same object. For example, you could write Tofu.marinate(Tofu.marinate(peanut, 60), 60). In this case, a batch of tofu will be marinated in peanut sauce for 60 minutes, then another batch of tofu will be marinated in that batch of marinated tofu (yum!). •

Exercises 13:1 Below are two patches. Read the help files associated with each object. You may not completely understand all the terms, such as mul and add, and how they work in the patch, but it's ok to suppress some questions and be satisfied with what you do know. In both, identify objects, messages, argument lists, functions, matched enclosures, floats, integers, and arrays. 13:2 Look at the help documentation for the second patch. See if you can make sense of the arguments in relation to the actual sound. Try to predict how the changes you make will change the sound before you run the code. I've commented ("//") values you might want to change. Feel free to experiment. There is indeed a danger of crashing the machine or getting error messages or possibly startling a resident pet. But otherwise you can experiment at will. If you make changes that result in an interesting sound you can easily save it to a file. To do this choose save as and name the file or copy and paste the patch into a new file. 13.9::

Experimenting with a patch

{SinOsc.ar(abs(LFNoise0.ar(10, 400, 800)), 0, 0.3)}.play ( 121

{ RLPF.ar( LFSaw.ar([8, 12], 0, 0.2), abs(LFNoise1.ar([2, 3].choose, 1500, 1600)), 0.05 ) }.play ) ::

13.10::

Rising Sines

// Rising Sines ( { CombN.ar( // Delay Ugen SinOsc.ar( // Sine Osc Ugen provides input for delay LFNoise1.kr( // Step Noise Ugen provides freq for Sine 4, // LFO of step noise, for "wandering" 24, // range in MIDI of step noise wandering LFSaw.kr( [8,7.23], //second LFO for "sweep" 0, 3, // range in MIDI of sweep 80 // offset in MIDI of sweep ) ).midicps, 0, 0.04 ), 0.2, // max delay 0.2, // actual delay 4 // delay decay ) }.play ) ::

122

14.

Concrète Procedures Splicing blocks and razor blades; no longer useful.

•

Concrète

To illustrate the impact recorded sound has had on music and communications, take a quick inventory of the music you listen do daily; live vs. recorded. Is it even close to 50/50? Would we have to extend the challenge to a week before we could include a single incidence of live performance? I've encountered students who have never attended a live concert of any kind, yet their collections contain thousands of works. Early recordings were used as parlor entertainment. Guests were shocked to hear, for example, horses in a room. I love classic science fiction, in particular tech spotting; taking note of what they got right, or how far off they were in other ways. (They were much too pessimistic.) No flying cars from the Jetsons, but video phones? There is one sitting next to me on the couch. Two that still seem to elude us are teleportation and time travel. Yet one could argue that recordings do both. They are time and space machines. The tape deck's intended contribution to the music world, reproduction, was quickly hijacked by the likes of Ussachevesky, Sheaffer, Leuning, McCartney, and the minimalist school. These pioneers were more interested in extending the capabilities of natural instruments—for example exploring the sound of a piano if the keyboard were several octaves lower, or higher— but were easily distracted by entirely novel treatment; sounds that had literally never been heard before, and were unique to the medium. The Beatles continued to explore Concrète treatment on nearly every album after Revolver, the most notable, striking, and ignored being Revolution 9. It would fit in nicely at most electro-acoustic concerts. Once dismissed as hair-brained, this technique is now common in Rock, a staple in Rap, Techno, Industrial, and similar styles. It is the foundation for sampling synthesis and sample libraries. Examples of Concrete treatment in popular media include Strawberry Fields and For the Benefit of Mr. Kite (Beatles), Another One Bites the Dust (Queen), pretty much everything Daft Punk does, but I like Face to Face, Vox Humana (Kenny Loggins), and the entire Year Zero album, NIN. I can't pick just one. Maybe God Given. There used to be a pronounced schism between purely electronic and Concrète composers. The two have since merged. The Concrète camp were drawn to the built in complexity due to the rich source material. Concrète and additive synthesis, (covered later— which can also be very rich and surreal) are my favorite electronic techniques.  Most think of Concrète as manipulating recordings of real sounds in artistic ways, any recorded sound is "set in concrete." That is in fact the inspiration for the term. Even a recording of a classical work with full orchestra is like a statue; set in stone. This definition reminds us that live music has a dimension recorded pieces do not; the potential for growth and change. A written work that has to be realized is not the performance, but the potential for a performance, or a failed performance, or a brilliant performance, or in the case of aleatory and improvisation, something completely new, never heard before. I believe too many people equate a recording with a performance. In a sense it has tainted our expectations: we attend a concert expecting what we heard on the CD. The Beatles were one of the first to 123

realize they could subvert the standard practice of releasing recordings just to bring people to concerts. They saw that the recordings could stand on their own merit. They became the medium, and the principal source of income. As digital rights management fails there could be a return to the previous model; give away the recording to attract audiences to live performance, which holds the promise of fresh material. Concrète introduced us to sounds that had never been heard before. Indeed all recordings, even those not intended for concrète treatment, have exposed us to rare and exotic sounds, previously only heard by natives of a particular environment. Exploring sounds unique to recorded reproduction is an important goal of this chapter. My first exposure to an EVI (electronic valve instrument) was a concerto that could have been, and probably should have been played on trumpet. I was disappointed that the performer never changed the timbre, never leapt six octaves, never bent the pitches, never fired off notes at the speed of sound. (Well, technically, yes.) Many electronic performances since have left me wondering; what is the point? I see the stage filled with mics, computers, wires, and speakers, but hear music that could have been performed before the electronic revolution. The power of Tchaikovsky, Bach, Vivaldi, and Hendrix lies in the exploitation of their chosen medium; the power of the piano, carefully crafted vocal lines, flashy violin arpeggios, controlled harmonic feedback. We should judge electronic music the same way; does it exemplify its unique power? For Concrète, we have to ask; what can digital manipulation do that can't be done on any other instrument? An additional appeal to concrète material is the inherent contextual baggage. If a sound is drawn from the human experience (and how can it not?), then the audience will bring that association to the work. You can dodge association through obfuscation or foreign languages, or you can embrace it as part of the work; juxtaposing, for example, a speech with iconic religious inflection over cartoon sound effects. Auto-tuning the news is an excellent example; it is precisely the intended hubris and arrogant context that makes vocalized parodies so effective.  Concrète is capable of four treatments (which we now take for granted) that cannot be performed on any other instrument; precise loops, pitch and time shift, reversal, and abrupt edits. •

Looping

… is rampant in popular media, evidenced by the Apple Loops library. Early loops were executed on magnetic tape decks, splicing a section of tape to itself then threading it around the room using mic stands for support. Tomorrow Never Knows, Beatles, is an early example of loops. (These were created by Paul on a newly purchased open reel tape deck. As far as I have read Paul worked this out on his own, without exposure to any of the art music experiments happening at Columbia. From Geoff Emerick's book: "Every tape machine in every studio was commandeered and every available EMI employee was given the task of holding a pencil or drinking glass to give the loops the proper tensioning. In many instances, this meant they had to be standing out in the hallway, looking quite sheepish. … Add in the fact that all of the technical staff were required to wear white lab coats, and the whole thing became totally surreal.") Looping appeals to me in two ways; the precision of repetition, and the rhythmic patterns possible with mathematically related loops. Our brains are finely tuned to recognize pattern. We recognize in electronic loops minute details. No other musical instrument is capable of such accurate repetition. 124

•

Pitch and Time Shift

… were, until only recently, inextricably linked. Change the speed of playback and the pitched changed. Today's wizardry can change one and not the other. (This is analogous to increasing the length of a book without changing the content.) Taking this process too far results in aliasing (in itself an interesting effect). Speed change with spoken samples borders on cliché, but other sounds—percussion, ambient noise, instruments, sounds from nature, still keep my attention. The Beatles also experimented with tape speeds. The solos on Misery were recorded with the tape at half speed, partly out of necessity (George was having difficulty with them), partly for effect. This is not readily apparent, but blatantly obvious when the track is played at half speed. But the backing track to Rain was intentionally recorded at a faster tape speed, then slowed down for effect, not out of convenience. •

Reversal

… also trickled from the experimental academic studios to popular recordings. Rain, Beatles (and nearly everything else on Revolver because of Rain) uses reversal. They hit on the idea because John mis-threaded a rough mix on his home machine. Other examples include Another One Bites the Dust, Queen, 2000 Light Years From Home, Rolling Stones and the entire left channel guitar solo in Castles Made of Sand, Hendrix. •

Abrupt Edits

… is another technique that is unique to recording technology. No natural sound or instrument can change it's timbre, subtly or radically, with millisecond accuracy. In Yellow Submarine, Beatles, Emerick snipped up the brass band section into small segments, threw them in the air, then spliced them back together again. His intention was to avoid copyright infringement, which worked, but also hit on a unique treatment. Daft Punk has built a career around sampling, abrupt edits and looping. Of course there are, and I encourage you to explore, excellent examples of concrète from the academic world. I use popular examples to illustrate their viability. These techniques were dismissed and even derided among academics. They were obviously wrong to do so. The 1998 documentary Modulations chronicles the link between experimental academia and the electronic styles that evolved in clubs and raves. There are two tools in Logic that give quick concrète results. They are designed as sampling synthesizers. Like recording technology in general, they are intended for exact reproduction of natural instruments, but can be repurposed for concrète. The first is the … •

EXS24

You can import samples and play them back on the keyboard for precise control of a large collection of loops.

125

 When you first launch the EXS24 there is no instrument assigned. (If there is an instrument, choose no instrument.) Pressing the edit button when no instrument is loaded will create a new instrument and open the editor. You can drag and drop samples onto this field. Use the bars just above the keyboard to specify which keys trigger a given sample. Set the Key field to where you want normal speed played back. In the local View menu choose Zone: Loop to turn looping on and to specify the start and end point (in frames) of each loop. Now you can play these samples on the keyboard and record them as MIDI events. They will play at different speeds for each key. The most interesting application, for me, is precisely coordinated loops. Try intervals such as a fifth, fourth, or octave. As discussed in Chapter 8, the samples will go in and out of sync at those ratios. •

UltraBeat

… has similar functionality with drag and drop simplicity. The advantage over the EXS24 is the built in looping. It can manage 24 separate samples, each with 32 step sequences, and virtually unlimited sets of patterns. It is a Concrète composer's dream. In a Logic session add an instrument track and insert UltraBeat in the I/O input menu. UltraBeat has three synthesizers to generate its drum loop sounds. We will work with the sampler at the bottom (see Figure 14.1, first image). In the Media/Library choose the Hip Hop 90s Kit preset or the Drag and Drop Samples preset. (I choose Hip Hop kit because most of the sounds are real samples that can be replaced, and it is pre-loaded with sequences and loops. Feel free to explore, others, but find something that uses primarily samples. Use Drag and Drop to start from scratch.) Press the power button, then the play button. Turn on pattern mode, set to toggle or one-shot trigger, then to the bottom left click on the pull down menu next to pattern to select one of the 24 sequences assigned to the lower octaves of a MIDI keyboard (C -1 through B 0). The far left of the UltraBeat window is a keyboard (see Figure 14.1), each key labeled with the name of the loaded sample, a mute, solo, and pan control. While playing back, load a new concrète sample by dragging it to the window displaying a wave form. Click the red arrow to reverse, click the piano key to test. Continue to experiment with different sequences and replacing existing drum sounds with your concrète material. This alone, if improvised, would be an interesting concert performance. The sequencer (Figure 14.1, second image) has 32 steps. Each can be shortened by dragging the bar just below the step numbers. Clicking on a number adds a gate to that step. A gate is a trigger with duration. The sound continues to play until the gate is released. A 4 second sample needs at least the same length of gate (though it can certainly be shorter). Once you have loaded and tested a Concrète sample, click on the numbers to add several gates. Click and drag across several trigger positions to create longer or shorter gates. Repeat for each new key and each new sample, building a collection of looped Concrète sequences.

126

Fig. 14-1 UltraBeat

To the right of the sample window is a complex dial (labeled OSC2) that allows you to modify the pitch, amplitude, and velocity. Experiment with the reverse switch as well as the pitch of the playback. Feel free to load the same sample into several key locations for different treatments (reverse, pitch, envelopes). In the lower left hand corner of the UltraBeat window click full view. This changes the display from synthesizer to a graph showing all sequences. Control click any track to reveal a handy set of editing features, including copy, paste, clear, randomized creation and replacement. No one takes me seriously, but seriously, you could stop coming to class and do an entire project with this plug-in alone. There is a mysterious grid icon just to the left of the MIDI key assignment pull down menu. Drag the small box to the arrange window to duplicate the pattern as a region. For a generative dimension, name all your samples 01.aif, 02.aif, etc., have a colleague collect new samples (perhaps before the concert from the audience) and also name them 01, 02, etc. Drag these new samples to your sample folder, relaunch the session and it will read and play the replacements. •

Concrète using SC

… is a little more involved, but the results, especially the granular treatment in the next chapter, are worth it. Since this is the first example in this chapter (and perhaps coming to this example afresh), remember to launch the server before running any examples. 14.1::

Loading an Audio Buffer

// Boot the server first b = Buffer.loadDialog; // Switch to sclang to open file [b.bufnum, b.numChannels, b.path, b.numFrames] // confirm that it opened correctly b.play; // test playing it back ::

Buffer.loadDialog brings up an OS X dialog (in the interpreter) which you will use to find an audio example. I suggest you use a mono, 16 bit, 44.1k aif file longer than 10 seconds. It 127

will play back only in the left channel. Stereo comes later. And while I'm suggesting, how about a two second clip from any cut out of Tom Waits' Bone Machine or a segment from Cage's brief interview on the activity of sound. The sound is stored at the variable b using the equals sign. The next line of code can be used to make sure the buffer loaded correctly. It displays the buffer number, the number of channels in the file, the path, and the number of frames. A frame is the collection of samples for all channels. So a 5 channel audio file will have five samples per frame, but that single frame represents a single sample of time. A one second stereo file will have 44100 frames, but 88200 samples. For now we will loop using PlayBuf. The help file for PlayBuf shows the third argument is the playback speed. Negative numbers will play backward. 14.2::

Pitch and Direction

// change the third argument to values between -2.0 and 2.0 {PlayBuf.ar(1, b, 1)}.play; // Be patient. You may have to listen a while. {PlayBuf.ar(1, b, LFNoise0.kr(1, 3))}.play; ::

The fourth and fifth argument of PlayBuf can be used to generate loops. The fourth is a trigger that resets playback to the start position. The fifth argument is the position, in frames, where the loop returns to. Example 14.3 is a loop browser. A MouseX returns values based on mouse position and is used to supply the playback position of each loop. It uses a minimum value of 0 and a max value of the total number of frames in the sample. Move the mouse to the left for the beginning of the sample, right for the end. The trigger argument is the frequency, in seconds, of the trigger to reset the loop. This essentially becomes the length (in time) of the loop. The example below uses Impulse. The LFNoise0 from the previous example has been removed so you can experiment with the playback start and trigger frequency without speed being changed. Replace it if you'd like. Try values between 0.1 and 4 for the trigger frequency. 14.3::

Loop, Playback Position, Triggers

// The numFrames message returns the number of frames. // It's used below for the MouseX min, max value. b.numFrames; // Impulse sends a single pulse as a trigger. // You will hear a tick {Impulse.ar(1)}.play // {PlayBuf.ar(1, b, 1, Impulse.kr(1), MouseX.kr(0, b.numFrames))}.play; ::

128

•

Envelopes

You might notice that some start positions produce a pop, or click at each loop. This is the same effect caused by poor edits; an abrupt change from positive to negative values. When fishing around for a start position, it's impossible to know where 0 crossings are, so the solution is to apply an envelope over each loop. Envelopes are essential components for instrument design. They describe the change of the sound over the duration of a single instance of the sound (usually amplitude, but also timbre and sometimes pitch). A piano, for example, will have a fast attack and a long decay. A violin can create complex envelopes, including a long attack and a very fast decay. Envelopes are set in motion with a trigger or gate. Pressing a piano key is a real world example of a trigger. It is a single event that sets the sound in motion. You have no control once the key is pressed. A gate is similar to an organ key. It will continue to produce sound until the key is released, at which point it will decay. Example 14.4 illustrates envelopes. (The plot windows will appear in the interpreter. You will have to switch back and forth.) The fist shows a plot of noise without an envelope. The second shows a millisecond of plotted noise to better illustrate the probability that the beginning and ends of the audio are not at 0 crossings. If this example were looped, you would hear a tick. The next two multiply the white noise source by an envelope generator. The first plots, so a trigger is unnecessary. The last uses a trigger (Impulse) at one time per second. 14.4::

Envelopes

// Steady state noise. {WhiteNoise.ar}.plot(1) {WhiteNoise.ar}.plot(0.001) // Same signal multiplied by an envelope with 0.1 second attack // 0.6 second sustain, and 0.1 second decay. {WhiteNoise.ar*EnvGen.kr(Env.linen(0.1, 0.6, 0.1))}.plot(1) {WhiteNoise.ar*EnvGen.kr(Env.linen(0.1, 0.6, 0.1), Impulse.kr(1) )}.play ::

•

Variables, Frequency and Duration

The single letters a through z are defined by SC as global variables during startup. Some of these are reserved by the program. The variable s is always used for the server. Example 13.3 used the variables a, b, and c. Variables store items that will be used several times in a patch or several patches. In these examples, the variable b retains the buffer for each new execution of a patch. Otherwise, it would require the load-buffer dialog to choose a file each time. Local variables are confined to the function where they are declared. They are used to link components of a patch together, and to organize code by breaking it down into smaller statements. In the example below, each section is stored in the variable mix. Think of it as a bowl into which you mix and combine ingredients; mix = eggs and flour; mix = mix + butter;. They are declared in the top line as var myVar1, myVar2, myVar3; 129

Periodic events have both a frequency and duration. They don't have to correspond. For example, a trigger with a frequency of 2 (times per second) will have a 0.5 second duration. Each event between those triggers can be independent of the trigger duration. They could have a duration of 0.5 seconds, but also 4 or 0.1 seconds. If 0.1, then the sound will die away between each event. If 4 seconds the events would overlap. Either way, there is a connection, and it is useful to set durations in relation to frequency. As discussed in an earlier chapter, frequency and duration are reciprocal. Invert each to convert one to the other; d = 1/f, and f = 1/d. A 10 second duration has a frequency of 1/10 (think of it as one in ten seconds). A 10 Hz frequency will result in a duration (between each cycle) of 1/10th of a second. To link the two, set the duration to some percentage of the frequency. For example, d = 1/f * 0.5. If the frequency changes the duration will change too, and will always be 50% of the time between each event. Example 14.5 introduces a handy convention. The first patch is everything between the first, the code that actually produces the sound, is everything from { // patch… and }.play;. But this entire patch is enclosed in parentheses. These parentheses have no real impact within the patch, but they define a region. You can double click on the top parentheses to select the entire patch, or use ⌘-return to select and execute. Try inserting the CombN used in the rising sines. Insert a line before the final mix; with something like CombN(mix, etc.). 14.5::

Variables

( { // patch without variables PlayBuf.ar(1, b, 1, Impulse.kr(1), MouseX.kr(0, b.numFrames))* EnvGen.kr(Env.linen(0.1, 0.8, 0.1), Impulse.kr(1)); }.play; ) ( { // same patch with shorter duration (higher frequency) Impulse PlayBuf.ar(1, b, 1, Impulse.kr(0.5), MouseX.kr(0, b.numFrames))*EnvGen.kr(Env.linen(0.05, 0.4, 0.05), Impulse.kr(0.5)); }.play; )

( { // Variables for clarification var mix, loop, duration; duration = 0.5; // duration in seconds of the loop loop = Impulse.kr(1/duration); // duration converted to frequency mix = PlayBuf.ar(1, b, 1, loop, MouseX.kr(0, b.numFrames)); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, 1, 1, duration); mix; }.play; ) 130

::

The last line of code is simply mix. This is called the return, or the final results of the function that are passed to the message play. Aside from spreading out this patch for clarity, the variables solve a critical problem. The envelope (in this case at least) needs to match the duration of each loop. In the second patch, not only did I have to change both Impulse frequencies from 1 to 0.5, but I had to calculate the attack, sustain, and decay of the envelope given a duration of 0.5. All that is done automatically in the last example. Mix, which is a single loop, is multiplied by the envelope, which in turn is scaled to the duration with the fifth argument of EnvGen; timeScale. The result is an envelope that changes for any given duration, and changes in the other channels (covered soon). Imagine doing 8 channels, each with a different duration, each with different attack, sustain, and release times. With a variable, when the value for duration is changed, the attack, sustain, and release is set to 10%, 80%, and 10% of the duration, whether 1/10th of a second or 5 seconds. Variables are also (albeit slightly) more efficient since this patch uses a single Impulse in two places rather than two Impulse UGens. •

User Defined Arguments

The problem with variables is they cannot be changed once sent to the server. Like existing Ugens, you can create your own arguments, which are variables that have a link to the outside world and can be changed once a synth is running. They are declared before variables, directly after the opening brace of a function. One additional component, the synth must be loaded into a variable so that we can refer to it when changing the parameter. (But see SynthDefs below.) The syntax is a.set(\argName, value). 14.6::

Arguments

// b = Buffer.loadDialog; ( a = { // Arguments for control arg speed = 1, position = 0.5; var mix, loop, duration = 0.5; loop = Impulse.kr(1/duration); // duration converted to frequency mix = PlayBuf.ar(1, b, speed, loop, position * b.numFrames); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, 1, 1, duration); mix; }.play; )

a.set(\speed, -0.5); a.set(\speed, 1.25); a.set(\position, 0.2); // must be between 0 and 1 a.set(\position, 0.75); a.set(\speed, rrand(-1.2, 1.2)); // execute these several times for different values a.set(\position, rrand(0, 1.0)); 131

::

•

Multi-Channel Expansion

 I love this trick: if any UGen argument, anywhere in a patch, is an array (e.g., [1, 3]), the entire patch is duplicated using the first value for the left channel, the second for the right. (There are some exceptions, discussed in later chapters.) If arrays are used for other arguments, they are likewise split between the two channels. This is an extremely efficient way to expand a patch into 2, 4, 8, heck, 16 channels. Steve Reich's Come Out illustrates both looping and phase shift, which became a staple of the minimalist style. The patch above is ready for similar treatment. We only have to change one value—the Impulse trigger—replacing the single 0.5 to an array of [1, 61/60]. This makes the duration of the left channel 1 second, the right 61/60 seconds, or slightly longer. Stop for a second and imagine what the resulting loops will do. Both channels will start in phase, slowly shift out of phase, then back over sixty seconds. In order to focus better on the phase shift I've replaced the MouseX.kr with a rrand, which picks a value between 0 and total number of frames. Change it back if you'd like. Also replace it with Line.kr(0, b.numFrames, 60) and it will not only move slowly out of phase, but also gradually advance across the entire buffer. Try running it several times. 14.7::

60 Second Phase

( { var mix, loop, duration; duration = [1, 61/60]; // duration in seconds of the loop loop = Impulse.kr(1/duration); // duration converted to frequency mix = PlayBuf.ar(1, b, 1, loop, rrand(0, b.numFrames)); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, timeScale: duration); mix; }.play; ) ::

You will know when the loop as come full cycle because it will sound like a mono signal; both left and right channels will be the same. Steve Reich slowed down one of the channels by pressing his thumb against the reel of tape. In SC, it's just as easy, but with much more flexibility and precision. Change the 60/61 to other ratios. May I suggest common interval ratios: 2/1, 3/2, 4/3, 8/5, 15/16, 45/32, etc. After working with this patch for a few days you may bolt up in bed realizing the flaw; a 1 second loop in the left channel and a 3/2 second loop in the right will have different durations, so the smoothing envelopes also should be adjusted. If, for example, the left channel is a frequency of 1 and the right has a frequency of 61/60, it's loop will be slightly shorter, or is it longer? and by how much? Well lay back down and rest easy. Since we used a variable it happens automatically. If the duration is an array, say [1, 1.234] then the left envelope will be 1 * the envelope, the right will be 1.234 * the envelope, because we used the timeScale argument. 132

•

Experiment

Now we can have some fun. First up are TRand, TIRand and TChoose, which return random values with each trigger. TIRand returns integers only, which will be useful for our ratios. TChoose selects items from an array. 14.8::

Generative Choices

( { var mix, loop, duration, rate, gen, start; gen = Impulse.kr(1/15); // every 15 seconds // choose another speed and direction at each gen trigger rate = TRand.kr(-2.0, 2.0, gen); // Choose values between -2 and 2 start = TIRand.kr(0, b.numFrames, gen); // start location duration = [1, 15/16]; // match the gen trigger; could I use a variable? loop = Impulse.kr(1/duration); // duration converted to frequency mix = PlayBuf.ar(1, b, rate, loop, start); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, 1, 1, duration); mix; }.play; ) ( { // Be patient; change happens gradually. var mix, loop, duration, rate, gen, start, baseloop, buffer; gen = Impulse.kr(1/20); // every 20 seconds baseloop = TRand.kr(0.5, 3.0, gen); // select a base loop time // choose another speed and direction at each gen trigger rate = TRand.kr(-2.0, 2.0, gen); start = TIRand.kr(0, b.numFrames, gen); duration = [baseloop, baseloop * TChoose.kr(gen, [3/2, 4/3, 5/4, 6/5])]; loop = Impulse.kr(1/duration); // duration converted to frequency mix = PlayBuf.ar(1, b, rate, loop, start); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, 1, 1, duration); mix; }.play; ) // Load a new buffer on the fly b = Buffer.loadDialog(bufnum: b.bufnum); // This will switch you over to the sclang interpreter. // To continue editing, or stop the sound, switch back to SuperCollider ::

The second differs slightly. The generative choice is lengthened to once every 20 seconds to allow each pattern to play through, but each pattern is shorter, between 1 and 4 seconds. The duration first takes a base loop time, chosen with a TRand, then the left channel uses that base 133

loop while the right uses the base loop multiplied by one of a collection of ratios, which is chosen by TChoose at each gen trigger. Finally, we wrap up the patch by loading a new buffer on the fly. •

Exercises 14:1 Modify any of the patches above to include arguments rather than variables so that you can change parameters on the fly. 14:2 In the studio, record samples of instruments or items to use in the experiments in the next few chapters. Save the files as 01.aif, 02.aif, etc., so that you can easily swap with other students. 14:3 With your prepared Concrète files handy, locate an existing preset in UltraBeat with built in loops. Start the loop and drag and drop your samples, gradually turning the drum beat into a sample beat. 14:4 Record yourself saying "come out to show 'dem." Import the sample into an empty EXS24 instrument. First, try to reproduce Reich's seminal work with a gradual phase shift. Continue with a study of phased patterns using intervals (octave, fifth, fourth, third, second) and no more than two notes at a time. 14:5 Finish the comments in the following patch, documenting what each line does and how it modifies the patch.

( { // This line ... var mix, loop, duration, rate, gen, start, baseloop, buffer; // Every 20 seconds gen... gen = Impulse.kr(1/20); // baseloop is... baseloop = TRand.kr(0.5, 3.0, gen); // rate changes... TRand chooses values... rate = TRand.kr(-2.0, 2.0, gen); // start resets the loop... TIRand returns... start = TIRand.kr(0, b.numFrames, gen); // duration is the duration of each... duration = [baseloop, baseloop * TChoose.kr(gen, [3/2, 4/3, 5/4, 6/5])]; // loop defines... loop = Impulse.kr(1/duration); // mix is used to combine... mix = PlayBuf.ar(1, b, rate, loop, start); // The envelope generator smooths... // The arguments are loop and duration, which... mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, 1, 1, duration); mix; }.play; )

134

15.

SC to Logic Recording direct to disk. Granular systems.

When Apple introduced a Graphic User Interface, the computer world became more accessible. But there was a trade off. Easier usually means slower, less flexible, and less depth. UltraBeat and EXS24 offer images, dials, icons, menus, and drop fields that greatly simplify building a sequence of loops. The disadvantage is being stuck with the interface as it was designed. It's like driving modern cars. You have a lot of control, but only what the manufacturer will allow. Despite incredible flexibility, sooner or later an idea will surface for which there is no dial or lever. SuperCollider is a car part warehouse, filled with not just spark plugs and gears, but machining tools with which you can make components from scratch. It allows you, requires you, to get your hands dirty. It does have GUI objects, which we will cover later. This chapter offers a few controls that are less involved. The first is a system for triggering events from the keyboard. The second shows how to link SC and Logic, to exploit the strengths of each: SC's flexibility and depth with Logics slick interface, routing, reverbs, and mix tools. •

KeyState.kr

… converts key presses from the computer keyboard into triggers. (See also; KeyboardWindow.) The first argument identifies each key with a keycode. Modifiers (shift, option, command) had no effect. It will still return the same number. You might expect a pattern (a = 1, b = 2, etc.), but there is none. Figure 15.1 maps them out, more or less as they appear on the keyboard, then in order. In Example 15.1 they are simply used as triggers to select a new Rand value. I use only three codes—left, right, and down arrow. 15.1::

Keyboard Triggers

( { SinOsc.ar( LFNoise0.kr( TRand.kr(4, 12, KeyState.kr(123)), // Left arrow triggers new rate mul: 6, // Right arrow triggers new octave add: TIRand.kr(4, 8, KeyState.kr(124))*12).midicps) }.play ) // If beginning from scratch // b = Buffer.loadDialog; ( { var mix, loop, duration, rate, start, baseloop, buffer; baseloop = TRand.kr(0.5, 3.0, KeyState.kr(125)); // Down arrow, new loop rate = TRand.kr(-2.0, 2.0, KeyState.kr(123)); // Left arrow, new speed 135

start = TIRand.kr(0, b.numFrames, KeyState.kr(124)); // Right arrow, new start duration = [baseloop, baseloop * TChoose.kr(KeyState.kr(125), [3/2, 4/3, 5/4, 6/5])]; loop = Impulse.kr(1/duration); mix = PlayBuf.ar(1, b, rate, loop, start); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, timeScale: duration); mix; }.play; ) :: Fig. 15-1 Keycode map

[1] 18 [2] 19 [3] 20 [4] 21 [5] 23 [6] 22 [7] 26 [8] 28 [9] 25 [0] 29 [-] 27 [=] 24 [q] 12 [w] 13 [e] 14 [r] 15 [t] 17 [y] 16 [u] 32 [i] 34 [o] 31 [p] 35 [ [ ] 33 [ ] ] 30 [\] 42 [a] 0 [s] 1

[d] 2 [f] 3 [g] 5 [h] 4 [j] 38 [k] 40 [l] 37 [;] 41 ['] 39

[z] 6 [x] 7 [c] 8 [v] 9 [b] 11 [n] 45 [m] 46 [,] 43 [.] 47 [/] 44 [space] 49 [ [ ] 33 [ ] ] 30 [-] 27 [,] 43 [;] 41 [.] 47 ['] 39 [/] 44 [\] 42 [=] 24

[←] 123 [→] 124 [↑] 126 [↓] 125 [space] 49

[↑] 126

[←] 123

[→] 124

[↓] 125

[0] 29 [1] 18 [2] 19 [3] 20 [4] 21 [5] 23 [6] 22 [7] 26 [8] 28 [9] 25

[a] 0 [b] 11 [c] 8 [d] 2 [e] 14 [f] 3 [g] 5 [h] 4 [i] 34 [j] 38

[k] 40 [l] 37 [m] 46 [n] 45 [o] 31 [p] 35 [q] 12 [r] 15 [s] 1 [t] 17

[u] 32 [v] 9 [w] 13 [x] 7 [y] 16 [z] 6

By now you should be generating material worth saving, to be used on its own or compiled into a collage style composition using Logic as an editing tool. There are several ways this can be done. The first is … •

Recording to Disk

The quickest method requires a GUI window for the server that can be created with the code s.makeWindow. When you boot the server, it is automatically assigned to the variable s. This variable can be used to send commands to the server, such as creating a GUI, starting a recording, or changing record options. Once created, simply click on the record button to start and stop the recording. Note also the mute and volume controls, Quit, and CPU information. When the server GUI is focused you can press 0 to reset the volume to 0, s to bring up a signal scope, m to mute, l to bring up a level meter, and f to open a frequency spectrum.

136

Fig. 15-2 Server GUI

Recording this way may be complicated by a number of factors. The first occurs if you are using mouse controls; the slightest change in position will alter the audio considerably. Relocating it to hit the record button will defeat the purpose and you may never find that precise location again. To remedy this, click on the server window and note whether the record button is highlighted with a border. If it is, you can start and stop recording with the space bar. If it is not, try pressing tab repeatedly until it is. (You may have to turn full keyboard access on in the system preferences.) Next, the server window belongs to the sclang application, so you will have to switch to it before engaging record. The last complication is more critical in a multi-user (and tightly administrated) lab: where are the files saved? Do you have write permissions for that folder? The default location is the Music folder of the home user. Another solution is to set the record directory explicitly. (You can also specify a file name, but this is more work, and slightly dangerous. If you neglect to change the file name each time you want to record your previous recording will be overwritten without warning.) Change your_user_name to, uh, your user name. 15.2:: Options for Record thisProcess.platform.recordingsDir = "/Users/Change_to_your_user_name/Desktop/" s.prepareForRecord; // or Server.local s.record; s.stopRecording; // Run these variations before to change file formats. s.recSampleFormat = "int16"; // change format s.recChannels = 1; // change from the default stereo // saves file with specific name in the default folder, will be overwritten if repeated s.prepareForRecord("myAudio.aif"); // Specific location s.prepareForRecord("/Users/your_user_name/Desktop/myAudio.aif"); ::

Once saved, these files can be opened in Amadeus or dragged into a Logic session. A more direct method is to record directly into Logic by connecting … •

SC to Logic using Soundflower

… which is an open source plug-in that allows you to change the routing of applications within the operating system. (See also: JackPilot.)

137

(Install Soundflower.) Launch Logic and SC. In Logic, open the audio preferences dialog. Choose Soundflower (16ch) as the input source. Add four audio tracks, set the inputs to 1, 2, 3, and 4, and click on the I to monitor the input. The whole point of this exercise is to combine the features in Logic with the sounds created in SC. To that end, create two auxiliary sends for all four tracks. On one of the resulting auxiliaries add a reverb, add a delay designer on the other. In SuperCollider, run the lines in Example 15.3. The first line has two statements. The first sets the output to Soundflower. The next reboots the server. The next line sends a test tone to the first four channels. You should see signal in the Logic tracks and hear signal in your speakers. This may take a little tweaking to get it right. Take your time. I can wait. Note that all buffers are cleared when the server is rebooted. You will have to load them again. Once you get the test-tone to work, try some of the previous patches. To switch the system back to the default audio device, just use "nil." 15.3:: SC to Soundflower to Logic s.options.device_("Soundflower (16ch)"); s.reboot; // Test tone {SinOsc.ar([400, 600, 800, 1000])*0.3}.play// // To switch back to the system default device, use nil s.options.device_(nil); s.reboot; // Note that if you've restarted the server to use a different routing, // you will have to reload all buffers. We will use two buffers in the next examples. b = Buffer.loadDialog; c = Buffer.loadDialog; ::

Add two more tracks in Logic Pro. Pan each track to a different position: hard left, left, barely left, barely right, right, hard right. The next patches expand previous examples to illustrate the possibilities using the outputs 0 through 5 of SC, and inputs 1 through 6 of Logic. (You read correctly. Computers start counting at 0, humans start at 1. Sorry.) You can change the outbus simply to route it to different tracks in Logic. For example, working in stereo you might record a patch using outbus: 0 to record to tracks 1 and 2, then switch to outbus: 2 to record to tracks 3 and 4, and so on. When working with generative looped concrète the output levels can change drastically. You may stumble on an interesting loop and mouse position that uses a quiet part of the original sample. Remember there is a volume adjustment in the server window. If you turn on full keyboard access you can tab to the slider and use the arrows to increase or decrease the output (so that you don't have to move the mouse from the position you want to capture). When you find an interesting sound, simply arm the logic tracks and click record. 138

•

Real-Time Analysis

… is another advantage to using SoundFlower. You can link SC to both Logic and Amadeus Pro and use analysis tools in either. In Amadeus Pro, open any of the real-time analysis windows under the Analysis menu. Click on Soundflower as an input and also Playthrough. Every patch you run in SC will be analyzed in the waveform, sonogram, spectrogram, etc. It's a slick and simple way to get a visual reinforcement of a particular process. It's also handy for teaching. Example 15.4 shows a number of variations based on Reich's original idea for… •

Come Out

Some of them use two buffers. But I'm not convinced it is effective. (Technology invites complexity. But remember sophisticated treatment is not intelligent treatment. Two buffers are difficult to focus on.) I prefer the variations that use one. I collect the buffers, as well as the buffer durations in two global arrays for convenience. Items in an array are recalled (referenced) with the syntax array.at(n), so in this example, ~durs.at(0). Remember that computers begin counting with 0. In addition to the global variables a through z, you can create your own by preceding a variable name with the tilde (~) character. Here I also introduce the commented alternatives technique. That is, a line of code is followed by an alternative line, but commented. To try this alternative, remove the comments. It is not necessary to comment the line that it replaces, since the new line will replace the one above when executed. 15.4:: Come Out Variations // Note that if you've restarted the server to use a different routing, // you will have to reload all buffers. b = Buffer.loadDialog; c = Buffer.loadDialog; // Collect the buffers into a single array. ~buffers = [b, c]; ~durs = [b, c]; // Test to see if you are getting signal to all channels in Logic. // Modify this to two items if you're working in stereo. {SinOsc.ar([400, 600, 800, 1000, 1200, 1400])*0.3}.play // Test the buffers in two channels. {PlayBuf.ar(1, ~buffers)}.play ( { // This uses a single loop, but different ratios or rate // be sure to engage down, left, and right arrows 139

var mix, loop, duration, rate, start, baseloop, buffer; baseloop = TRand.kr(0.5, 3.0, KeyState.kr(125)); // Down arrow rate = TRand.kr(-2.0, 2.0, KeyState.kr(123)); // Left arrow start = TIRand.kr(0, b.numFrames, KeyState.kr(124)); // Right arrow // start = LFTri.kr(1/60, 0, 0.5, 0.5)*(c.numFrames); duration = 1; // duration = baseloop*[1, 3/2, 4/3, 5/4]; // duration = [20, 21, 22, 23, 24, 25]/20; loop = Impulse.kr(1/duration); mix = PlayBuf.ar(1, b, rate, loop, start); // mix = PlayBuf.ar(1, c, rate*[1, 3/2, -3/2, 2], loop, start); // mix = PlayBuf.ar(1, ~buffers, rate, loop, start); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, timeScale: duration); mix*0.3; }.play(outbus: 0); ) // Load new buffers on the fly b = Buffer.loadDialog(bufnum: b.bufnum); c = Buffer.loadDialog(bufnum: c.bufnum); ( { // Single loop, more gradual phase shift over 30 seconds, six channels. // This one uses a mouse control var mix, loop, duration, rate = 1, start, baseloop = 1, buffer = b; start = LFTri.kr(1/[60, 61, 62, 63, 64, 65], mul: 0.5, add: 0.5) * buffer.numFrames; baseloop = [30, 31, 32, 33, 34, 35]/30; baseloop = MouseX.kr(1, 25); // Grains duration = 1/baseloop; loop = Impulse.kr(1/duration); mix = PlayBuf.ar(1, buffer, rate, loop, start); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, timeScale: duration); mix*0.3; }.play(s); )

( { // Single loop, start position phase shift. // Mouse control size of loop and speed of shift. var mix, loop, duration, rate = 1, start, baseloop = 1, buffer = b; start = LFTri.kr(MouseY.kr(1, 60)/[60, 61, 62, 63, 64, 65], mul: 0.5, add: 0.5) * buffer.numFrames; baseloop = MouseX.kr(1, 25); // When loop is very small, they become "grains." duration = 1/baseloop; loop = Impulse.kr(1/duration); mix = PlayBuf.ar(1, buffer, rate, loop, start); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), loop, timeScale: duration); mix*0.3; 140

}.play(s); ) ( { // This example uses channels 0 and 1, adjust panning. var srate = 44100, start = 0, end = 3, duration, rate = 1, signal; duration = abs(end - start); // or // end = [2.3, 3.5]; signal = BufRd.ar(1, b, // Buffer 0 LinLin.ar( LFSaw.ar(rate/duration, 1), -1, 1, start, end)*srate )*EnvGen.kr(Env.linen(0.01, 0.98, 0.01), timeScale: duration, gate: Impulse.kr(1/duration)); SinOsc.ar(LFNoise1.kr([0.4, 0.43], mul: 200, add: 200))*signal; // or SinOsc.ar(LFNoise0.kr([12, 15], mul: 300, add: 600))*signal; // or // SinOsc.ar(LFNoise1.kr([0.4, 0.43], mul: 500, add: 1000))*signal; }.play(s) )

// Pulsing in and out ( { // Also only stereo. Adjust panning var srate = 44100, start = 0, end = 3, duration, rate = 1; var pulse; pulse = [6, 10]; duration = abs(end - start); BufRd.ar(1, b, // Buffer 0 LinLin.ar( LFSaw.ar(rate/duration, 1), -1, 1, start, end)*srate )*EnvGen.kr(Env.linen(0.01, 0.3, 0.01), timeScale: duration/pulse, gate: Impulse.kr(pulse/duration)); }.play(s) ) // Filtered ( { // Stereo var srate = 44100, start = 0, end = 3, duration, rate = 1, signal; duration = abs(end - start); signal = BufRd.ar(1, c, // Buffer 0 LinLin.ar( LFSaw.ar(rate/duration, 1), -1, 1, start, end)*srate )*EnvGen.kr(Env.linen(0.01, 0.98, 0.01), timeScale: duration, 141

gate: Impulse.kr(1/duration)); RLPF.ar(signal, LFNoise1.kr([12, 5], mul: 700, add: 1000), rq: 0.05)*0.2; }.play(s) ) ::

•

Granular Synthesis

The first examples above have a loop time between 0.5 and 3 seconds. In the last few examples the size of the loops can be very small; 1/25th of a second. This wasn't possible in the days of tape splicing, though they certainly dreamed about it. Of course now it is possible, but we use the term grains, or granular synthesis. TGrains is similar to PlayBuf, but the envelope (interpolation) is not as steep. The grains can overlap, and there is a pan argument, which we can put to good use. Since there is a pan, the minimum value for numChannels is 2. The centerpos is analogous to the start position in PlayBuf. The trate in 15.5 moves according to the mouse position. Near the top is very slow; 1 in 4 seconds, so you can hear what a grain is, and what is going on at very high rates, such as 200 Hz. As with the previous examples, the duration of each grain can be different from the rate. A duration of 1/trate leaves quite a bit of space between the grains because of the interpolation. A 4/trate is 4 times the length of each grain, but this overlap makes the grains more effective at faster speeds. In this example we load the buffer using read rather than loadDialog. This is faster if you are using the same audio file over and over. Note that it must be a mono audio file to work. Replace "REPLACE_WITH_User" with your user name. Alternately; you can drag the file anywhere in the editor and it will produce the path name in quotes. 15.5:: Granular Synthesis b = Buffer.read(s, "/Users/REPLACE_WITH_User/Desktop/Cage_quote.aif"); // or drag the file to the second argument position below: read(s, HERE); b = Buffer.read(s, ); b.play;

( { var trate, trigger, dur, center; trate = MouseY.kr(1/4,200,1); dur = 4 / trate; trigger = Impulse.kr(trate); center = MouseX.kr(0,BufDur.kr(b)); // numChannels, trigger, bufnum, rate, centerPos, dur, pan, amp, interp TGrains.ar(2, trigger, b, 1, center, dur, 0, 0.5, 4); }.play; ) :: 142

The next example engages the pan position and speed using a TRand as before. Pan can be -1 (left channel) to 1 (right channel). If you are still routing the signal through Logic, adjust the pan on the first two channels to hard left and hard right. The first (commented) alternative is an interesting twist. The center position pops around the file using a random generator. Another center alternative is below the panpos. It links the center location with the pan position, so that the beginning parts of the file will be left, the middle, center, etc. The speed alternative links rate with panpos also; fast grains on the outsides, slow ones in the middle, backward loops on the left, forward on the right. Try that with just a tape deck and splicing blocks. 15.6:: Grains with Pan and Speed ( { var trate, trigger, dur, center, panpos, speed; trate = MouseY.kr(1/4,200,1); trigger = Impulse.kr(trate); dur = 4 / trate; center = MouseX.kr(0,BufDur.kr(b)); // center = TRand.kr(0, BufDur.kr(b), trigger); // center = Dseq([0, 0.1, 0.9, 0.7, 0.4, 0.5, 0.6, 0.4, 0.3, 0] * b.numFrames, inf); panpos = TRand.kr(-1.0, 1.0, trigger); // panpos = Dseq([-1, -0.3, 0, -0.6, 0, 0.1, 1, 0.5, 0.3, 1], inf); speed = TRand.kr(-2.0, 2.0, trigger); // center = (panpos /2 + 0.5) * BufDur.kr(b); // speed = panpos * 2; // speed = Dseq([-1, 1, -0.1, 0.5, -0.5, 3, 1, 0.3, -0.3, -1], inf); TGrains.ar(2, trigger, b, speed, center, dur, panpos, 0.5, 4); }.play; ) ::

Have I mentioned there is absolutely nothing on earth that can generate these kinds of sounds. (I'm speaking of granular synthesis in general, not SC specifically.) But there are precious few synthesis packages that can do granular synthesis. •

Sequenced Granular Synthesis

The last set of commented lines uses a Dseq, which returns an array of values in sequence with each triggered event. (More on this later—after I figure out how it works.) The first argument is the array of values, the second is the number of repeats. In this patch it is set to inf or infinite repeats. The current version has 10 values for each array. This ensures the patterns match. Making them different lengths—8, 9, and 10—will set the patterns into a phase shift. Setting center, panpos, and speed to a Dseq is too repetitive. Setting them to different lengths is too much variation. As with all music, search for a balance between variation and repetition. I prefer leaving at least one with a mouse control.

143

•

Exercises 15:1 Choose your favorite version of the patches above and document each line with comments. You may have to change the values, then listen to the patch as you document. 15:2 In Logic, add auxiliary reverbs, delays, and possibly serial effects to the channels. When you arrive at an interesting variation, commit it to disk by recording it in Logic. Continue in Logic to build a foundation for an electro-acoustic work. 15:3 Use the granular synthesis patch with mouse controls to add a "lead" to your composition. Perform it live and record several takes. Choose from the takes using the comp features in Logic.

144

16.

Wave Generators Analog synthesizers; but now digital.

Visual arts use color, shape, texture, and brightness. Similarly, music can be broken down into frequency, amplitude, and timbre, as described in previous chapters. The following illustrates how to control these three characteristics using SC wave form generators. •

Frequency, Amplitude, Timbre

The arguments for a SinOsc are frequency, phase, mul, and add. Frequency is the number of cycles per second. Amplitude (mul) changes the height of the wave. Phase changes where in the cycle the wave begins. (Changing the offset moves the wave up or down. This makes no sense, and shouldn't be done with a simple audio wave. Offset is used for control sources, covered later.) Changing the frequency and amplitude translate into pitch and volume. We don't hear a change in phase, but when several waves of different phase are combined, we hear the results of the interaction. •

Voltage Control

Early synthesizer modules were labeled VCO, VCA, and VCF—acronyms for voltage controlled oscillator (to control pitch), amplifier (to control amplitude), and filter (to control upper harmonic structure). Nearly all previous patches have used voltage controls (MouseX, TRand, LFNoise, etc.). The term voltage is not technically correct for what happens inside a computer, but the results are the same, so the term lingers. They are also what you will see on analog synthesizers and most Logic instruments. Control simply means changing a parameter. When you turn the volume of your stereo up you are manually controlling the amplitude parameter. When you press your finger down on the string of a guitar you manually control the length of the string and therefore the frequency. Controlling these parameters with the precision, accuracy, speed, and flexibility inherent in electrical circuitry is what inspired early pioneers. The limitations inherent in real instruments and humans playing them (how high we can sing, how fast we can play, how loud a piano can be) could be surpassed with electrically generated and controlled sound. They were right, and today there are virtually no limits to what you can do in computer music. They can easily exceed our ability to perceive. The limits, dear Brutus, is not in the machines, but in ourselves. Example 16.2 shows three SinOsc patches. The first has a static value for frequency, the second replaces that static value with a MouseX control (which is still, technically, manual control), the third replaces it with a TRand. TRand supplies a string of values to the SinOsc, which uses them for the frequency argument, at the rate specified by the Impulse trigger. This is a basic voltage controlled frequency. The frequency is being supplied by the TRand. 145

16.1::

VCO SinOsc

{SinOsc.ar(freq: 440)}.play {SinOsc.ar(freq: MouseX.kr(200, 2000))}.play {SinOsc.ar(freq: TRand.kr(200, 2000, Impulse.kr(5)))}.play ::

The arguments for TRand are lo, hi, and trigger. The trigger is Impulse.kr. It takes only one argument for frequency. There are, therefore, two frequencies in the third patch. This is often confusing. There is the frequency of the SinOsc, which are changing values between 200 and 2000. This is the frequency we hear as pitch. The frequency of the Impulse is 5 Hz, or 5 times per second. We don't hear this frequency as a pitch, but as a tempo; the number of times per second a new pitch is heard. Frequencies below 20 Hz are called LFOs; low frequency oscillators. Example 16.2 shows controls of frequency, amplitude, and timbre using three TRands as VCO, VCF, and VCA in a Blip. The arguments for Blip are frequency, number of harmonics (more give a brighter timbre, fewer a duller sound, so essentially a filter), and mul, or amplitude. Recall the exercise where you sang into an Amadeus oscilloscope and changed frequency but not timbre or amplitude, then amplitude only, then timbre only. This is the same thing, but using an automated, voltage control. 16.2:: VCO, VCF, VCA using Blip // Controlling frequency {Blip.ar(TRand.kr(200, 2000, Impulse.kr(5)))}.play // Controlling filter (number of harmonics) {Blip.ar(300, TRand.kr(0, 20, Impulse.kr(5)))}.play // Controlling amplitude {Blip.ar(300, 4, TRand.kr(0.1, 0.9, Impulse.kr(5)))}.play // All three ( {Blip.ar( TRand.kr(200, 600, Impulse.kr(5)), TRand.kr(0, 20, Impulse.kr(5)), TRand.kr(0.1, 0.9, Impulse.kr(5)))}.play ) ::

•

Scale and Offset

Notice that all three TRands differ only in lo and hi value. In the last example, the three TRands are combined as VCO, VCF, and VCA. If this were a much more complicated patch with hundreds of components, this duplication could eat up CPU. It would be more efficient to use a single TRand for all three, but we cannot do that if the values required are so disparate. 146

The solution is offset and scale; use a single TRand with a range of 0 to 1 and modify the range with simple math. Example 16.3 shows how a TRand with a range of 0 to 1 can be transformed using offsets and scales. It uses poll, which prints the values to the data window, so you can monitor the output. Run the first line. You should see values between 0 and 1. The second example uses a single TRand but the values are multiplied by 100. The last adds 100. Adding or subtracting changes the offset. Multiplying changes the scale. 16.3:: Scale and Offset Examples {(TRand.kr(0, 1.0, Impulse.kr(10))).poll(label: 'rand')}.play // range of 0 to 1 {(TRand.kr(0, 1.0, Impulse.kr(10)) * 100).poll(label: 'rand')}.play // range of 0 to 100 {(TRand.kr(0, 1.0, Impulse.kr(10)) + 100).poll(label: 'rand')}.play // range of 100 to 101 [0, 0.5, 1, 0.5, 0, -0.5, -1, -0.5, 0] .plot("default", minval: -5, maxval: 5);

// triangle

// scale by 3 ([0, 0.5, 1, 0.5, 0, -0.5, -1, -0.5, 0] * 3) .plot("*3", minval: -5, maxval: 5); // scale by 2, offset by 3 ([0, 0.5, 1, 0.5, 0, -0.5, -1, -0.5, 0] * 2 + 3) .plot("*2, + 3", minval: -5, maxval: 5); // scale by 100, offset by 300 ([0, 0.5, 1, 0.5, 0, -0.5, -1, -0.5, 0] * 100 + 300) .plot("*100, +300", minval: -500, maxval: 500); ::

You can think of any wave generator as a stream (or array) of numbers describing the shape of the wave (because, uh, that's what it is). The second half of example 16.3 shows how offset and scale can change a wave. The first array of values describe, albeit at a very low sample rate, a triangle wave. The message plot realizes them in a plot in the sclang program. (The size of the plot is -5 to 5. The wave shows up between -1 and 1.) When scaled by 3, the shape of the wave doesn't change, but the range of values produced is greater: -3 to 3. The next shows a scale of 2, then offset of 3. Note that the triangle retains it's shape, only the range that it covers changes. The last, a scale of 100, offset of 300, resulting in a range of 200 to 400. Figure 16.1 shows the resulting plots. Of course, this triangle wave is no longer useful as an audio output, since 1 and -1 are the maximum values of the audio output. Scaling and offsetting in this way adjusts a single source so that it can be used with arguments that require different ranges, typically, as controls. The -3 to +3 could be used to control playback speed of a concrète patch. The next example, with a range of 1 to 5, might be used to control the number of overtones in Blip. The last example has a range of 200 to 400. These numbers are appropriate as a control source for frequency.

147

Fig. 16-1 Scale and Offset

•

Precedence

 … in SC is different from what you might expect. In a math class 10 + 5 * 6 and 6 * 5 + 10 would be the same (40) because the rules of precedence are; multiplication then the addition. In SC the rules of precedence are simple, but very different; operations are performed from left to right. So 10 + 5 * 6 = 90, but 6 * 5 + 10 = 40, and 10 + 6 * 5 = 80. If you need to force precedence, use parentheses (10 + (5*6) = 10 + (6*5) = (6*5) + 10 = 40). In the patch below a single TRand and single Impulse are stored in the variable control. That single control, producing values between 0 and 1, is used as a VCO, VCF, and VCA. The VCA (mul) doesn't need to be scaled or offset, since it's values are already 0 to 1. The VCF (number of harmonics) is scaled by 20. The VCO (frequency) is first scaled by 500, then offset by 500. Each control is enclosed in parentheses and passed to the poll message, which prints the value ten times a second. Once you've confirmed that the values are correct for frequency, harmonics, and amplitude, try faster rates—4, 10, 20—for the Impulse. It is worth patching this example through SoundFlower, then monitoring it using Amadeus' oscilloscope. This allows you to see the controls in the wave. Frequency will change the number of peaks in the scope, amplitude the height of the peaks, timbre the shape (how sharp the peak is). 16.4:: Scale and Offset Patch ( { var control; control = TRand.kr(0, 1.0, Impulse.kr(1)); Blip.ar( (control * 500 + 500).poll(10, "freq"), (control * 10).poll(10, "timbre"), control.poll(10, "amp") ) }.play ) 148

::

•

Keyword Assignment

Arguments have to be in the correct order. The fourth argument for a SinOsc is mul. If you want to give it a value, you need to supply values for freq, and phase even if the defaults are sufficient. They are needed as place markers so that the mul is in the correct position. This becomes a problem if there are, say, six arguments, and you want to change the sixth. You would have to enter arguments 1-5 just so the sixth argument is in position six. This is not only cumbersome, but invites error.  The solution is keyword assignment. You can precede an argument with a keyword and colon. Keyword names can be found in the documentation files and are seen in the help bubbles while typing. The documentation for SinOsc.ar shows: SinOsc.ar(freq, phase, mul, add). The keywords then are freq, phase, mul, and add. Using keywords not only allows you to enter a single argument out of order, but to mix the order of the arguments. Here are several versions of the SinOsc example written using keywords. All of them have precisely the same meaning and will sound the same. 16.5:: Keywords {SinOsc.ar(freq: 440, phase: 0, mul: 0.4, add: 0)}.play; {SinOsc.ar(phase: 0, freq: 440, add: 0, mul: 0.4)}.play; {SinOsc.ar(freq: 440, mul: 0.4)}.play; ::

 Another good reason for using keywords is clarity. The keyword provides an explanation of what each value means, like a short comment. The last reason for using keywords is portability. As mentioned earlier, argument lists are not interchangeable. Consider these two lines: Saw.ar(100, 500, 600) and SinOsc.ar(100, 500, 600). The arguments for Saw are freq, mul, and add. But for SinOsc they are freq, phase, mul, and add. So the two lists are not interchangeable. I can certainly use 100, 500, and 600 as arguments for the SinOsc, but they don't mean the same thing as they did with the Saw. A phase of 500 makes no sense, and a mul of 600 without an add will produce negative values, probably not what I want. Keywords (provided the two UGens have the same keywords) make the objects interchangeable: SinOsc.ar(freq: 100, mul: 500, add: 1000) and Saw.ar(freq: 100, mul: 500, add: 1000) will have a similar effect in a patch. Warning: this can be either dangerous or a source of serendipity. You should always double check the keywords and understand what they mean in a patch. Here is the first patch using keywords followed by dangerous swapping. 16.6::

First Patch using Keywords

{SinOsc.ar(freq: LFNoise0.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play 149

{Saw.ar(freq: LFNoise0.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play {SinOsc.ar(freq: LFNoise1.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play {Pulse.ar(freq: LFNoise1.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play {Pulse.ar(freq: LFSaw.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play {LFTri.ar(freq: LFPulse.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play {LFTri.ar(freq: LFTri.ar(freq: [10, 15], mul: 400, add: 800), mul: 0.3)}.play ::

•

Vibrato

… is a simple voltage control common on most synthesizer patches. On a live instrument, vibrato is a slight oscillation (convenient, since we have oscillators) in either the amplitude (in the case of voice) or pitch (in the case of stringed instruments). String musicians generate vibrato by rolling the finger forward and backward on the string. Given an A 440 the actual pitch with vibrato would move between, say 435 and 445, essentially moving in and out of tune about 5 times per second. In some previous SinOsc patches the frequency argument was a static value (such as 440). We need to replace the static value with some function or UGen that will change smoothly over time between 435 and 445. A sine wave moves back and forth between two values at a periodic and predictable rate, but not 435 and 445. We can use a slow moving SinOsc, scaled and offset correctly, as the input or freq argument for another SinOsc. The example below shows SinOsc UGens attached to a poll to print the values it produces. The frequency is 1 cycle in 4 seconds. Its default output is similar to the triangle wave earlier. It starts at 0 then moves to 1 then back to 0, then negative 1, then 0. More precisely, it's range is 0 +/-1 (zero plus and minus one). The difference between frequency and output is very confusing to beginners. The frequency is the number of times the wave moves up and down in one second. The output is the string of numbers that are actually generated by the up and down motion. In each of these examples the frequency is the same: 1/4 (once every four seconds). The output, however, is modified in each one. While the oscillation is still 1 in 4 seconds, the output of the first example is 0 +/-1. In the second example the frequency is still 1 in 4, but the output is now 0 +/-5. In the third example the output is 440, +/-1. The last example has both: scale by 5, offset by 440. Finally, the plot. 16.7:: SinOsc poll {SinOsc.ar(1/4).round(0.01).poll; 0}.play // -1 to 1 {(SinOsc.ar(1/4) * 5).round(0.01).poll; 0}.play // -5 to 5 {(SinOsc.ar(1/4) + 440).round(0.01).poll; 0}.play // 339 to 441 {(SinOsc.ar(1/4) * 5 + 440).round(0.01).poll; 0}.play // 395 to 405 150

{(SinOsc.ar * 5 + 440)}.plot ::

Example 16.8 illustrates how this offset and scaled LFO sine oscillator can be linked to a second SinOsc to create a subtle vibrato. Use a MouseX to determine which parameter is responsible for the speed of the vibrato, which is the depth, and which is the pitch we hear. Try higher depths. Try faster speeds until you find a natural sounding vibrato. Try a lower speed to clearly see the poll results. (Do not try speeds or depths higher than 20 Hz! In fact, it is irresponsible of me to bring it up. Forget I said anything.) 16.8::

VCO Vibrato

( { var control; control = SinOsc.kr(4) * 5 + 600; SinOsc.ar(control.poll); }.play ) ::

When I first tried a VCO control on a vintage machine the offset and scale were absolutely clear because I had a knob at my finger tips that changed them both. Sometimes that tactile feedback helps. Example 16.9 creates a GUI (covered later, and I'm taking a lot of shortcuts here) with sliders to control the three key components of a low frequency control; the rate or frequency of the controlling wave, the scale of the wave, and the offset. Note that none of these sliders control the parameters of the sine that we hear. Not directly, at least. They are only linked to the controlling wave. The audio wave (the one we hear as a pitch) gets its frequency directly from the controlling oscillator. Again, it would be useful to watch the change using Amadeus' real time sonogram. To stop this, close the window, switch back to SC and press ⌘period. Sorry for the kluge. 16.9:: Rate, Scale, and Offset Sliders ( var inst, gbutton; var gkfreq, gscale, goffset, gWin; inst = {arg rate = 4, scale = 30, offset = 800; SinOsc.ar(SinOsc.kr(rate, mul: scale, add: offset), mul: 0.3)}.play; gWin=Window.new.front; gWin.view.decorator = FlowLayout(gWin.view.bounds); gkfreq = EZSlider(gWin, 400@20, "Rate", [1, 20, \lin, 0.1, 4].asSpec) .action_({|sl| inst.set(\rate, sl.value)}); gscale = EZSlider(gWin, 400@20, "Scale", [10, 1000, \lin, 10, 30].asSpec) .action_({|sl| inst.set(\scale, sl.value)}); goffset = EZSlider(gWin, 400@20, "Offset", [100, 3000, \exp, 10, 800].asSpec) .action_({|sl| inst.set(\offset, sl.value)}); ) 151

::

Offset and scale are so integral to synthesis that most UGens have a mul and add argument built in. In fact, the first line should have been written SinOsc.kr(4, mul: 5, add: 440). (Note the keywords to avoid using the phase argument.) The results are the same. This is a fairly believable vibrato, if a bit crude, manufactured, or flat; like a lot of early electronic music. (And to be fair, we all knew it. Many early works seem comical; like the first attempts at moving pictures. This explains, in part, the appeal of Concrète. At the time tape work was much more sophisticated, just because the sounds were so complex. But we also saw the potential of analog synthesis. Logic Pro illustrates how it has matured, and how we benefit from the vision of analog pioneers.) Real musicians don't dive right into a vibrato. They usually start with a pure tone then gradually increase the speed and width of the vibrato. Which values would you change to make the vibrato more natural? How would you change them? The envelope Line.kr can make a gradual change over time. The example below adds a second layer of controls. Supply the missing arguments. 16.10:: Vibrato Envelope ( { var vibrato, speed, depth; speed = Line.kr(start, end, duration); depth = Line.kr(start, end, duration); vibrato = SinOsc.kr(speed, mul: depth, add: 440); SinOsc.ar(vibrato); }.play ) ::

Lest you get too comfortable in your grasp of offset and scale, I'm going to add the following complication. Values for offset and scale differ depending on the type of UGen … •

Unipolar/Bipolar

 The first KRand used to illustrate offset and scale was unipolar; it generated only positive values between 0 and 1. The SinOsc is bipolar, and returns positive and negative values between -1 and 1. Most UGens are bipolar, but many are unipolar. The difference is important when calculating offset and scale. Below are two arrays of values. The first is unipolar, the second bipolar. They are both scaled by 100 and offset by 100. But notice how different the results are. [0, 0.1, 0.5, 0.4, 0.9] * 100 = [0, 10, 50, 40, 90] + 100 = [100, 110, 150, 140, 190] [0, -0.1, -0.5, 0.4, 0.9] * 100 = [0, -10, -50, 40, 90] + 100 = [100, 90, 50, 140, 190] An easy way to remember is that in both cases the offset replaces the 0 and the scale replaces the 1s. A unipolar UGen's output is 0 to 1. When offset by 100 then scaled by 100, 0 becomes 100 and 1 becomes 200. A bipolar UGen's output is 0 +/-1. When offset by 100 then scaled by 100, 0 becomes 100, -1 becomes 0, and 1 becomes 200. Said a different, possibly (but I won't guarantee) less confusing way; offset is what 0 used to be, scale is the deviation from that. The EnvGen is a unipolar control. It creates, by default, values between 0 and 1. 152

Fig. 16-2 Unipolar Output

When scaled, zero also stays the same, but 1 is replaced by the value of scale. Notice that this does not produce negative values (when offset is 0) nor values below the offset (when offset is different from 0). Fig. 16-3 Unipolar Scaled by 5

Fig. 16-4 Unipolar Offset by 100, Scaled by 5

Offset and scale are not temporary asides along our path to the real topics of synthesis. It took me several days. You have to have the same eureka moment. They are the core of every patch we will build. Skip ahead and complete exercise 16.1 before moving onto the Theremin. •

Audio UGen, Control UGen (Carrier, Modulator)

The vibrato example uses two SinOsc UGens but in very different ways. One is the sound you hear (carrier) and the other changes the sound you hear (modulator). Each has a frequency that determines the aggregate sound, but in different ways. The freq of the modulator determines the speed of the vibrato. The freq argument of the carrier is the pitch we hear. The scale of the carrier (5) is the width of the vibrato. The offset (600) is the pitch that we hear, + and - 5 Hz. Change each of these values to confirm how they are related to the vibrato.  The difference is the rate. The carrier SinOsc uses the ar message while the modulator SinOsc uses kr. The ar stands for audio rate while kr is control rate. The carrier SinOsc is generating actual audio so it needs to be audio rate (44.1k). But controls that are not heard as audio, but rather transfer their shape as a control, or that trigger other UGens, do not need such high resolution and are more efficient if generated at a lower rate. A SinOsc.kr will generate a 153

sine wave at 1/64th the rate of a SinOsc.ar. In short, kr is more efficient, and should be used for control UGens. •

Theremin

The Theremin was one of the first universally popular electronic instruments. It's the only instrument I know that is played without touching. One of the first film scores to included the Theremin won an academy award; Spellbound. Hitchcock didn't like it, but became so pervasive and closely linked to the alien worlds of early sci-fi that it surfaces in stand up comedy (Robert Klein) and MST3K riffs ("I know that Theremin is around here somewhere"). Most people think they recognize it in the Beach Boys' Good Vibrations. In fact, they were unable to find a real Theremin, so they used an analog synthesizer to imitate the iconic sound. I include it here as an application of vibrato, because the two characteristics we most identify with this instrument are the glissando between pitches (a continuum rather than discrete pitches) and a solid wide vibrato. The vibrato was not the result of a voltage control (as in these examples), but generated by the performer's hand. Perhaps they felt the need for some added dimension to compensate for the cold character of a pure sine wave and inevitable difficulties with intonation. Below is a virtual Theremin with controls connected to mouse position: up and down for amplitude, left and right for pitch. In the first vibrato, the control wave generates frequencies. The problem with this method is that the width of a vibrato should be wider with higher pitches. A range of +/-5 worked for 440, but as we've noted previously, we perceive pitch logarithmically, not linearly. Given a pitch of 1000 a width of +/-100 will sound more correct. For that reason, the control oscillator has an offset (add) of 1 and scale (mul) of 0.02. This means it moves 0.02 above and below 1. The frequency is multiplied by that value, so if you hold the mouse at 800 Hz, the vibrato will move above and below 800 by 2% (16 Hz), but if you are at 3000, it will deviate by 60 Hz. 16.11:: Theremin ( { var vibrato; vibrato = SinOsc.kr(6, mul: 0.02, add: 1); SinOsc.ar( freq: MouseX.kr(200, 3200, 1, 0.5) * vibrato, //Vibrato mul: MouseY.kr(1, 0.02) //Amplitude; top is louder ) }.play ) ::

154

•

Exercises 16:1 Fill in the following charts. The first presents an offset and scale. You provide the bipolar and unipolar results. The second offers a common control and a range that would work for that control, you supply the mul (scale) and add (offset). Offset 500 800 1 0 1000 10

Scale 100 1000 0.5 0.01 1000 20

Bipolar range 400 to 600

Unipolar range 500 to 600

Control type MIDI Amplitude Frequency Harmonics Phase Duration

Typical range 48 to 72 0.1 to 0.9 100 to 900 2 to 12 0 to 6.28 1 to 6

Bipolar offset/scale mul: 12, add: 60

Unipolar offset/scale mul: 24, add: 48

16:2 Do not, repeat, do not even think of changing the Theremin patch vibrato to speeds higher than 20 and widths between 0.1 and 0.9. Are you mad? 16:3 In the Theremin patch, add keywords to all arguments. Then swap out the control SinOsc with LFTri, LFSaw, LFPulse, LFNoise0, and LFNoise1. 16:4 Modify the Blip patch by replacing the TRand UGens with properly scaled and offset SinOsc, LFTri, LFSaw, LFPulse, LFNoise0, or LFNoise1.

155

17.

Sources of Control Impart their shape to the sound; smooth or chaotic.

•

Periodic Waves

… repeat. Aperiodic waves do not (this is technically untrue). Both types of waves are useful as controls. Figure 17.1 shows graphs of sine, saw, triangle, pulse, interpolated random and step random waves. The first block of code in example 17.1 produces this plot. The second block of lines applies each wave shape as a frequency control in a SinOsc. It uses a GUI (covered in chapter 18). Press start to begin the patch, then select (with the slider or the number box) the control. Sine wave is 1, saw is 2, triangle 3, pulse 4, interpolated noise 5, and step noise 6.

Fig. 17-1 Wave shapes

17.1::

Wave Shapes

// Periodic waves {[SinOsc.ar, LFSaw.ar, LFTri.ar, LFPulse.ar, LFNoise1.ar(3000), LFNoise0.ar(3000)]}.plot; // Each shape above used to control frequency. MouseY chooses different controls ( SynthDef("Waves", { |con=0| Out.ar(16 - con, [SinOsc.ar(2), LFSaw.ar(2), LFTri.ar(2), LFPulse.ar(2), LFNoise1.ar(8), LFNoise0.ar(8)] * 100 + 500); Out.ar([0, 1], SinOsc.ar(In.ar(16, 1))*0.8); } ).add; ControlSpec.specs[\con] = ControlSpec(0, 5, \lin, 1, 0, "Control"); SynthDescLib.global[\Waves].makeWindow 156

) ::

•

VCA

UGens intended for amplitude control have a range of 0 to 1, therefore do not need to be scaled or offset. Bipolar UGens must be scaled by 0.5 then offset by 0.5. Alternately, a bipolar control can be run through a max function with 0 as the first argument. This sets the floor at 0. Anytime the control moves to negative values they are replaced by the maximum 0. This creates longer periods of 0 volume. This control source can then be applied to the mul of a SinOsc. 17.2:: VCA {LFNoise1.kr}.plot(0.1); {max(0, LFNoise1.kr)}.plot(0.1, minval: -1, maxval: 1); {SinOsc.ar(1000, mul: max(0, LFNoise1.kr(5)))}.play; ::

As before, we now layer another control inside the LFNoise to modify it's amplitude also; a Line.kr. The first line in 17.3 plots the control, the next plays with a duration of 10 seconds. This single sound, on it's own, is not very appealing. But adding a bunch of them together in example 17.4 using a Mix.ar with random frequencies is, and illustrates how a simple idea can become complex very fast. 17.3::

VCA With LFNoise and Line

{max(0, LFNoise1.kr(mul: Line.kr(0.9, 0, 0.1)))}.plot(0.1, minval: -1, maxval: 1) {SinOsc.ar(1000, mul: max(0, LFNoise1.kr(5, Line.kr(0.9, 0, 10))))}.play ::

The next example uses the single LFNoise1, offset and scaled properly to control both the amplitude and the frequency of the SinOsc. It also places the function inside a Mix.fill. The first argument is the number of items to mix, the second is the function that generates each instance to be mixed. Since it is a function, the random elements will be executed for each instance, therefore different for each instance. (Place a .postln after the rrand(100, 800) to confirm this.) This means the control will have a range of 100 with the value supplied by rrand as the center, a value between 100 and 800. Also, each instance will have an independent LFNoise1. Try changing the speed of the LFNoise1 (1/3), the duration of line (30), and swap LFNoise0. 17.4:: VCO, VCA With LFNoise And Line ( { Mix.fill(30, { var control; control = LFNoise1.kr(freq: 1/3, mul: Line.kr(1.0, 0, 30)); 157

Pan2.ar(SinOsc.ar(control * 100 + rrand(100, 800), 0, max(0, control/30)), 1.0.rand2)}) }.play ) ::

•

Aperiodic Triggers, Complex Envelopes

Periodic and aperiodic sources are constantly in motion. Triggered controls such as KRand and envelopes supply values only when triggered. The trigger can be periodic or aperiodic. The first few patches used the periodic trigger Impulse.kr. Dust.kr is an aperiodic trigger. Though not their intended application, you can play both of these triggers. You will hear a tick at each trigger. Note the stereo expansion. 17.5:: Aperiodic Triggers {Impulse.ar([3, 2])}.play {Dust.ar([5, 6])}.play ::

The Line UGen in the previous patch is a type of envelope. Envelopes are a single event (as opposed to periodically repeating events) that change over time. They are periodic only if triggered periodically.  Common envelopes have attack, decay, sustain, and release sections (ADSR). Simpler envelopes may include only AR, or attack and release. Below is a graph showing an envelope with an attack and decay. Below are graphs of common envelopes. You can also specify the type of curve between each breakpoint; linear, exponential, welch, sine, etc. In SC, envelopes provide an additional argument: done action, or what to do when the envelope closes. The default action is to stop playing that UGen, which frees up processing. Without this feature the events in a composition would add up and eventually choke the CPU.

Fig. 17-2 AR, ASR, ADSR, Aussie Opera

 Fixed duration envelopes are always the same duration, typically the attack and release combined. The duration of a sustain envelope will change depending on the length of a gate, e.g. how long a key is held down. Percussion instruments such as cymbals, marimba, chimes, etc., have fixed duration envelopes. Once they are set in motion you can't control the decay. Organs, voice, violins, brass, etc., have sustaining envelopes. The sustain time is provided by some type of gate, for example holding down a key. Each note on an organ will sound until the key is released. At that point the pitch dies away. Complex envelopes look like Aussie opera houses. 158

While amplitude is an obvious application of envelopes, most instruments also change timbre over the duration of a note. The patch below uses a PMOsc (phase modulation oscillator, covered later) as a VCO, VCA, and VCF. The first argument is pitch (the carrier), the second is the modulator, which effects the character or shape of the wave, the third argument is the pmindex which effects the number of harmonics (or sidebands), the mul argument is amplitude. A trigger is defined, which is used in the Linen envelope. The trigger is also used to generate new pitches (MIDI values) which are converted to cycles per second, and also for a random choice of mod. The key is how the envelope is used. It is placed, unmodified, in the mul, argument, which accounts for amplitude. But it is also used as the third argument—index—which determines the number of harmonics. A higher number is brighter, lower dull. It requires values well above 1, which is the default output of the unipolar Linen. To achieve correct values it is scaled by 6. An array is used for the first argument for TIRand, which makes the patch stereo. 17.6:: VCA and VCF Envelope ( {

var envelope, rate = 1, pitch, trigger, mod; trigger = Impulse.kr(rate); envelope = Linen.kr(trigger, 0, 1, 1/rate); pitch = TIRand.kr([48, 48], 72, trigger).midicps; mod = TChoose.kr(trigger, [3/2, 4/3, 5/4, 6/5, 7/6]); PMOsc.ar(pitch, pitch*mod, envelope * 6, mul: envelope) * 0.5 }.play ) ::

The rate is 1 so that you can focus on the decay in both the volume and the timbre. Once you've listened closely to the character, feel free to change it to higher rates; 10 to 15. Note also that rate is given a value in the same line where it is declared. Below is a sonogram of the results. Color is amplitude, showing how the sound dies away with the envelope, but also since the envelope is applied to both amplitude and timbre (pmindex) notice that the harmonics are tapered smoothly as the amplitude dies away.

159

•

Dseq

… illustrates another common, classic, control source; the sequencer, which runs through an array of values when called by Demand.kr. Dseq takes two arguments; the sequence, in the form of an array, and the number of repetitions before it stops. 17.7:: Dseq ( { var rate = 7, trigger, duration; trigger = Impulse.kr(rate); duration = 1/rate; SinOsc.ar( Demand.kr(Impulse.kr(rate), 0, Dseq([60, 62, 64, 65, 67, 69, 71].midicps, 5)))* Linen.kr(Impulse.kr(rate), 0, duration * 0.3, duration * 0.5) }.play ) // Used to control timbre ( { var rate = 7, trigger, duration; trigger = Impulse.kr(rate); duration = 1/rate; Blip.ar(600, Demand.kr(Impulse.kr(rate), 0, Dseq([1, 4, 2, 7, 9, 20, 1, 4, 2, 6, 1].midicps, 5)))* Linen.kr(Impulse.kr(rate), 0, duration * 0.3, duration * 0.5) }.play 160

) ::

These control sources can be nested—a sequence of sequences—for greater variety: Demand.kr(trigger, [Dseq(etc.), Dseq(etc), Dseq(etc], inf). As a generative composer, I'm rarely invested in a specific series. I prefer to be surprised by the outcome. The generative process embraces all the possibilities within a range of instructions as equally valid. This does not mean completely random choices, but rather choices within a range of appropriate values for the desired effect. As an example, random frequencies would not be heard as pitch or any scale. For that reason the choice for pitch might be limited to MIDI values, or even a major scale using MIDI values. Sequences require arrays. A quick and easy way to fill an array with values is … •

Array.fill The two arguments are size of the array and a function for generating values.

Functions are enclosed in braces. The function runs its enclosing code from scratch each time it is called. Note the difference between the first two lines in example 17.8. The dup message duplicates its object. The first chooses a random number and that object is duplicated. The second duplicates the function of choosing a random number, and each function returns a different random value. The second becomes [10.rand, 10.rand, 10.rand, 10.rand, 10.rand, etc.]. Array.series is also useful. The arguments are size, start, and step. This is useful for creating a harmonic series. The final example fills an array with a pentatonic scale, the last scrambles a pentatonic scale (to avoid repeated notes, if that's the desired effect). In the actual code, an octave would be chosen then added to the MIDI note: 60 + 4, 48 + 7, etc. 17.8:: Filling Arrays 10.rand.dup(10) // No function, duplicates a single random choice. {10.rand}.dup(10); // Function repeats, 10 random numbers, not ten of a random number Array.fill(10, 10.rand) // no function, a single random number repeated Array.fill(10, {10.rand}) // the function repeats, so random numbers Array.series(10, 1, 1) Array.series(10, 1, 1) * 440 // returns harmonic series Array.fill(30, {[0, 2, 4, 7, 9].choose}) // 30 items from a pentatonic scale (any octave) [0, 2, 4, 7, 9].scramble; // scramble a series ::

This is why all of our patches were enclosed in a function. A SinOsc on its own would just generate a single value; the first value of a sine wave. But placing it in a function allows it to repeat the process of generating values in the shape of a sine wave. 161

The {function}.dup(n) seems more efficient, but the Array.fill includes a useful feature. It passes a value to the function as it runs. These … •

Function Arguments

… can be used inside the function as it repeats. A real life example might be instructions for someone to repeat an action, say, cut up ten apples. If you wanted them all cut the same way, you could use a single sentence: "cut all apples in half." But what if you wanted each one cut differently; leave the first one whole, cut the second one in half, the third in threes, the fourth in fourths, and so on. Rather than list explicit instructions for each apple, you could just include in the instructions to count the number of repetitions and cut each apple into that many pieces. In order to do this, the system has to keep count so as to know which iteration it is on. Array.fill has a built in counter, passed to the function as an argument. It can have any name, but must be declared as an argument. The first example simply fills the array with the count of each repetition; not very useful. But the counter can be used, like Array.series, to generate harmonics. Example 17.9 illustrates filling an array using the count argument. 17.9:: Array.fill counter Array.fill(20, {arg count; count}) // fill the array with the count argument Array.fill(20, {arg count; count+1*440}) // Use count to generate a harmonic series Array.fill(20, {12.rand}) // 20 random MIDI pitches, 0 through 12 Array.fill(20, {48.rand + 36}) // 20 random pitches spread out over 4 octaves // This line first picks a value from the first array, then adds it to the second // The first is a pentatonic scale, the second octave Cs. This preserves the // pentatonic scale. Array.fill(20, {[0, 2, 4, 7, 9].choose + [36, 48, 60, 72].choose}) // An array of Dseq's each filled using Array.fill. Array.fill(6, {Dseq(Array.fill(10, {[60, 62, 64, 67, 69].choose}), 5)}) ::

The last line fills an array with Dseq UGens. Each of the Dseq's is in turn filled with an array of 10 values, chosen from a pentatonic scale, with a repeat value of 5. You could change this single 5 to another random choice, say rrand(2, 6). This can now be placed in the Demand UGen, which is placed in the previous patch. It is now generative. As I said previously, I'm satisfied with any of the resulting variations. Why impose one particular sequence over another? If you love your music, set it free.

162

17.10::

Complex Demand Sequence

( {

var envelope, rate = 9, pitch, trigger, mod, seq, oct; trigger = Impulse.kr(rate); envelope = Linen.kr(trigger, 0, 1, 1/rate); seq = Array.fill(10, {Dseq(Array.fill(12, {[0, 2, 4, 7, 9].choose}), rrand(2, 6))}); oct = Array.fill(10, {Dseq(Array.fill(12, {[60, 72].choose}), rrand(2, 6))}); pitch = Demand.kr(trigger, 0, Dseq(seq, inf) + Dseq(oct, inf)).midicps; mod = TChoose.kr(trigger, [3/2, 4/3, 5/4, 6/5, 7/6]); PMOsc.ar([pitch, pitch*2], pitch*mod, envelope * 6, mul: envelope) * 0.3 }.play ) ::

•

VCO, VCF, VCA in Logic Instruments

 Most Logic instruments are laid out in a VCO, VCF, VCA model. The skin is even designed to look as though the oscillator section on the left flows into the filter in the middle, then out again into the amplifier to the right. Figure 17.3 shows the ES2 broken down into VCO, VCF, then VCA. Following that are images of the ES1, ES2, EVOC, and EXS24, which all have a similar layout. Launch Logic, select the ES1 or ES2 as an instrument and scroll through the presets using the caps lock keyboard to play back (press the [caps lock] key) and the Next Plugin Setting shortcut (either [ ] ] or ⌘-down arrow) to choose a new instrument. When you find something interesting, explore each component. Use the left section (VCO) to choose a different wave form to confirm how the shape is linked to timbre. Adjust the resonance and cutoff to explore the filter in the middle section (VCF). The right controls effect volume. Fig. 17-3 Logic Voltage Control Instruments

163

•

Control Sources in logic

 The control sources are at the bottom of each of these instruments. The ES1 has an ADSR, Mod Envelope, and an LFO (low frequency oscillator). The ADSR is patched to the VCA. Adjust it to shape the attack, decay, sustain level, and release of the amplitude. The Mod envelope is patchable, and changes whatever is selected in the router just to the left. The amount of control is set with the split slider. The bottom sets the control at low keyboard velocity, the top at high velocity. (The effect is passed through the velocity control.) For our experiments keep them the same. Try controlling pitch first, since it is easy to hear. The image to the right shows the LFO set with a 2-second attack for pitch. When using it to control cutoff or resonance, be sure to adjust the cutoff and the resonance in the filter itself so that the envelope has room to work. If the cutoff is set too high the envelope will have little effect.  The Low Frequency Oscillator has a similar layout. Choose the shape of the control with the wave selector, the amount of control with the split slider, and the destination in the router. Again, try pitch first. This image shows an aperiodic step wave controlling pitch at a rate of 8.01 Hz. The SC equivalent would be ({SinOsc.ar(LFNoise0.kr(8.01) * 600 + 1000)}.play). The left side of the rate dial syncs the rate to the current tempo; a very useful feature. Try controlling resonance and cutoff.  The EXS24 and ES2 have a more sophisticated router. Each has a source, via, and destination. Click on the Src to choose a source, via to set an intermediate control, and Dest to choose a destination. Use the split slider to set the amount of change. The center of this slider is 0 change. The top part of the slider is positive values, the bottom are negative (which typically inverts the wave). This image shows LFO1 controlling pitch, by way of Ctrl #1. Ctrl #1 is a controller on a synthesizer. It will change with each studio setup. If the split sliders are at different points, Ctrl #1 will change the amplitude of the LFO1 that is applied to the pitch. If they are together, the Ctrl will have no effect, and the full value of the LFO will be sent to pitch. Once a control is assigned, use the LFO1 and LFO2 (bottom image) to choose a wave and rate. 164

The EXS24 control sources can be used to modify the Concrète samples loaded into an instrument (see previous chapter). •

Synchronized Controls

Events have, as discussed earlier, both frequency and duration. (I've covered this several times, but we are about to apply it, so a quick review is in order.) They are often linked, so it is useful to know how to express both as a ratio. In the case of frequency the number of events is the numerator and the time (usually in seconds) is the denominator. For example, 40 Hz is 40 events in 1 second, or 40/1. A frequency of 1 Hz is 1/1. Frequencies lower than 1 Hz can be expressed as a ratio or decimal. 1/10 and 0.1 are both a frequency of 1 event in 10 seconds. The ratio seems clearer to me. In short: events/seconds. Duration can also be expressed as a ratio. An event that lasts 5 seconds can be expressed as 5/1 (five seconds for one event). Events less than a second long can also be fractional or decimal: 5/1 is 5 seconds for one event, 1/2 is one second for two events, each event lasting ½ second (0.5). Since SC understands both decimal and fractional notation it may seem easier to always use fractional notation and let SC do the math. Using fractions facilitates synchronization of LFO controls. In example 17.11 the LFO SinOsc UGens have frequencies of 1 in 5, 2 in 5, 3 in 5, so that they will all come into sync every 5 seconds. Uncomment the first three lines to add more sines. The second example is similar, with triggers that are synched every 5 seconds. This is a nice illustration of harmonic structure. When the individual bells sound at once, we hear them blend into a single bright tone. This example also illustrates how a composer can exploit the unique capabilities of a computer. This kind of precision is extremely difficult in analog systems. Both of these patches repeat the SinOsc line for clarity. It would be more efficient to use an Array.fill. See if you can make the change on your own. 17.11::

Synched LFO

( { var scale = 600, offset = 1000; Mix.ar([ Pan2.ar(SinOsc.ar(SinOsc.ar(6/5, mul: scale, add: offset), mul: 0.1), -1), Pan2.ar(SinOsc.ar(SinOsc.ar(5/5, mul: scale, add: offset), mul: 0.1), -0.5), Pan2.ar(SinOsc.ar(SinOsc.ar(4/5, mul: scale, add: offset), mul: 0.1), -0.2), Pan2.ar(SinOsc.ar(SinOsc.ar(3/5, mul: scale, add: offset), mul: 0.1), 0.2), Pan2.ar(SinOsc.ar(SinOsc.ar(2/5, mul: scale, add: offset), mul: 0.1), 0.5), Pan2.ar(SinOsc.ar(SinOsc.ar(1/5, mul: scale, add: offset), mul: 0.1), 1), ]) }.play ) ( // synchronized triggers { Mix.ar([ Pan2.ar(SinOsc.ar(100, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(1/5))), -1), Pan2.ar(SinOsc.ar(300, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(2/5))), -0.66), 165

Pan2.ar(SinOsc.ar(500, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(3/5))), -0.33), Pan2.ar(SinOsc.ar(700, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(4/5))), 0), Pan2.ar(SinOsc.ar(900, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(5/5))), 0.33), Pan2.ar(SinOsc.ar(1100, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(6/5))), 0.66), Pan2.ar(SinOsc.ar(1300, mul: EnvGen.kr(Env.perc(0, 1), Impulse.kr(7/5))), 1)]) * 0.1 }.play ) ::

Figure 17.3 shows graphs of the synchronized elements of each patch.

Fig. 17-4 Synchronous LFOs

Mix.fill combines Mix and Array.fill. Examples below. 17.12:: Mix.fill

( //frequency linked to envelope length //high notes short, low long // uncomment lines for alternatives { var frequency, trigger; Mix.fill(40, //number of items {arg count; frequency = exprand(40, 5000); // expon. frequency choice for each 166

// frequency = count + 1 * 30; // harmonic series // frequency = exprand(count+1*40, count+2*40); // enharmonic, but evenly dispersed trigger = Dust.kr(0.05); // trigger = Impulse.kr(count + 1 / 20); // trigger = Impulse.kr(count + 1 / 10); // trigger = Impulse.kr(rrand(1, 40)/20); Pan2.ar( // pan position SinOsc.ar( // sine osc frequency, // freq for this sine osc mul: EnvGen.kr( // envelope (VCA) // the frequency is used to determine duration of envelope Env.perc(0.001, 500/frequency), trigger, // aperiodic trigger 0.6 // overall volume / (count + 1) // amp of this harmonic ) ), rrand(-1.0, 1.0) // pan position ) } ) }.play ) // this one spawns involuntary maniacal laughter

( // blends from random to ordered { Mix.fill(40, //number of items {arg count; Pan2.ar( // pan position SinOsc.ar( // sine osc exprand(40, 1000), // freq for this sine osc mul: EnvGen.kr( // envelope (VCA) // the frequency is used to determine duration of envelope Env.perc(0.001, rrand(0.2, 3)), Dust.kr(0.5) * Line.kr(1, -1, rrand(20, 40));, 0.6 // overall volume / (count + 1) // amp of this harmonic ) ), rrand(-1.0, 1.0) // pan position )+ Pan2.ar( // pan position SinOsc.ar( // sine osc count + 1 * 100, // freq for this sine osc mul: EnvGen.kr( // envelope (VCA) // the frequency is used to determine duration of envelope Env.perc(0.001, rrand(0.2, 3)), Dust.kr(0.5) * Line.kr(-1, 1, rrand(10, 40)), 0.6 // overall volume / (count + 1) // amp of this harmonic 167

) ), rrand(-1.0, 1.0) // pan position ) } ) }.play )

(//reversed envelope {var frequency; Mix.ar({ frequency = rrand(100, 3000); Pan2.ar(SinOsc.ar(frequency, mul: EnvGen.kr(Env.perc(200/frequency, 0.0001), Dust.kr(0.05), 0.2)), rrand(-1.0, 1.0)) }.dup(100))}.play ) (// {var frequency; Mix.ar({ frequency = rrand(100, 1000); Pan2.ar(SinOsc.ar(frequency, mul: EnvGen.kr(Env.perc(frequency/500, 0001), Dust.kr(0.05), 0.05)), rrand(-1.0, 1.0)) }.dup(100))}.play ) ::

168

•

Exercises 17:1 Browse and find an interesting patch in the ES1, ES2, or EXS24. Adjust the waveforms or filters. Once you discover an interesting variation, use the patch bay to attach a voltage control to the parameter. 17:2 Import Concrète samples into the EXS24. Attach controls for pitch. 17:3 Begin with your favorite SC patch from this section. Document as much as you can using comments. Try to identify control sources and destinations. Look up the help files and place keywords in front of all parameters. Then swap out the control sources. For example, change the LFSaw to SinOsc, LFPulse, LFNoise1 or LFNoise0. Each time you hit on an interesting variation, chose "save as" and save the patch for future use. 17:4 Listen to each patch in the 17.11 and 17.12. Draw what you think the sonogram might look like. Open the preferences and set the output to Soundflower16. In Amadeus open a real-time sonogram. Compare your diagram with the actual sonogram.

169

18.

SynthDef and GUIs An interface for humans. Perhaps not worth it?

•

SynthDef

Until now we have been doodling on scratch paper, so to speak—crumpling up each patch and tossing it in the trash when we're done. Each time we send an audio patch to the server, it creates a virtual synth based on our specifications and plays it until ⌘-. is pressed. The post window displays the ad hoc synth definitions with temporary names and node numbers such as Synth("temp__23" : 1000). Using a SynthDef creates a synth, like a preset in Logic Pro, and saves it with a specific name. They take a little more code, but they have several advantages: 1) They are saved to the hard drive and can be recalled even after restarting the computer. 2) They can be called and controlled independently. So far, we've been stopping each example with command period, which halts all processes. 3) The definition can include arguments that can be changed after the synth is running. Finally, (4) since parameters can be changed after the fact, GUI interfaces can be used to add greater control during live performance. The first two arguments for the SynthDef are name and the UGen function. The UGen function is what was attached to the .play message in previous examples with one new wrinkle. In our ad hoc examples the output was automatically set to 0. In definitions the patch must be enclosed, sooner or later, in an Out.ar UGen. SynthDef understands the play and add message. Both are temporary definitions. They will not persist after restart. 18.1::

SynthDef

// ad hoc synth {SinOsc.ar(400, mul: SinOsc.kr(5, mul: 0.3, add: 0.5))}.play; // saved as a "preset" SynthDef("testtone", {Out.ar([0, 1], SinOsc.ar(400, mul: SinOsc.ar(5, mul: 0.3, add: 0.5)))}).add; // Play the synth, and assign it to the variable 'a' a = Synth("testtone") // This time don't use ⌘-period to stop it. Use the following line. a.run(false); ( b = SynthDef("computer", {Out.ar(0, Mix.ar( Array.fill(5, {SinOsc.ar(SinOsc.ar(1/10, rrand(0, 6.0), 200, 500))}))*0.03)}).play ) ( c = SynthDef("string", 170

{Out.ar(1, RLPF.ar(Dust.ar(12), LFNoise1.ar(1/3, 1500, 1600), 0.01))}).play ) a.run(true); b.free; c.free; a.free; ::

•

SynthDef arguments

… allow you to change, as we did with examples in chapter 14, parameters after a synth is running. They are similar to those we've been using, for example, with SinOsc and LFNoise0. To illustrate the internal workings (and the declaration of arguments) in the SinOsc Ugen, the source code can be examined by pressing ⌘I. SinOsc is in the Osc.sc file, and shows: SinOsc : PureUGen { *ar { arg freq=440.0, phase=0.0, mul=1.0, add=0.0; ^this.multiNew('audio', freq, phase).madd(mul, add) }

In this core code file the SinOsc Ugen is defined with arguments that are similar to variables. They are declared and given default values with arg freq=440.0, phase=0.0, etc. The same can be done with synth definitions. The key difference is, and the value is, variables can't be changed once a synth is running, arguments can. If a synth definition is a box that makes sounds, the arguments are the knobs that can be adjusted to change the sound. For the next examples we use load which defines and also writes the def to a file (in the user library). Use play or add for temporary experiments, load for synths you want to continue to use. 18.2:: SynthDef arguments ( SynthDef("RisingSines", // create a synth def { arg out=0, kfreq=0.1, scale=200, offset=500; Out.ar(out, Mix.ar( Array.fill(5, { Pan2.ar(SinOsc.ar(SinOsc.ar(kfreq, rrand(0, 6.0), scale, offset)), LFNoise1.kr(1))}))*0.03)}).load ) // Run these in sequence a=Synth("RisingSines"); // start the synth a.set(\kfreq, 1); // change kfreq a.set(\kfreq, 0.1); a.set(\scale, 1000); a.set(\offset, 1200); // change a group a.set(\out, 1); // change output channel ( SynthDef("GratingString", { arg out=1, trig=12, kfreq=0.3, scale=1500, offset=1600; Out.ar(out, RLPF.ar(Dust.ar(trig), LFNoise1.ar(kfreq, scale, offset), 0.01))}).load 171

) b=Synth("GratingString"); b.set(\kfreq, 4); b.set(\offset, 3000); a.free; b.free; // Start and change several instances of a single synth. Run each line in succession. a = Synth("RisingSines", [\kfreq, 0.2, \scale, 200, \offset, 400]) b = Synth("RisingSines", [\kfreq, 0.03, \scale, 700, \offset, 1000, \out, 1]) b.set(\out, 0); a.set(\out, 1); b.set(\kfreq, 4); a.set(\kfreq, 1, \scale, 4000); a.free; b.free; ::

A SynthDef is not required to create arguments. They can be added to any ad hoc synth and changed after it is running. The SynthDef just commits it to memory. 18.1::

Ad Hoc Arguments

( c = { arg out=0, kfreq=0.1, scale=200, offset=500; Mix.ar( Array.fill(5, { Pan2.ar(SinOsc.ar(SinOsc.ar(kfreq, rrand(0, 6.0), scale, offset)), LFNoise1.kr(1))}))*0.03}.play; ) c.set(\kfreq, 1); c.set(\kfreq, 0.1); c.set(\scale, 1000); c.set(\offset, 1200); c.set(\out, 1); ( d = { arg out=1, trig=12, kfreq=0.3, scale=1500, offset=1600; RLPF.ar(Dust.ar(trig), LFNoise1.ar(kfreq, scale, offset), 0.01)}.play ) d.set(\kfreq, 4); d.set(\offset, 3000); c.free; d.free; :: 172

Which parameters should be set as arguments? It's up to you. I try to keep them manageable by skipping those that are integral to the patch, and unlikely to change. In the "string" patch, for example, 0.01 is the resonance, and probably needs to stay at 0.01. Even though we can now make changes while the synth is running, typing in each change by hand is cumbersome. In a live performance it can be used to carefully orchestrate specific changes, but it does not lend itself to improvisation or exploration. For that, we need a … •

GUI slider

Following is a very simple example. To link an interface with a synth three things have to happen. Start the synth and give it a variable. Create the interface and place it in a variable. Finally, link the action of the interface with the argument you want to control. The first argument for EZSlider is the parent window. In this case, there is none, so it is set to nil. The second is the size of the slider; 400 is the width, 24 is the height. The third is the name of that slider. The forth is a ControlSpec, which sets the minimum, maximum, warp, step value, default, and units. All GUIs belong to the sclang application. You will have to switch back and forth for these experiments. 18.2:: Single Slider ( a = Synth("GratingString"); g = EZSlider(nil, 400@24, " kfreq ", [1/4, 12, \lin, 0.1, 1].asSpec); g.action_({|ez| a.set(\kfreq, ez.value)}); ) ::

A button is a little more involved. It requires an if statement, covered later. This tests to see what state the button is, 0 or 1, and performs one action if true, the other if not. A button can't exist on its own, as the slider example above. It needs to be placed in a parent window. This example assumes the synth definition "string" has already been sent to the server. 18.3:: GUI Button, Parent Window ( b = Button(Window.new.front, Rect(20, 20, 340, 30)) .states_([["Start"],["Stop"]]) .action_({|view| if (view.value == 1) {a = Synth("GratingString")} {a.free}}); ) // Essentially, this, wrapped in a button, wrapped in a window: a = Synth("GratingString"); a.free; ::

Next we can place some sliders in the same window by replacing the 'nil' with a GUI window. The Flowlayout manages each item added to the window and places it in the next logical location, so the order in which you create each interface item matters. We will be dealing with two sets of variables/arguments. It's useful to give them similar names, (e.g. scale), so that 173

it's easy to keep things straight, but not exactly the same. The window will have a scale slider that will be linked to the scale of the "sines" synth. It would probably work to name them both scale, but if you ever want to replace one or the other, they are no longer unique. For that reason I've named the GUI variables with a preceding 'g'. So it's easy to see the relationship in this line: \kfreq, gkfreq.value. But the first kfreq is the argument from the 'sines' synth and the gkfreq is the value returned from the slider. The names could be completely different, but the similarity makes for faster coding, and for slicker shortcuts, covered later. To illustrate how SynthDefs can be reused, the first example builds an interface assuming the "RisingSines" synth has already been defined. If you're starting from scratch, run 18.2 again. The second patch repeats the Theremin earlier but with more arguments, more sliders, an amplitude control, and a control that allows you to select the output bus. 18.4::

Sines and Theremin with Interface

( // Build the GUI interface var inst, gbutton; var gkfreq, gscale, goffset, gWin;

// Create and name a window gWin=Window.new.front; // This line ensures each component is added to the window // in an orderly way. gWin.view.decorator = FlowLayout(gWin.view.bounds); gbutton = Button(gWin, 100 @ 24); gbutton.states = [["Start", Color.black, Color.green(0.7)], ["Stop", Color.white, Color.red(0.7)]]; gbutton.action = {|view| if (view.value == 1) {inst = Synth( "RisingSines")} {inst.free}}; // This looks complicated, but it's just duplication. // The specs array is low value, high value, warp, step, and initial value gkfreq = EZSlider(gWin, 400@20, "KFreq", [0.05, 4, \lin, 0.1, 0.1].asSpec) .action_({|sl| inst.set(\kfreq, sl.value)}); gscale = EZSlider(gWin, 400@20, "Scale", [10, 1000, \lin, 10, 200].asSpec) .action_({|sl| inst.set(\scale, sl.value)}); goffset = EZSlider(gWin, 400@20, "Offset", [100, 3000, \exp, 10, 600].asSpec) .action_({|sl| inst.set(\offset, sl.value)}); ) // The theremin example is more complicated, but again, mostly duplication. ( SynthDef("Theremin", { 174

arg speed=6, depth=0.02, out=0, lDelay = 1.123, rDelay = 1.1, amp = 0.5; var mix; mix = SinOsc.ar( freq: MouseX.kr(200, 3200, 1, 0.5) * SinOsc.kr(speed, mul: depth, add: 1), mul: MouseY.kr(1, 0.02)); mix = mix + CombN.ar(mix, 3.0, [lDelay, lDelay*rDelay], 6); Out.ar(out, mix * amp) }).load ) // After defining the synth, build the interface. ( var synthNode, gbutton, gmenu; var gspeed, gdepth, gout, glDelay, grDelay, gamp, gWin;

// Create and name a window gWin=Window.new.front; // This line ensures each component is added to the window // in an orderly way. gWin.view.decorator = FlowLayout(gWin.view.bounds); gbutton = Button(gWin, 100 @ 24); gbutton.states = [["Start", Color.black, Color.green(0.7)], ["Stop", Color.white, Color.red(0.7)]]; gbutton.action = {|view| if (view.value == 1) {synthNode = Synth( "Theremin", // These lines set the instrument to the current slider location. // This preserves settings when stopping and starting. [\out, gmenu.value * 2, \speed, gspeed.value, \depth, gdepth.value, \lDelay, glDelay.value, \rDelay, grDelay.value, \amp, gamp.value ])} {synthNode.free}}; gmenu = EZPopUpMenu(gWin, 150@24, " Output ", [\1 -> {synthNode.set(\out, 0)}, \3 -> {synthNode.set(\out, 2)}, \5 -> {synthNode.set(\out, 4)}, \7 -> {synthNode.set(\out, 6)}]); // Here we've added a static text that allows a little longer explanation // of each slider. StaticText(gWin, 400@12).string = "Vibrato Speed"; gspeed = EZSlider(gWin, 400@20, "Speed", [0.5, 50, \exp, 0.1, 4].asSpec) .action_({|sl| synthNode.set(\speed, sl.value)}); StaticText(gWin, 400@12).string = "Vibrato Depth"; gdepth = EZSlider(gWin, 400@20, "Depth", [0.01, 50, \exp, 0.01, 0.02].asSpec) .action_({|sl| synthNode.set(\depth, sl.value)}); StaticText(gWin, 400@12).string = "Left Delay"; glDelay = EZSlider(gWin, 400@20, "Left D", [0.1, 3, \lin, 0.1, 0.5].asSpec) 175

.action_({|sl| synthNode.set(\lDelay, sl.value)}); StaticText(gWin, 400@12).string = "Right (percentage of left)."; grDelay = EZSlider(gWin, 400@20, "Right %", [1.0, 2.0, \lin, 0.1, 1.001].asSpec) .action_({|sl| synthNode.set(\rDelay, sl.value)}); StaticText(gWin, 400@12).string = "Volume."; gamp = EZSlider(gWin, 400@20, "Amp ", [0.1, 1.0, \lin, 0.1, 0.5].asSpec) .action_({|sl| synthNode.set(\amp, sl.value)}); ) ::

The button can be toggled with the space bar when the window is front, and the button is highlighted (press the tab key to move focus). The sliders respond to a number of key presses when highlighted: [r] chooses a random value, [c] centers, [n] sets to 0 (nil), [x] to highest value, [ [ ] and [ ] ] as well as the arrow keys increment and decrement the value. If these examples look intimidating, notice they are mostly duplication. The next example is more verbose than previous examples. I'm always too anxious to get to the sounds and use quick variables such as variables a, b, etc. I typically regret it when a year later I can't figure out what my own code does. One could argue that more verbose methods don't take that much more time, they certainly don't use up much more hard drive space, or make any difference in CPU. Also, using more descriptive variables are better if you ever decide to change a screwball name such as synthLFNoiseFreq to lf, it's easier to do a search and replace. If you name all your variables a, b, or c you'd have to replace them by hand. There's always a point in a patch where I decide I'm committed and take the time to clean up and add better variables and comments. 18.5:: Synth with GUI: 2666 // First define the synth ( SynthDef("RisingSines", { arg out=0, sawRate=1.5, sawScale=50, noiseRate=0.5, noiseScale=500, noiseOffset=600, amp=0.1, delay=1.35, panRate=0.1; var mix, effect; mix = SinOsc.ar( abs(LFSaw.kr(sawRate, mul: sawScale, add: LFNoise1.kr(noiseRate, noiseScale, noiseOffset))), mul: amp); mix = Pan2.ar(mix, LFNoise1.kr(panRate)); effect = CombN.ar(mix, 3.0, delay, 6 ); Out.ar(out, mix + effect) }).load; ) // Then create the GUI ( var synthNode, gbutton, gmenu; var gnoiseRate, gsawRate, gpanRate; var gnoiseScale, gsawScale, gnoiseOffset; var gdelay, gamp, gWin; 176

// Create and name a window gWin=Window.new.front; // This line ensures each component is added to the window // in an orderly way. gWin.view.decorator = FlowLayout(gWin.view.bounds); gbutton = Button(gWin, 100 @ 24); gbutton.states = [["Start", Color.black, Color.green(0.7)], ["Stop", Color.white, Color.red(0.7)]]; gbutton.action = {|view| if (view.value == 1) {synthNode = Synth( "RisingSines", [\out, gmenu.value * 2, \noiseRate, gnoiseRate.value, \sawRate, gsawRate.value, \panRate, gpanRate.value, \noiseScale, gnoiseScale.value, \sawScale, gsawScale.value, \noiseOffset, gnoiseOffset.value, \delay, gdelay.value, \amp, gamp.value ])} {synthNode.free}}; gmenu = EZPopUpMenu(gWin, 150@24, " Output ", [\1 -> {synthNode.set(\out, 0)}, \3 -> {synthNode.set(\out, 2)}, \5 -> {synthNode.set(\out, 4)}, \7 -> {synthNode.set(\out, 6)}]); StaticText(gWin, 400@12).string = "Rate that all the saws wander."; gnoiseRate = EZSlider(gWin, 400@20, "Rate", [0.1, 10, \lin, 0.1, 0.1].asSpec) .action_({|slider| synthNode.set(\noiseRate, slider.value)}); StaticText(gWin, 400@12).string = "Frequency of each saw."; gsawRate = EZSlider(gWin, 400@20, "Rate", [0.1, 20, \lin, 0.1, 1].asSpec) .action_({|slider| synthNode.set(\sawRate, slider.value)}); StaticText(gWin, 400@12).string = "Rate of panning."; gpanRate = EZSlider(gWin, 400@20, "Rate", [0.1, 20, \lin, 0.1, 2].asSpec) .action_({|slider| synthNode.set(\noiseScale, slider.value)}); StaticText(gWin, 400@12).string = "Scale of wandering."; gnoiseScale = EZSlider(gWin, 400@20, "Scale", [50, 1000, \lin, 10, 600].asSpec) .action_({|slider| synthNode.set(\noiseScale, slider.value)}); StaticText(gWin, 400@12).string = "Scale of each saw."; gsawScale = EZSlider(gWin, 400@20, "Scale", [50, 1000, \lin, 10, 500].asSpec) .action_({|slider| synthNode.set(\sawScale, slider.value)}); StaticText(gWin, 400@12).string = "Offset of wandering."; gnoiseOffset = EZSlider(gWin, 400@20, "Offset", [10, 1000, \lin, 10, 50].asSpec) .action_({|slider| synthNode.set(\noiseOffset, slider.value)}); StaticText(gWin, 400@12).string = "Delay in seconds."; gdelay = EZSlider(gWin, 400@20, "Delay ", [0.1, 3.0, \lin, 0.01, 1.23].asSpec) .action_({|slider| synthNode.set(\delay, slider.value)}); StaticText(gWin, 400@12).string = "Volume."; gamp = EZSlider(gWin, 400@20, "Amp ", [0.01, 0.2, \lin, 0.01, 0.1].asSpec) .action_({|slider| synthNode.set(\amp, slider.value)}); ) 177

::

What the heck was I thinking? This was such a simple patch to begin with. My first computer music instructor often observed that 95% of coding is interface. Telling a computer what to do is fast and simple. Deciding what you want it to do takes months. Pandering to humans takes years. To illustrate, here is the original (brief) patch. It takes 138 characters; 5% of the 2666 above. 18.6:: No GUI: 138 Characters ( { var out, delay; out = SinOsc.ar( abs(LFSaw.kr(1.5, mul: 50, add: LFNoise1.kr(0.5, 500, 600))), mul: 0.1); delay = CombN.ar(out, 3.0, [1.35, 0.7], 6 ); out + delay }.play ) ::

So I usually find a way around using GUIs. For example, though I could wrap a GUI around nearly every example in this text, I don't. The code required just for the GUI would obscure the components of synthesis, which is our focus. •

Iteration

… can be used to automate code. Wherever there is repetition, there is opportunity for elegance and refinement. I love elegant solutions, and that is what usually keeps me up late at night. But for the most part, I'm a sledgehammer programmer. Previous examples used Array.fill and Mix.fill to automate generating a collection of sounds or elements. Both have two arguments; the number of iterations, and the function used to fill the array. In all these examples the results of each repetition were independent of one another. That is, Array.fill(10, {100.rand}) generates 100 random values and none of them are related. It is often useful to generate an array of sequential values, or lines of code that are related. Array.series and Array.geom allow greater control over the contents, but only in terms of numbers. Another method of iteration is simply called do. The expression 10.do({"ping".postln}) will post 'ping' 10 times. Both of these functions have a built in counter that increments with each repetition. This value is transferred to the function by way of an argument, similar to a variable. The name can be anything. Remember that computers begin counting with 0. I often write the do in functional notation: do(10, {"ping".postln}). In receiver notation the first argument is moved ahead of the message and dot. In functional notation the object is placed as the first argument. The two versions are precisely the same, and this can be done with any Object message pair: 100.rand, rand(100), Array.fill(10, {10.rand}), fill(Array, 10, {10.rand}). Which do you use? It's a matter of style and context. 178

18.7::

Array, Do, Counters

Array.fill(10, {arg myName; myName}); // fill the array with the counter Array.fill(10, {arg anotherName; anotherName*100}); // Since the counter starts at 0, you often have to add 1, for // example with harmonics. Remember precedence is // left to right: harm + 1, then * 100. Try harm * 100 + 1 Array.fill(10, {arg harm; harm + 1 * 100}); 50.do({arg anyNameButAlwaysCounter; "Count: ".post; anyNameButAlwaysCounter.postln}) do(50, {arg count; count.postln}) // This sounds more like English: "do 50." ::

 In addition to numbers, you can do an array of items. In this case two arguments are sent to the function. The first is each item in the array being "done" (used in some sort of function). The second is a counter. Again, they can have any name, but they will always be each item in the array and a counter in that order. Note that the array can be a collection of unrelated data types. There is a shortcut for naming the arguments: |one, two|. Those are pipes. 18.8::

Array Iteration

["One", 2, {10.rand}, [1, 2]].do({ | arrayItem, count | count.post; " is ".post; arrayItem.postln}) [0, 1, 2, 3, 4, 5, 6].do({ | item, count | [item, count].post; " is redundant".postln;}) [0, 1, 2, 3, 4, 5, 6].do({ | item, count | [item + 1 * 500, count].post; " is not redundant".postln;}) // Swapping the names does not swap the meaning. The first argument // will always be each item, the second a counter. ["this", "that", "other", "item"].do({ | count, item | [count, item].post; " still item, count".postln;}) // This generates merged lines of code. ["SinOsc", "LFSaw", "LFNoise1"].do({arg each; ("{" ++ each ++ ".ar}.plot").postln}) // This generates merged lines and executes them. ["SinOsc", "LFSaw", "LFNoise1"].do({arg each; ("{" ++ each ++ ".ar}.plot").interpret}) ::

The counter is useful for generating harmonics, but the array can also contain text or values that differ in the repetitious section of the code; a type of document merge. Multiple sliders are a perfect example. So much of each is repeated that it can be encapsulated in an iteration. The more uniformity we can create the better, so here are two additional shortcuts. First, synth arguments can be set by index number. The second argument of the synth "RisingSines" was sawRate. This value can be set with a.set(\sawRate, 4) or a.set(1, 4). The 179

second work around is naming each slider by number then putting a text message at the bottom explaining what each slider does after the iteration. 18.9::

Sliders Using Do()

( // Create the parent window var gWin; gWin=Window.new.front; gWin.view.decorator = FlowLayout(gWin.view.bounds); // Create duplicate sliders. 10.do({arg each, count; EZSlider(gWin, 350@20, count+1, [0, 10, \lin, 1, 0].asSpec).action_({|slider| slider.value.postln}); }); StaticText(gWin, 350@24).string = "Slider 1 = offset, 2 = scale, 3 = rate, 4 = amp"; ) ::

It's a trade off. This code is more concise, but the resulting dialog is less refined. Choose your battles. I prefer a simpler interface and cleaner code. There are several parameters of the slider that are worth adjusting in each. The specs array; low and high value, warp, step value, and initial value should be different for each. If you left them all 0 to 10, then you would have to write the synth with offsets and scales to change the 0-10 range to, e.g., 1000 and 3000. That seems like a reasonable strategy. But then you wouldn't see the values in the number boxes, which I think is useful. •

Synth Browser

… offers a slicker method. We've used add, play and load when defining a synth. I like play because I'm constantly experimenting and it gives immediate feedback. But for the trick below, you have to use add, or load. Turns out, they can be chained, but add or load must come first. Here is the Theremin patch with load and play. Since we've already defined the Theremin, the file will be overwritten. Once defined, you can browse this and all the synths that have been added using SynthDescLib.read.global.browse. (The read messages loads the saved files. See below for a way to automate this at startup.) 18.10::

SynthDef.load.play

( SynthDef("Theremin", { arg out=0, speed=6, depth=0.02, lDelay = 1.123, rDelay = 1.1, amp = 0.5; var mix; mix = SinOsc.ar( freq: MouseX.kr(200, 3200, 1, 0.5) * SinOsc.kr(speed, mul: depth, add: 1), mul: MouseY.kr(1, 0.02)); 180

mix = mix + CombN.ar(mix, 3.0, [lDelay, lDelay*rDelay], 6); Out.ar(out, mix * amp) }).load.play ) SynthDescLib.read.global.browse ::

At the top of the window is a test and window button. Test starts the synth. Window brings up a simple GUI with an interface for each argument. Select both default and Theremin and press the window button. These interfaces are very similar to those in the previous examples. Notice the sliders for the default instrument (freq, amp, pan) have specs appropriate for each value. The Theremin window only offers a logical specs for the amp slider. The other values just have number boxes. Furthermore, if you try using the up and down arrows, you get values far out of the appropriate range. How does SC know to format the specs of a slider for freq, amp, and pan, but not speed, depth, and delays?

Fig. 18-1 Synth Browser Windows

The answer is in the default control specifications. Out of the box, SC comes with the control specs listed below. It is an identity dictionary (pairing of a symbol with a set of values). When a GUI window is created, if any of the arguments in the synthdef match the default control specs, it uses those specs. Note that freq is 20 to 20000 with 440 as a default starting point, as seen in the default window. \symbol = [minval, maxval, curve, stepval, start, units] \amp = [0, 1, 'amp', 0, 0, ""] \audiobus = [0, 127, 'linear', 1, 0, ""] \beats = [0, 20, 'linear', 0.0, 0, " Hz"] \bipolar = [-1, 1, 'linear', 0.0, 0, ""] \boostcut = [-20, 20, 'linear', 0.0, 0, " dB"] \controlbus = [0, 4095, 'linear', 1, 0, ""] \db = [-inf, 0.0, 'db', 0.0, -inf, " dB"] \delay = [0.0001, 1, 'exp', 0, 0.3, " secs"] \detune = [-20, 20, 'linear', 0.0, 0, " Hz"] \freq = [20, 20000, 'exp', 0, 440, " Hz"] \lofreq = [0.1, 100, 'exp', 0, 6, " Hz"] \midfreq = [25, 4200, 'exp', 0, 440, " Hz"] \midi = [0, 127, 'linear', 0.0, 64, ""] \midinote = [0, 127, 'linear', 0.0, 60, ""] \midivelocity = [1, 127, 'linear', 0.0, 64, ""] \pan = [-1, 1, 'linear', 0.0, 0, ""] \phase = [0, 6.2831853071796, 'linear', 0.0, 0, ""] \rate = [0.125, 8, 'exp', 0, 1, ""] 181

\rq = [0.001, 2, 'exp', 0, 0.707, ""] \unipolar = [0, 1, 'linear', 0.0, 0, ""] \widefreq = [0.1, 20000, 'exp', 0, 440, " Hz"] We got lucky with the amp argument in Theremin. Choosing the name amp, which is in the defaults, allows SC to pair it with the specs above. If we change the name of the argument speed to rate, it will have a nice slider too. Using the defaults is less flexible. A synth may require four rates, and the default rate can be used this way only once. Also, the defaults for minimum, maximum, step, and default may not correspond with how the patch works. One solution is to define your own. Once a spec is defined you can and should reuse it in future synth definitions. Example 18.11 adds control specs for all the patches in this text. The arguments are minval, maxval, warp, step, default, and units (used by some GUIs as a label). If units is blank, the GUI will use the symbol. The last line launches a window for the Theremin synth without having to use the browser window button. Now that speed, depth, lDelay, and rDelay have been added as control specs, they will have sliders for more accurate adjustments. You can run it several times to create several synths, each controlled by its own window. (See also Instr, and Patch from the Crucial library.) 18.11::

Add Control Specs

( ControlSpec.specs.addAll( [ \out -> ControlSpec(0, 16, \lin, 1, 0, " chan"), \in -> ControlSpec(0, 16, \lin, 1, 0, " chan"), \loop -> ControlSpec(0.1, 16, \lin, 0, 1, " chan"), \speed -> ControlSpec(0.1, 24, \lin, 0.1, 6), \depth -> ControlSpec(0.001, 1, \lin, 0.02, 1), \fadd -> ControlSpec(60, 20000, \exp, 1, 440, " offset"), \send1 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \send2 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \send3 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \fmul -> ControlSpec(1, 2000, \lin, 1, 100, " scale"), \lDelay -> ControlSpec(0.001, 3.0, \lin, 0.001, 1.123, " ld"), \rDelay -> ControlSpec(0.001, 3.0, \lin, 0.001, 1.6, " rd"), \dur -> ControlSpec(0.1, 5.0, \lin, 0.1, 1, " secs"), \damp -> ControlSpec(-1.0, 1.0, \lin, 0.1, 0.3), \slider1 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \slider2 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \slider3 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \slider4 -> ControlSpec(0, 1, \lin, 0.01, 0.5), \lfospeed -> ControlSpec(0.1, 200, \exp, 0.1, 1), \lfoscale -> ControlSpec(100, 500, \exp, 1, 200), \pindex -> ControlSpec(1, 50, \lin, 0.1, 2.5), \fmlfo -> ControlSpec(0.1, 200, \exp, 0.1, 1), \fmscale -> ControlSpec(100, 2000, \exp, 1, 1500), \fmoffset -> ControlSpec(2000, 4000, \exp, 10, 3000), \lfo2speed -> ControlSpec(1/4, 10, \lin, 0.1, 4, " lfo2sp"), 182

]); // Create a GUI window for this synth SynthDescLib.global[\Theremin].makeWindow; ) ::

An alternate method is to modify an existing specification. Example 18.12 illustrates a patch using the default freq argument after its step value, minval, and maxval have been changed. But if you're going to change that many arguments, might as well define a new spec. 18.12::

Modify Control Spec

( \freq.asSpec.step = 100; \freq.asSpec.minval = 1000; \freq.asSpec.maxval = 2000; SynthDef("ChangeSpec", {|freq = 1100| Out.ar(0, SinOsc.ar(freq))}).add; SynthDescLib.global[\ChangeSpec].makeWindow; ) ::

•

Startup File

As you become more adept at SC you will want to do some customization. Control specs, record defaults, variables, synth browser, or synth windows can be loaded at startup. For example, since I work in so many different locations, I use the Theremin as a test tone, checking my system for sound. To create or edit an existing startup choose File/Open startup file. Place the ControlSpec lines (and, if you like, the Theremin synth and SynthDescLib line) in the file and it will be run each time you launch or recompile SC. I use the following file to set the record depth, open the server window, create a test tone, and add control specs. 18.13::

Startup.rtf

s.recSampleFormat="int16"; s.makeWindow; SynthDescLib.read; ControlSpec.specs.addAll( [ \out -> ControlSpec(0, 16, \lin, 1, 0, " chan"), \speed -> ControlSpec(0.1, 24, \lin, 0.1, 6, " speed"), \depth -> ControlSpec(0.001, 1, \lin, 0.02, 1, " depth") // etc. ]); // A handy test tone SynthDescLib.global[\Theremin].makeWindow; 183

::

•

Exercises 18:1 Create a startup file with customized code. Add a synth definition and control specs. 18:2 Choose your favorite patches from previous chapters and create a synth definition with GUI interface.

184

19.

Filtering, SoundIn The methods of synthesis; starting with filters.

Real instruments have resonant qualities. The body of the instrument filters and shapes the character of the sound. Humans have one of the most complex and flexible filters attached to the top of their necks. The raw sound produced by vocal chords alone resemble a kazoo, but are still a rich fundamental source. (Open wide and say "aah." That's what you would sound like without a head.) The tongue, sinus cavities and shape of the mouth filter the raw vocal chords into the consonants and vowels of speech. Filtering, or subtractive synthesis begins with a sound or wave that has a rich spectrum and removes harmonics. •

RLPF

… (resonant low pass filter) filters frequencies above a given threshold (hence, lowpass—see also high-pass filter and band-pass filter). The arguments for RLPF are input, frequency cutoff, and rq. The cutoff is where filtering begins. The rq is the width of the band. In musical terms a narrow rq (low number) focuses the filter so sharply on the cutoff that nearly any source will resonate on that pitch. These are the same parameters we explored in Logic's EQ plug-ins. The big difference, the point in fact, is that an RLPF is voltage controlled. To illustrate, we will use SoundIn to listen to and filter the ambient noise in your environment. (You might want to move to a noisy location, but also try speech, instruments, etc.). First, since we've experimented with Soundflower, set the audio device back to your default input using device(nil). In example 19.1 the poll message posts the value of MouseY. Note at which value a pitch emerges. Be careful: of both feedback and too narrow of rq. Both can have startling results. Use headsets and turn the volume down. 19.1::

SoundIn, RLPF

s.options.device_(nil); s.reboot ({RLPF.ar(SoundIn.ar, MouseX.kr(100, 5000, [1, 1]), MouseY.kr(0.05, 1.0).poll)}.play ) ({RLPF.ar(SoundIn.ar, LFNoise0.kr([10, 6], mul: 300, add: 500), MouseY.kr(0.05, 1.0).poll)}.play ) ({RLPF.ar(Saw.ar(40, 0.2), LFNoise0.kr([10, 6], mul: 300, add: 500), 0.05)}.play) ::

The second example places an LFNoise0 control on the frequency cutoff. It sounds like a VCO (controlling frequency), but that is just the narrow Q, which brings out a pitch. Try replacing the Saw with LFPulse, LFTri, and PinkNoise. Be sure to check the parameters to see if 185

they match. In Saw they are frequency (40 is used so there are plenty of harmonics in the 200 to 800 range) and mul. To illustrate how effectively a narrow Q can resonate, the next example uses a single trigger: Dust as a "rich" source. The first line is Dust without any filtering. The second is Dust with a very narrow filter. This example sounds quite natural because many real instruments work this way; some type of excitation, the source (a pluck, tick, strike or bow) that is filtered by the resonance of the instrument's body. 19.2::

RLPF With Dust

{Dust.ar([13.1, 15.6])}.play {RLPF.ar(Dust.ar([12.1, 15.6]), LFNoise1.ar(1/5, 1000, [1100, 2200]), 0.02)}.play ::

•

Logic ES1, ES2

Both the ES1 and ES2 are primarily filtering synthesizers. The examples above can be recreated in the ES1 using the settings in Figure 19.1. For a noise source set the Wave Mix to Sub, choose the noise waveform. The resonance is set very high and the cutoff is about in the middle (this gives the filter some headroom so that the LFO control has room to work). The LFO is a step random wave, 7 Hz, set to control the cutoff. Try adjusting the Resonance, the Cutoff, the Rate, the LFO Wave, and the Oscillator waveform. Fig. 19-1 ES1, Filter

With either the ES1 or ES2, peruse the presets, find something interesting and make your own modifications. See if you can predict what each dial will control, or the sound that will result from a modification. •

Complex Resonance

Resonant low pass filters are single pole, that is, they only filter a single frequency. Though distinct, and perhaps retro (used heavily in the 70s—to warn an unsuspecting teen of the slasher in a closet), they now sound cliché, immature. Fig. 19-2 Single Pole Filter Sweep

Figure 19.2 shows a sonogram of a filter sweep using a single pole resonant filter (like Bright Swell Pad in the ES1 or . Compare with the sonogram below of a real bell in Figure 19.3. Note that each harmonic has a different and independent envelope. One harmonic does not even appear until several others have died away. For more complex filtering, we can use …

186

•

Klank

… a component modeling filter that imitates the resonant nodes of a body or space. It allows you to specify a set of frequencies to be filtered (resonated), matching amplitudes, and independent decay rates. Fig. 19-3 Complex Resonance

The first argument for Klank is the frequency specifications. It is an array of arrays (two dimensional array). That is each item in the array is also an array (like a box filled with boxes). The outer array begins with a "`" (back tick? accent grave?) to disable multi-channel expansion. (If this character were left off, SC would expand the array into different channels. Go ahead and try it.) The first inner array is the set of frequencies that resonate, the second is the amplitude of each frequency (default is 1), and the last is the decay rates of each frequency. The first example in 19.3 is a single frequency in the array. Filtering a single frequency has the exact result of an RLPF in the examples above. The second example fills the arrays using Array.series. The third example loads an array with random frequencies. The last example adds an amplitude control: an LFTri. Its output is bipolar, that is, -1 to 1. Placing it inside a max with 0 as an argument returns the greater of 0 and the LFTri. Negative numbers become 0, creating a nice envelope with silence between. Run this last example several times to confirm the frequency array is random and to create a group of rotating wheels. 19.3::

Klank

// Single pole {Klank.ar(`[[300]], PinkNoise.ar(MouseX.kr(0.01, 0.04)))}.play ( { // Harmonic resonance Klank.ar( `[[100, 200, 300, 400, 500, 600, 700, 800, 900, 1000]], PinkNoise.ar([0.01, 0.01])) }.play ) ( { // Using Array.series Klank.ar( `[Array.series(10, 50, 50).postln, Array.series(10, 1.0, -0.1).postln], ClipNoise.ar([0.005, 0.005]) ) }.play ) ( { // Enharmonic Spectrum; 187

Klank.ar( `[{exprand(60, 10000)}.dup(15)], PinkNoise.ar([0.005, 0.005]) ) }.play; )

( SynthDef("Wheel", { arg amp=0.7, out=0; Out.ar( out, Pan2.ar( Klank.ar( `[{ExpRand(60, 10000)}.dup(15)], PinkNoise.ar(0.005) * // Try LFNoise1 and LFNoise0 max(0, LFTri.kr(Rand(1/3, 5)))), Rand(-1.0, 1.0))) }).load; ) Synth(\Wheel); 6.do({Synth(\Wheel)}) ::

The sound is rich, natural, evolving, because the excitation source (input) is noise. A violin bow has a similar effect. The physical body or space described by Klank extracts patterns from the noise, but the noise gives variety to the body or space. The noise source by itself is too chaotic to be interesting. The single pole filter sweep is too predictable and manufactured. This last example strikes, for me anyway, a satisfying balance between the two. •

Bells

… are resonant bodies. The clapper striking the side of a bell is a burst of noise. (Try holding a bell so it doesn't resonate and hit the striker. You just hear a short tick.) The shape and material of the bell or chime resonates on certain frequencies. Those frequencies are sustained and all others die away quickly. The remaining frequencies are what make up the character of the bell or chime. The bell patch below begins with an extremely short burst of noise. Fig. 19-4 Noise Burst

The image in Figure 19.4 shows three of these bursts. Each is unique. Though they are only a thousandth of a second, our ears (brains) can distinguish them. The first example in 19.4 produces these bursts. Notice the slight differences in each. Because the bursts are different they excite the resonance of frequency banks in Klank differently, and each strike of the bell produces similar but distinct sounds. Likewise, run the last example several times to create different bells. 188

19.4::

Chimes

( { var burstEnv, att = 0, burstLength = 0.0001, strike; burstEnv = Env.perc(0, burstLength); strike = PinkNoise.ar(EnvGen.kr(burstEnv, gate: Impulse.kr(1))); strike; }.play ) ( { var chime, freqSpecs, burst, totalHarm = 10; var burstEnv, att = 0, burstLength = 0.0001; freqSpecs = `[ {rrand(100, 1200)}.dup(totalHarm), //freq array // .normalizeSum ensures sums to 1 {rrand(0.3, 1.0)}.dup(totalHarm).normalizeSum, //amp array {rrand(2.0, 4.0)}.dup(totalHarm)]; //decay rate array burstEnv = Env.perc(0, burstLength); burst = PinkNoise.ar(EnvGen.kr(burstEnv, gate: Impulse.kr(1))); Klank.ar(freqSpecs, burst)*MouseX.kr(0.1, 0.8) }.play ) ( { // The triggers are random, so you may have to wait for sound var chime, freqSpecs, burst, totalHarm = 10; var burstEnv, att = 0, burstLength = 0.0001; freqSpecs = `[ {rrand(100, 1200)}.dup(totalHarm), //freq array {rrand(0.3, 1.0)}.dup(totalHarm).normalizeSum, //amp array {rrand(2.0, 4.0)}.dup(totalHarm)]; //decay rate array burstEnv = Env.perc(0, burstLength); burst = PinkNoise.ar(EnvGen.kr(burstEnv, Dust.kr(1/3))); Pan2.ar( Klank.ar(freqSpecs, burst)*MouseX.kr(0.1, 0.8), 1.0.rand2) }.play ) :: 189

Experiment with a different range of harmonics and fewer or more total harmonics. Use Array.fill to tune the harmonic frequencies as with the examples above. If the harmonics are detuned slightly, it will sound a little more like a tuned bell. How would you detune the harmonics by + or – 10%? Change the range of choices for decay rates. These chimes sound very natural, realistic. That is because they are chaotic. Each burst of noise is different, and the resonant body is excited differently, and the harmonics ring differently. The code does not describe a single sound, but the process of creating a sound. It describes the body of the instrument, its nature, then sets it in motion with a chaotic event. Though Klank is a filter, it is also an example of physical modeling, covered later. The previous examples show both harmonic and inharmonic collections of frequencies. How can you be sure the frequency collection is what you wanted (random, odd, even, detuned, etc.)? Likewise, how do you go about isolating problems in a patch that isn't doing what you want? By … •

Debugging

When something isn't quite right, one of my first steps is to isolate and evaluate small pieces of code. The example below shows an entire line of code with smaller sections in red. These smaller sections can be selected and evaluated individually. An earlier patch is reproduced below it. See if you can find sections that can be evaluated in isolation. 19.5:: Array.fill(5, Array.fill(5, Array.fill(5, Array.fill(5,

Isolate, Evaluate {exprand(10, {exprand(10, {exprand(10, {exprand(10,

100)}).round(1) 100)}).round(1) 100)}).round(1) 100)}).round(1)

* * * *

([1, ([1, ([1, ([1,

2, 2, 2, 2,

5, 5, 5, 5,

7].choose) 7].choose) 7].choose) 7].choose)

( { // Run this several times Pan2.ar( Klank.ar( `[{exprand(60, 10000)}.dup(15)], PinkNoise.ar(0.005) * max(0, LFTri.kr(rrand(1/6, 5)))), 1.0.rand2) }.play; ) ::

You have to be careful about commas and variables. For example, if you select and evaluate this section: {rrand(100, 1200)}.dup(totalHarm), you will get an error because of the comma and missing variable declaration for totalHarm. But if you replace the totalHarm with 10 and select everything but the comma, it can be evaluated separately and SC will print the results in the post window.

190

Two other methods for finding problems: you can copy and paste a section of code to another file or outside your patch, fiddling with it until it produces what you need. For example, the max(0 LFTri.kr(rrand(1/6, 5)) could be placed in a function with a plot. Also, ⌘E brings up a command line for evaluating code sections immediately. You can also isolate problems by replacing suspect parts of the code with safer values, and by removing a section with comments. Comments are identified by "//" or "/*" to begin the comment and "*/" to end the comment. Again, you have to be careful about semicolons and commas. If you suspect a problem with the freqSpecs arrays you could insert a single "safe" value in place of the random array: {0.2 /*rrand(0.3, 1.0)*/}.dup(totalHarm). This will fill the entire array with 0.2 rather than the rrand(0.3, 1.0). Array.fill(5, {exprand(10, 100)}) /* .round(1) * ([1, 2, 5, 7].choose) */ Lines of code are evaluated from top to bottom. This means you can replace suspect lines by simply placing new lines just below. Place the cursor on the second line and you now have an A/B switch that is turned on and off by toggling comments (⌘/). Run the first example, then uncomment the "this I know works" and run it again. The sentence "this is experimental" is simply replaced. 19.6::

Safe Alternatives

( a = "this is experimental"; // a = "this I know works"; a; ) ( // an example in real code { var chime, freqSpecs, burst, totalHarm = 10; var burstEnv, att = 0, burstLength = 0.0001; freqSpecs = `[{rrand(100, 2000)}.dup(10)]; // experimental // freqSpecs = `[[100, 200]]; // safe Klank.ar(freqSpecs, PinkNoise.ar(0.01)); // SinOsc.ar // test tone if I get no sound at all }.play ) ::

When the second freqSpecs is commented it is not included in the patch. If you remove the comments then the safe section will replace the experimental section. (You don't have to comment out the experimental section, it is just reassigned.) If the patch then runs correctly, there is something wrong with the experimental part. If there is no sound at all, you can narrow down the issue (e.g., is it a system problem, an SC malfunction, or something in the code) with a simple SinOsc.ar as the last line of the function and that line will be played while everything above it is ignored. 191

•

Monitoring Data

… will also help to isolate issues. The next example uses post and poll messages to print the results of an operation to the Data window. This is similar to selecting and evaluating a section. The difference is that you can evaluate several places at once. There are intentional errors in the patch below. When you run it, it will explode into several screen's worth of error data. But the first lines should reveal the items connected to a postln. Can you spot them by examining the posted data? (Hint: nil means no data has been assigned.) Press ⌘P to clear the post window. 19.7::

Post And Poll

( // intentional errors { var chime, freqSpecs, burst, totalHarm = 10; var burstEnv, att = 0, burstLength; freqSpecs = `[ rrand(100, 1200).dup(totalHarm).postln, {rrand(0.3, 1.0)}.dup(totalHarm).normalizeSum.round(0.1).postln, {rrand(2.0, 4.0)}.dup(totalHarm).round(0.1).postln]; burstEnv = Env.perc(0, burstLength.postln); //envelope times burst = PinkNoise.ar(EnvGen.kr(burstEnv, gate: Impulse.kr(1))/*.poll*/); Klank.ar(freqSpecs.postln, burst)*MouseX.kr(0.1, 0.8)/*.poll*/ }.play ) ::

The first error is the fourth item posted—burstLength, which was nil. The second error is revealed by the first postln; all the values in the array are the same because the rrand is not enclosed in a function. Once these errors are fixed, uncomment the poll messages to post the values of MouseX and EnvGen as the patch runs. The postln message can be nested within a message chain, as shown in this example. You can remove them using find and replace. 19.8::

Nested Postln

( [[60, 62, 68], [65, 67], [61, 63, 55, 23]].choose.postln.midicps.postln .round(0.1).postln.reciprocal.postln.round(0.0001).postln ) ::

The post message prints to the post window, postln prints the object then begins a new line. If you use post repeatedly the information is merged into one blob. The obvious solution is

192

to use postln, but that eats up the post window quickly. It is also difficult to examine related data if everything is on a separate line. The solution is to format the output by placing it in an array or to use scatArgs, which concatenates a string with a list of arguments, and finally postf, which has as its first argument a string containing text interspersed with the "%" character. Following that are additional arguments that will be placed in the string where the "%" characters were (in order). For the second example there are three methods. In this case, when you uncomment a new alternative, you must comment the current postln line. Otherwise they will both print. 19.9::

Formatting Posted Information

( // too spread out var pitch, duration, amplitude, voice; 20.do({ pitch = [60, 62, 64, 65, 67, 69, 71].choose.postln; duration = 4.0.rand.round(0.1).postln; amplitude = 1.0.rand.round(0.1).postln; voice = 10.rand.postln; }) ) ( var pitch, duration, amplitude, voice; "pitch, duration, amplitude, voice".postln; 20.do({ pitch = [60, 62, 64, 65, 67, 69, 71].choose; duration = 4.0.rand.round(0.1); amplitude = 1.0.rand.round(0.1); voice = 10.rand; // Collect variables in an array [pitch, duration, amplitude, voice].postln; // or prepend identifiers // "p d a v".ccatArgs(pitch, duration, amplitude, voice).postln; // interleve the identifiers %. Also, \t = tab, \n = new line // "p %\t d %\t a %\t v %\n".postf(pitch, duration, amplitude, voice) }) ) ::

In strings a backslash has special meaning. It is an escape character which gives the character immediately following special meaning. The most useful are \t, \n, \\, and \".

193

•

Exercises 19:1 Begin with two ES1 presets. Generate three new patches by adjusting each of these items: Oscillator Wave and Sub, Cutoff and Resonance, Low Frequency Wave and Rate, and ADSR. When you create something interesting, save it as a preset. 19:2 Open the ES2 instrument. Experiment with the randomize feature. Save three of your favorite results. 19:3 Modify the "Wheel" patch. Change the number of harmonics and the range of harmonics. How many harmonics push the CPU to the limit? Try a wide and narrow range of harmonics. Pick three that represent the most difference from the original and name the patch accordingly; bent metal, crotale, sheet metal, explosion, etc. 19:4 Add "safe" values to the "Wheel" patch. Find sections of code that can be isolated. Add monitoring points with formatted posts.

194

20.

Phase Modulation Formerly AM, FM. Rich, and efficient.

The vibrato example in Chapter 16 used one oscillator in low frequency to control another in audio range. Hopefully you ignored my warning not to exceed 20 Hz for the speed of the control frequency. When the control rate of this kind of patch moves into audio range something unexpected, almost magical occurs: new frequencies appear above and below the frequency being controlled. The vibrato morphs into a blend of frequencies. Several examples in this chapter (and from now on) assume you have run example 18.12 to add control specs. They will run without this, but the GUI window is not as accurate. •

Frequency Modulation

Example 20.1 illustrates the transition from low frequency modulation to high frequency modulation. As you increase the LFO speed the wide vibrato will slowly evolve into a much more complex sound. You might recognize an antique radio being tuned because controls that enter audio range become frequency and amplitude modulation; the same technology used by AM and FM bands for radio transmissions. 20.1::

From LFO to FM

( SynthDef("LFOtoFM", {|lfospeed = 1, lfoscale = 200| Out.ar(0, SinOsc.ar(SinOsc.ar(lfospeed, mul: lfoscale, add: 800))) }).add; // Run 18.12 first SynthDescLib.global[\LFOtoFM].makeWindow; ) ::

Fig. 20-1 LFO to FM

 The magical frequencies are upper and lower sidebands; the product of the two modulated frequencies. The example below uses scope. The scope belongs to the sclang program. Once it appears, move it to the lower right hand corner then return to SuperCollider. Alternately, you can replace all scopes with play and use SoundFlower to route audio to Amadeus, then open oscilloscope and sonogram. Notice in the scope that you can create patterns identical to the interference patterns of intervals from chapter 8. 195

20.2::

AM Synthesis

{SinOsc.ar(500, mul: SinOsc.ar(MouseX.kr(1, 400), mul: 0.5))}.scope(1) {Saw.ar(500, mul: SinOsc.ar(MouseX.kr(1, 400), mul: 0.5))}.scope(1) ::

First impressions of AM and FM are distinctly sci-fi, alien. What starts out as a natural vibrato, when extended to extremes, becomes a uniquely synthetic sound. Yet FM modulation occurs in the natural world, specifically, the songs of birds—Mother Nature's little FM and AM modulation synthesizers. Their vocal chords (a detailed description of which, even if within my grasp, is not going to happen here) are arranged such that one can modulate the other. 20.3 is a patch that simulates the AM and FM effect in bird song. The speed of the two controls are set to MouseX and Y UGens. Slight movement seems most effective. The second example adds a GUI, though I think the mouse version is better. 20.3::

AM, FM, Bird Song

( { var fm, am; fm = SinOsc.ar(MouseY.kr(1, 100), mul: 1500, add: 3000); am = SinOsc.ar(MouseX.kr(1, 100), mul: 0.5, add: 0.5); Pan2.ar(SinOsc.ar(fm, mul: am * max(0, Pulse.kr(1/2)))) }.play )

( SynthDef("Syrinx", { |fmlfo = 15, fmscale = 1500, fmoffset = 3000, lfospeed = 20, lfo2speed = 4| var fm, am, pitchmod; fm = SinOsc.ar(fmlfo, mul: fmscale, add: fmoffset); am = SinOsc.ar(lfospeed, mul: 0.5, add: 0.5); pitchmod = LFNoise1.kr(lfo2speed, 0.3, 1); Out.ar(0, Pan2.ar(SinOsc.ar(fm * pitchmod, mul: am * max(0, Pulse.kr(1/2))))) }).add; // Run 18.12 first SynthDescLib.global[\Syrinx].makeWindow; )

::

 The oscillator that is being controlled is the carrier and the controlling oscillator is the modulator. In amplitude modulation the sidebands produced are the sum and difference of the two. If the carrier is 500, and the modulator is 50, then the two sidebands will be 550 and 450. The carrier will also be heard; three frequencies total. 196

 In FM the sidebands are also the sum and difference of the carrier and modulator, but many more sidebands can be generated. The exact number depends on the modulation index. The modulation index is how far the modulating frequency deviates from the carrier (the amplitude of the modulating wave). If the carrier is 500, and the modulator is 50, then the sidebands will be 550, 600, 650, etc., and 450, 400, 350, 300, etc. (depending on the index). In SC both the carrier and modulator frequencies can appear as arguments in a single SinOsc. For clarity I've used this form: 400 + SinOsc(124, mul: 100). The 400 is the carrier frequency, 124 is the modulator frequency and 100 is the index. 20.4::

FM Synthesis

{SinOsc.ar(400 + SinOsc.ar(124, mul: 100), mul: 0.5)}.play ::

FM synthesis was developed by John Chowning during the 60s and used extensively in the 80s, specifically in the Yamaha DX7 series. Its success comes from its efficiency; using only two waves to generate theoretically unlimited sidebands of harmonic or inharmonic (but still mathematically related) spectra. In subtractive synthesis you can only remove what is already there. The advantage of FM is the ability to generate and change the harmonic content. The most common sources for filtering are noise or waves with harmonic spectra; square, saw, triangle. These overtones are whole number multiples of the fundamental; x1, x2, x3, x4, x5, etc. In FM, the sidebands (ergo, 'harmonics') are the determined by the ratio of the carrier to the modulator. A 440 Hz carrier and 67 Hz modulator (a ratio of 0.15227) will generate one set of sidebands. Changing the carrier only slightly—65 Hz (a ratio of 0.1477)—will produce a completely different set of sidebands. Below is a chart showing sidebands for a 440 Hz carrier and modulators of 67, 65, and 65.13. Modulator

Sidebands

67 65 65.13

[ 38, 105, 172, 239, 306, 373, 507, 574, 641, 708, 775, 842 ] [ 50, 115, 180, 245, 310, 375, 505, 570, 635, 700, 765, 830 ] [ 49.22, 114.35, 179.48, 244.61, 309.74, 374.87, 505.13, 570.26, 635.39, 700.52, 765.65, 830.78 ]

Remember that the character of a wave is determined by the presence and strength of harmonics (or sidebands). Each of these sets of sidebands will have a different character. The possible variations in sound quality are literally endless. The code for calculating sidebands is ((-n..n)*m+c).abs.sort.round(0.01), where m is modulator, c is carrier, and n is the number of sidebands produced. 20.5::

Complex Sets of Sidebands

20.do({((-20..20)*rrand(100, 1000) + rrand(100, 1000)).abs.sort.round(0.01).postln}); ::

And of course the ratio can be voltage controlled, therefore changed in real time. With filtering, even if you began with a source containing any of these sets of harmonics, you can only remove or add them. You can't modulate them on the fly. 197



•

To summarize the appeal of FM: • Efficiency—rich spectra with only two oscillators • Complex spectra—limitless patterns of harmonics • Voltage controlled spectra—harmonics can be changed in real time. Phase Modulation

SC's PMOsc (phase modulating oscillator—Marvin the Martian had one) is a more efficient cousin to FM. The difference is academic, and the results are the same. Example 20.6 illustrates the efficiency of PM. The first uses a mouse control for exploration, similar to the FM example. The second uses static values, but note the CPU and number of UGens in the server window. The third creates a similar set of sidebands using additive synthesis (covered later). It uses nearly ten times the number of UGens and three to four times the CPU. 20.6::

Phase Modulator Efficiency

( SynthDef("PMOscEx", {|fmlfo = 100, pindex = 2.5| Out.ar(0, PMOsc.ar(400, [fmlfo, fmlfo*2], pindex), mul: 0.5)}).add; SynthDescLib.global[\PMOscEx].makeWindow; ) {PMOsc.ar(1000, 1367, 12, mul: EnvGen.kr(Env.perc(0, 0.5), Impulse.kr(1)))}.play ( { Mix.ar( SinOsc.ar(abs((-20..20)*1367 + 1000), mul: 0.1*EnvGen.kr(Env.perc(0, 0.5), Impulse.kr(1))) )}.play ) ::

 The ratio of the carrier and modulator determine the sidebands and hence the quality or timbre, the index controls the number of sidebands, hence the brightness of the wave. The carrier is more or less the fundamental pitch. Changing the carrier while the modulator and index remain the same will sound like one instrument with a changing pitch. Changing the modulator only will sound like different instruments playing the same pitch. Changing the index will sound like one instrument being filtered. Of course, you can change all three. 20.7::

Phase Modulation

( {PMOsc.ar(LFNoise0.kr(5, 300, 700), // carrier (pitch) 134, 4, mul: 0.4) }.play ) ( {PMOsc.ar(700, 198

LFNoise0.kr(5, 500, 700), //modulator (timbre) 12, mul: 0.4 )}.play ) ( {PMOsc.ar(700, 567, LFNoise0.kr(5, 6, 12), //index (filter) mul: 0.4 )}.play ) ( // All three. {PMOsc.ar(LFNoise0.kr([9, 9], 300, 700), LFNoise0.kr([9, 9], 500, 700), LFNoise0.kr([9, 9], 6, 12), mul: 0.5 )}.play ) ::

Similar to the ratios that give us musical intervals, if the sidebands are related by low ratios, then they will blend into a clear pitch. For this reason it is useful to calculate the modulating frequency as a ratio of the carrier frequency (freq*ratio). Example 20.8 uses a slider interface with a numerator and denominator (super-particular just out of convenience). Notice that low ratios—2/1, 3/2, 4/3—are mild, clear pitches. But as the ratio increases, the timbre darkens. Each of these ratios is reflected in the value of the fine tune slider. Use this slider to adjust the ratio to more complex numbers (e.g., 2.293). This will generate more distantly related sidebands and more obscured pitch. The result is more metallic, bell-like timbres. Values that are very close to a simple ratio, such as 2.01, will detune a harmonic spectrum slightly. The other advantage of using a ratio is that the modulator is linked to the carrier. When the carrier (which is the pitch we hear, if a simple ratio) changes, the collection of sidebands will remain the same in relation to the carrier. The practical result is a consistent timbre, even though the pitch changes. Notice I am using MIDI numbers for frequency. 20.8::

PM Ratio

( SynthDef("PM", { arg ratio=1, rate=1, amp=0.4, low=24, hi=48, index=10; var freq, mix, env; env = EnvGen.kr(Env.perc(0, 2/rate), Impulse.kr(rate)); freq = TIRand.kr(low, hi, Impulse.kr(rate)).midicps; mix = PMOsc.ar(freq, freq*ratio, //modulator index*env, mul: amp); Out.ar(0, mix*env); }).add; 199

)

( // Interface w=Window.new.bounds_(Rect(20, 400, 500, 160)).front; w.view.decorator=FlowLayout(w.view.bounds); w.view.background = Color.rand; Button(w, 100@20) .states_([["Start"],["Stop"]]) .action_({|view| if (view.value == 1) {a = Synth("PM")} {a.free}}); EZSlider(w, 392@16, " ", [1, 12, \lin, 1, 3].asSpec, {|ez| var ev; ev = ez.value; h.value_(ev + 1); i.value_(ev); g.value_((ev+1)/ev); a.set(\ratio, (ev+1)/ev)}, numberWidth: 0); h = EZNumber(w, 100@20, "num ", [1, 12, \lin, 1, 1].asSpec); i = EZNumber(w, 100@20, "den ", [1, 12, \lin, 1, 1].asSpec); g = EZSlider(w, 392@16, "fine ", [0.0, 4.0, \lin, 0.001, 1].asSpec, {|ez| a.set(\ratio, ez.value)}); EZSlider(w, 392@16, "rate ", [1/10, 12, \lin, 0.1, 3].asSpec, {|ez| a.set(\rate, (ez.value))}); EZRanger(w, 392@16, "range ", [12, 96, \lin, 1, 24].asSpec, {|ez| a.set(\low, ez.value.at(0), \hi, ez.value.at(1))}, [24, 48]) ) ::

These timbres are not possible using noise as a source or any periodic waveform alone. They are unique to FM/PM, at least in the synthesis realm. They do occur naturally in bells, birds, (bangles and beads). But we're not done yet. Controlling the ratio, therefore the sidebands, in real time produces results that cannot be found anywhere else, except maybe the Krell laboratory on Altair IV. 20.9::

Voltage Controlled PM Ratio

( { var env, trig, rate, carrier, mix; rate = 1/2; trig = Dust.kr(rate); env = EnvGen.kr(Env.perc(1/rate, 0.1), trig); carrier = LFNoise1.kr(1/20, 500, 2000); mix = Pan2.ar( PMOsc.ar( carrier, carrier * LFNoise0.kr(12, 0.2, 2), 10 * env ) * env, LFNoise1.kr(1/20)); mix + CombN.ar(mix, 3, 0.167); 200

}.play ) ::

Here is an earlier example that pairs each of the parameters with a demand sequence. 20.10::

PMOsc and Demand Sequence

( {

var envelope, rate = 9, pitch, trigger, mod, seq, oct; trigger = Impulse.kr(rate); envelope = Linen.kr(trigger, 0, 1, 1/rate); seq = Array.fill(10, {Dseq(Array.fill(12, {[0, 2, 4, 5, 7, 9, 10].choose}) + 48, rrand(2, 6))}); pitch = Demand.kr(trigger, 0, Dseq(seq, inf)).midicps; mod = Demand.kr(trigger, 0, Dseq([1.5, 1.5, 1.5, 1.65, 1.25, 2.223, 2.567], inf)); PMOsc.ar(pitch, pitch*mod, envelope * 6, mul: envelope) * 0.3 }.play ) ::

•

EFM1

Logic, of course, has an FM synthesizer—the EFM1. Having worked through the previous examples, you should recognize the carrier frequency, the modulator frequency, and modulation envelope. The FM depth is equivalent to index in PMOsc. Selecting fixed links the modulator and carrier as we did with a ratio. Scroll through the presets and note the similarities in the bell-like patches to our experiments. Of course, SC allows much more flexibility in designing control sources for each parameter. There is no random LFO in the EFM1, but there is a randomize button, which you should try. •

Logic Pro Modulators, Ringshifter

In addition to the EFM1 instrument, Logic has a good collection of modulation plugins. Most are variations of chorus, flangers, and phase shifters. These all work by duplicating the signal and shifting the phase, exactly what we said you should avoid when placing mics in pairs. Choruses are generally used to thicken the sound. Flangers are more extreme versions that created a distinctive swirling effect; Experienced, Hendrix, Listen to the Music, Doobies (just before the 'lazy flowing river'), Owner of a Lonely Heart, Yes ('oh much better than…'), and Killer Queen, Queen ('gunpowder, guillotine, dynamite with a laser beam'). By far the most interesting Logic insert is the Ringshifter, which applies FM techniques to any track. Start with the presets and explore by adjusting the knobs. The side chain always looks promising, but it's never given me satisfying results. •

Latch (Sample and Hold)

While we're listing missing features, sample and hold is a control source that was lost in the transition from modular synthesizers to digital workstations. They are a type of sequencer, but capable of a greater variety in patterns; more variety than a sequencer, but more cohesive 201

than a noise source. Though I haven't researched it, I'm pretty sure sample and hold units led to the notion and subsequent revolution of digital audio. It is certainly the logical extension. Digital audio is simply a sample and hold with a very high sample rate. The effect of Latch is similar to aliasing, and the results are aliased patterns. Below is a patch with an upper graph that shows the control values that are sent to the frequency parameter of the carrier sine wave. 20.11::

Sample and Hold Patterns

( var gWin, audio, aNode, gbutton, xrate=0.1, yrate; yrate = 0.01; aNode = SynthDef("Alias", {arg rate = 5, freq = 0.25; Out.ar(0, SinOsc.ar( Latch.ar(SinOsc.ar(freq, mul: 200, add: 500), Impulse.ar(rate))) * [0.2, 0.2]) }).play; gWin = Window.new("Aliasing Illustration (rough draft)", Rect(20, 600, 1160, 350)).front; gWin.view.decorator = FlowLayout(gWin.view.bounds); x = MultiSliderView(gWin, Rect(10, 38, 1110, 150)).drawLines_(true) .value_(Array.fill(100, ({arg i; (i * xrate).sin * 0.5 + 0.5}))) .thumbSize_(5).gap_(6).valueThumbSize_(4).colors_(Color.red(0.7)) .background_(Color.new255( 92, 206, 255, 40 )) .action_{|v| v.value.round(0.01)}; y = MultiSliderView(gWin, Rect(10, 38, 1110, 150)).drawLines_(true) .value_(Array.fill(1000, ({arg i; (i * yrate).sin * 0.5 + 0.5}))) .background_(Color.new255( 209, 95, 238, 40)) .thumbSize_(0.1).valueThumbSize_(4).colors_(Color.blue(0.5)) .action_{|v| v.value.round(0.01)}; g = EZSlider(gWin, 1000@24, nil, [0, 4.0, \lin, 0.001, 0].asSpec) .action_({|ez| x.value_(Array.fill(100, ({arg i; (i * ((ez.value * 10 + 1) * xrate)).sin * 0.5 + 0.5}))); y.value_(Array.fill(1000, ({arg i; (i * ((ez.value * 10 + 1) * yrate)).sin * 0.5 + 0.5}))); aNode.set(1, (0.25 + (ez.value * 1.25))); }); gbutton = Button(gWin, 100 @ 24); gbutton.states = [["Stop Audio", Color.white, Color.red(0.7)], ["Start Audio", Color.black, Color.green(0.7)]]; gbutton.action = {|view| if (view.value == 1) {aNode.run(false)} {aNode.run(true)}}; ) ::

The patterns that emerge from a sample and hold are the product of the sample rate and frequency of the sampled wave. If either is set to a very gradual voltage control, the shifting 202

patterns seem improvised. They have a logic; predictable yet surprising. Figure 20.2 shows a sonogram of a latched saw wave (from patch 20.12). The first few patterns reflect the saw that is being sampled, but as the speed of the wave changes, the patterns change. They are complex, but logical.

Fig. 20-2 Sample and Hold (Latch) Patterns

Patch 20.12 contains latch variations. The first uses a SynthDef, allowing the creation of a window. The frequency of the sampled wave (and therefore the patterns) can be explored using the slider (or click the number box and use the up and down arrows). The next uses a Line, followed by a slowly changing LFNoise in the third, and a step noise in the fourth. The very last example combines a rhythmic pattern that adds to the sense of direction. It is accomplished by simply turning notes on or off. That is, rather than determining patterns by constructing a set of durations, a potential note is created at each pulse, then they actually fire, or don't fire according to some random (or sequenced) pattern. In this case, it is an LFNoise0 filtered through a max(0, LFNoise0) which returns 0 when LFNoise0 falls below 0. In short, and with fewer 0s, negative numbers don't "fire", positive values do. 20.12::

Sample and Hold Improvisation

( ControlSpec.specs.addAll([\latchlfo -> ControlSpec(0.1, 24, \lin, 0.1, 6, " latchlfo")]); SynthDef("Latch", { arg lowfreq=10, latchlfo = 0.1; var trig, env, out; trig = Impulse.kr(lowfreq); env = Linen.kr(trig, 0, 1, 1/lowfreq); out = SinOsc.ar(Latch.kr(LFSaw.kr(latchlfo, 0, [500, 250], 600), trig), mul: env); Out.ar(0, out) }).add; SynthDescLib.global[\Latch].makeWindow; ) ( { // Line control var env, trig, speed=10; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); 203

SinOsc.ar( Latch.kr( // spread out a little LFSaw.kr(Line.kr(0.1, 20, 60), 0, [500, 250], 600), trig), mul: env ) }.play ) ( { // Ramp noise control var env, trig, speed=10; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); SinOsc.ar(Latch.kr(LFSaw.kr(LFNoise1.kr(1/20, 10, 10.1), 0, [500, 250], 600), trig), mul: env) // SinOsc.ar(Latch.kr(LFSaw.kr(LFNoise0.kr(1/3, 10, 10.1).poll, 0, [500, 250], 600), trig), mul: env) }.play ) ( { // Choosen from a predetermined set: patterns will return. // Dial in the balance of repetition and variation by the // number of sets var env, trig, speed=10, set, setSize=5; set = Array.fill(setSize, {rrand(1.1, 2.7)}).round(0.001).post; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); SinOsc.ar(Latch.kr(LFSaw.kr( Demand.kr(Impulse.kr(1/5), 0, Drand(set, inf)), 0, 12, [60, 72]).midicps, trig), mul: env) }.play )

( { // controlling modulator and carrier freq (timbre) var env, trig, speed=10; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); f = Latch.kr(LFSaw.kr(MouseX.kr(1.1, 30), 0, 1000, 1100), trig); PMOsc.ar([f, f*3/2], f*2, 12, mul: env) }.play ) ( { // controlling index (filter) var env, trig, speed=10; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); i = Latch.kr(LFSaw.kr(MouseX.kr(1.1, 30), 0, 5, 5), trig); PMOsc.ar([300, 350], 356, i, mul: env) }.play 204

) ( { // controlling ratio (pitched timbre) var env, trig, speed=10; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); r = Latch.kr(LFSaw.kr(MouseX.kr(1.1, 30), 0, 2.0, 3.0), trig); PMOsc.ar(300, [300*r, 400*r], 12, mul: env) }.play ) ( { // and of course, scaled to control all three var env, trig, speed=10; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); c = Latch.kr(LFSaw.kr(MouseX.kr(1.1, 30), 0, 0.5, 0.5), trig); f = c*1000+200; o = PMOsc.ar(f, f*(c*3+4), c*5+6, mul: env); [o, CombL.ar(o, 2, 2)] }.play ) ( SynthDef("SH_Variations", { // out is included as an argument so you // can switch the output to other tracks if // in Logic (if using Soundflower). arg speed=6, out=0; var env, trig, set, setSize=10, mix; set = Array.fill(setSize, {rrand(1.1, 2.7)}).round(0.001).post; trig = Impulse.kr(speed); env = Linen.kr(trig, 0, 1, 1/speed); c = Latch.kr( LFSaw.kr(Demand.kr(Impulse.kr(1/5), 0, Drand(set, inf)), 0, 0.5, 0.5), trig); f = (c*48+24).round(1).midicps; // Tone down the ratio r = LFNoise1.kr(1/5, 2.0, 3.0).round(0.1); mix = PMOsc.ar(f, f*r, c*5+6 * env, mul: env); // add an on/off decision + amplitude variation mix = mix * max(0, TRand.kr(-1.0, 1.0, trig)); // Kanon // mix = [mix, CombL.ar(mix, 2, 2)] // Triple Kanon // mix = FreeVerb.ar(mix, LFNoise0.kr(1/3, 0.3, 0.3), 0.6); mix=Pan2.ar(mix, -1) + Pan2.ar(CombL.ar(mix, 2, 2), 0) + Pan2.ar(CombL.ar(mix, 4, 4), 1); 205

Out.ar(out, mix*0.3) }).add ) SynthDescLib.global[\SH_Variations].makeWindow ::

Another method for generating interest is to sample a more complex waveform. Even though complex, it is periodic, and patterns in the wave will be evident in the control values. In this example three waves are mixed together to create a more complex, but still periodic, wave. The first line of code shows four waves at 100, 200, 300, and 550 Hz, mixed to one wave. The second part places the mixed wave into a Latch. 20.13::

Complex Wave as Source

{Mix.ar(SinOsc.ar([100, 200, 300, 550], mul: 0.1))}.plot; ( //Used in a sample and hold { f = Latch.kr( Mix.ar(SinOsc.ar([100, 200, 300, 550], mul: 100, add: 110)), Impulse.kr(7)); e = EnvGen.kr(Env.perc(0, 0.2), Impulse.kr(7)); m = PMOsc.ar(f, f*1.25, e*5, mul: e*0.3); [m, CombL.ar(m, 2, 2)] }.play ) ::

In case you haven't guessed, I really like these systems. We should use them more. I can't think of any other method of organization that generates such interesting evolving patterns so quickly and elegantly. In the example below try an attack of 1/rate and a decay of 0. Uncomment the alternate latchrates, which move either gradually or stepwise through random values between 1.53 and 1.63. The patch is expanded to stereo with the [freq, freq * 1.5]. This creates a fifth in the right channel. Try other intervals such as 1.2, 1.7, 2, 1.25, etc. Also uncomment the alternative frequency, and ratios, or add your own. The key is to find that sweet spot where you can still follow the emerging patterns, yet have enough variety to keep your attention for long periods of time. Isn't that really what composition is all about; a blend of predictability and surprise? The second example does not use an S&H or sequencer (maybe you could add one?), but closes out this chapter by illustrating PM. It generates PM events with complex ratios, therefore inharmonic spectra (like a crotale) with a collection of envelopes (fast attack, long decay, and long attack, fast decay, which sounds reversed) for amplitude and modulation index. 206

The code that actually generates crotale events is a routine that uses random functions, but remember it gets values from a random number generator. Read further on to see how to reset the seed of the random number series, reproducing precisely a given sequence of values. The third task does this, and the occasional repetition gives a sense of cognizant improvisatory gestures. It sends "seeds" periodically, either chosen from the clock (70 percent of the time), or repeating the previous seed (30%). You can think of this task as a conductor, instructing the performer to either move on to something new (from the collection of billions of possible gestures) or repeating something you already did. More on this later. Try both examples at once. Adjust the overall volume of the first one. Comment out the FreeVerb lines and use Soundflower to send the output to Logic Pro and use it's more sophisticated spatial effects. This completes our discussion of typical electro-acoustic control sources. Later we will cover external data and probabilities, but I think of them as composition tools, not synthesis sources. Below is a review. •

Control Source Review

Periodic Wave Forms Aperiodic Wave Forms Sequencers Sample and Hold External data Probabilities 20.14::

Character Exact repetition, classic (cliché) Random, classic (cliché) Exact, but custom, complex repetition Evolving patterns, blend between variation/repetition Complex, meaning through context, natural patterns transfer Artificially intelligent

Examples Sine, Saw, Pulse LFNoise1, LFNoise0 Dseq Latch Data files Data tables, if()

Sample and Hold, PM

( // This example is not placed in a SynthDef because there are commented variations. // Once you settle on an interesting combination, create a SynthDef. { var freq, latchrate, index, ratio, env, speed = 9; speed = MouseX.kr(2, 20); latchrate = speed*1.61803399; // latchrate = speed*LFNoise1.kr(1/7, mul: 0.03, add: 1.6); // latchrate = speed*LFNoise0.kr(1/3, mul: 0.03, add: 1.6); index = Latch.kr( LFSaw.kr(latchrate, mul: 4, add: 8), Impulse.kr(speed) ); freq = Latch.kr( LFSaw.kr(latchrate, mul: 36, add: 60), Impulse.kr(speed) ).round(1).midicps; // freq = 200; //uncomment this line to hear just the index ratio = 2.01; // ratio = LFNoise1.kr(1, mul: 2.3, add: 3.0); // ratio = LFNoise0.kr(1, mul: 2.3, add: 3.0); 207

// ratio = LFNoise1.kr(1/5, mul: 2.0, add: 5.0); env = EnvGen.kr(Env.perc(0, 2/speed), gate: Impulse.kr(speed)); Out.ar(0, PMOsc.ar([freq, freq * 1.5], [freq*ratio, freq*1.5*ratio], index, mul: env*0.5) )}.play ) // Variation. I love this example. ( SynthDef("SH_PMv", {|speed = 9| var freq, latchrate, index, ratio, env, mix; speed = 9; latchrate = speed*LFNoise0.kr(1/10, mul: 0.03, add: 1.6); index = Latch.kr( LFSaw.kr(latchrate, mul: 5, add: 6), Impulse.kr(speed) ); freq = Latch.kr( LFSaw.kr(latchrate, mul: max(0, LFNoise1.kr(1/5, 24, 10)), add: LFNoise0.kr(1/7, 12, 60)), Impulse.kr(speed) ).round(1).midicps; ratio = LFNoise1.kr(1/10, mul: 2.0, add: 5.0); env = EnvGen.kr(Env.perc(0, LFNoise0.kr(speed, 1, 1.5)/speed), Impulse.kr(speed), LFNoise1.kr([5, 5], 2, 1).max(0).min(0.8)); mix = PMOsc.ar( [freq, freq * 1.5], freq*ratio, index, mul: env); Out.ar(0, FreeVerb2.ar(mix.at(0), mix.at(1), 0.2, 0.6)) }).add; SynthDescLib.global[\SH_PMv].makeWindow; )

// // // //

Crotale Run this code first to define the instrument, then execute the "task" below as many times as you want.

( SynthDef("crotale", { arg param = #[500, 3, 2, 0, 6, 5, 0, 0.9]; 208

var factor, env, out, freq, index, dur; var bus, ratioa, ratiob, attack, decay, panCont; freq = param.at(0); index = param.at(1); dur = param.at(2); bus = param.at(3); ratioa = param.at(4); ratiob = param.at(5); attack = param.at(6); decay = param.at(7); env = Env.perc(attack, decay); factor = gcd(ratioa, ratiob); ratioa = div(ratioa, factor); ratiob = div(ratiob, factor); panCont = (EnvGen.kr(env, timeScale: dur*1.1, levelBias: -1, levelScale: 2)) * (IRand(0, 1) * 2 - 1); // 0*2-1 = -1, 1*2-1 = 1 out = PMOsc.ar( ratioa*freq, //or try ratioa*freqCont, ratiob*freq, //or try ratioa*freqCont, pmindex: EnvGen.kr(env, timeScale: dur, levelBias: 1, levelScale: index), mul: EnvGen.kr(env, timeScale: dur, levelScale: 0.3)); out = Pan2.ar(out, panCont); out = out * EnvGen.kr(env, timeScale: 1.3*dur, levelScale: Rand(0.1, 0.5), doneAction:2); out = FreeVerb2.ar(out[0], out[1], 0.3, 0.5); Out.ar(0, out); //Out.ar(bus, out); }).add; ) // Define the instrument by running the code // above, then run this task to actually play // the instrument. ( r = Task({ var freq, indexDepth, indexRange, synthIndex, dur, repeat; var next, count, countDown, offset, ratioa, ratiob, envs, env; var range = 60, outBus = 0; count = 0; countDown = 0; offset = 0; envs = [[0, 0.9], [0.01, 0.9], [0.1, 0.8], [0.8, 0.01]]; repeat = Array.fill(10, {[rrand(range, range+24).round(1).midicps, 3, 2.1 - exprand(0.1, 2.0), 0, 1, 1, 0, 0.9]}); next = Array.fill(10, {[3, 0.75, 0.5, 0.25, 0.125].choose}); freq = rrand(range, range*2); // these two are just starting points indexDepth = 1; inf.do({ 209

if(countDown 10).and(countDown ControlSpec(60, 20000, \exp, 1, 440, " offset"), \fmul -> ControlSpec(1, 2000, \lin, 1, 100, " scale"), ]); ) ( SynthDef("BusEx1", { arg speed=1, fmul=500, fadd=2000, out = 0; var mix; mix = Pan2.ar( SinOsc.ar(LFSaw.kr(freq: speed, mul: fmul, add: fadd)), LFNoise1.kr(1)); mix = mix + CombL.ar(mix, 1.0, [0.4, 0.6], 1); Out.ar(out, FreeVerb.ar(mix, LFNoise1.kr(1)*0.5+0.5, 0.6)*0.5) }).add; SynthDescLib.global[\BusEx1].makeWindow; SynthDef("BusEx2", { arg speed=1/3, fmul=200, fadd=400, out = 2; var mix; mix = Pan2.ar( SinOsc.ar(SinOsc.kr(freq: speed, mul: fmul, add: fadd)), LFNoise1.kr(1)); mix = mix + CombL.ar(mix, 1.0, [0.4, 0.6], 1); Out.ar(out, FreeVerb.ar(mix, LFNoise1.kr(1)*0.5+0.5, 0.6)*0.5) }).add; SynthDescLib.global[\BusEx2].makeWindow; SynthDef("BusEx3", { arg speed=4, fmul=200, fadd=800, out = 3; var mix; mix = Pan2.ar( SinOsc.ar(LFPulse.kr(freq: speed, mul: fmul, add: fadd)), LFNoise1.kr(1)); mix = mix + CombL.ar(mix, 1.0, [0.4, 0.6], 1); Out.ar(out, FreeVerb.ar(mix, LFNoise1.kr(1)*0.5+0.5, 0.6)*0.5) }).add; 245

SynthDescLib.global[\BusEx3].makeWindow; ) ::

When working with Logic we learned that serial reverbs and delays are inefficient for two reasons: each reverb uses CPU, and they have to be controlled and modified independently. That is, if you wanted to change the depth of the virtual room, you will have to make the change in each patch (that is, provided you want them all to be in the same space). All of the previous examples have a delay and reverb built into the patch; a serial reverb, so to speak. The next example rewrites both patches without the reverb or the delay. Instead, both effects are defined as synths on their own with a bus input. For clarity I've declared the variables ~reverbBus and ~delayBus as 15 and 16. This simply allows me to use a name rather than a number for each bus, which is less confusing. Each of the instruments below are defined as synths, and opened using makeWindow. In this case the order of execution doesn't matter, because the window for each has a start button. The order in which you press the start buttons does matter. You have to start the reverb and delay first. Otherwise the other synths will not reach those destinations. The routing happens in the last few lines, the Out.ar statements. These are analogous to the main output and the auxiliary sends in Logic. The first Out.ar sends signal to bus 15 (~reverbBus), the second to 16 (~delayBus). The last routes the dry signal to out 0. For clarity, you might want to switch to the sclang and close all the previously created windows. 23.3::

Auxiliary Send Using Out and In

~reverbBus = 15; ~delayBus = 16; SynthDef("Delay", {arg lDelay = 0.3, rDelay = 0.5, audioBus = 0, amp = 0.6; Out.ar(audioBus, CombL.ar(In.ar(~reverbBus), 4.0, [lDelay, rDelay], 4)*amp); }).add; SynthDescLib.global[\Delay].makeWindow(w); SynthDef("Verb", {arg delay = 0.6, audioBus = 0, amp = 0.6; Out.ar(audioBus, FreeVerb.ar(In.ar(~delayBus), 1, delay)*amp); }).add; SynthDescLib.global[\Verb].makeWindow; SynthDef("Blip", { arg amp = 0.6, send1 = 0.5, send2 = 0.5; var mix; mix = Pan2.ar( SinOsc.ar(LFNoise0.kr(12, 300, 800), mul: LFPulse.kr(freq: 1/2, width: 0.1)), LFNoise1.kr(1)); Out.ar(~reverbBus, mix*send1); // reverb send Out.ar(~delayBus, mix*send2); // delay send Out.ar(0, mix * amp); // dry mix }).add; 246

SynthDescLib.global[\Blip].makeWindow; SynthDef("PMBlip", { arg amp = 0.6, send1 = 0.5, send2 = 0.5; var mix; mix = Pan2.ar( PMOsc.ar(300, LFNoise1.kr(1, 100, 300), 7, mul: LFPulse.kr(freq: 1/2, width: 0.1)), LFNoise1.kr(1)); Out.ar(~reverbBus, mix*send1); Out.ar(~delayBus, mix*send2); Out.ar(0, mix * amp); }).add; SynthDescLib.global[\PMBlip].makeWindow; ::

Once executed, you should see four windows. Again, they are on top of each other. The windows with amp and send1 sliders are the two synths. The one marked delay only is the reverb, the LDelay and RDelay is the delay. Start the delay and verb synths first, then launch the PMBlip and Blip. Adjust the controls to confirm that send1 is routed to the delay and send2 is routed to the reverb. Adjust the amplitude or delay times of each component. On the delay and reverb change the AudioBus to something other than 0. This will route the output to a track other than 1 and 2 in Logic. Of course, having routed these examples to Logic you have access to all the on board delays and reverbs, making this example a bit academic. •

Control Busses

… can be used to swap patching between different control sources. The next example first creates a synth with controls only. The first is an LFNoise sent to control bus 25, the next is a simple Saw sent to 26, then a Latch, similar to what we used with phase modulators, then a demand sequence, and finally a phase modulator. The second synth definition is the receiving synth. It is a playback buffer with a single control for speed (and direction). The control is linked to one of the busses 25 through 28 using In.kr. All of these are accessible from the newly created synth window because we have defined the arguments as control specs, or we have chosen arguments that already have control spec definitions. For that reason some may not make sense, for example the second argument of the PMOsc is "rate", but it's actually the modulator. The control specs for rate happen to be in the correct range. Run the first two lines to load a buffer, then play it just to make sure it loaded correctly. Run the final group, which will create two windows: one is the control, the other is the synth. There is only one slider on the control synth. It is used to change the rate of all of the controls. I've done this as a matter of convenience. In an actual composition you would probably want to create more elaborate arguments.

247

Select a different control source by changing the controlBus parameter. You can use the slider, but it is more accurate to just click on the number box and use the up and down arrows to switch busses. Note that you can also control the output buss of the buffer synth. You should see it appear in different tracks of logic. These examples are routed to output 0 and 1. If you are still using Logic, be sure you are using stereo tracks, or if mono tracks, pan track 1 hard left and track 2 hard right for the stereo effects. 23.4::

Control Busses

// Load an audio buffer ~playBuf = Buffer.loadDialog; // Check playback {PlayBuf.ar(1, ~playBuf.bufnum, 1, loop: 1)}.play ( // Create the controls synth, each routed to a different control bus SynthDef("Controls", {arg rate = 1; Out.kr(25, LFNoise0.kr(rate)); Out.kr(26, LFSaw.kr(rate)); Out.kr(27, Latch.kr(LFSaw.kr(25), Impulse.kr(rate))); Out.kr(28, Demand.kr(Impulse.kr(rate), 0, Dseq(Array.fill(12, {rand2(1.0)}), inf))); Out.kr(29, PMOsc.ar(60, rate, 7)) }).add; SynthDescLib.global[\Controls].makeWindow.bounds_(Rect(20, 400, 440, 120)).name_("Controls").front;

// Start the playback synth SynthDef("Buffer", { arg kbus25 = 25, out = 0, loop = 1; var mix, kontrol, trig; trig = Impulse.kr(loop); kontrol = In.kr(kbus25) * 2; mix = PlayBuf.ar(1, ~playBuf.bufnum, [kontrol, kontrol.neg], loop: loop); mix = mix * EnvGen.kr(Env.linen(0.1, 0.8, 0.1), trig, timeScale: loop); Out.ar(0, mix); }).add; SynthDescLib.global[\Buffer].makeWindow.bounds_(Rect(20, 600, 440, 120)).name_("Buffer").front; ) // Additional example using modeled0 bells. With a random trigger, you may have to wait a while. // Don't turn the volume up! // Start the Delays (destination) first. ( SynthDef("Delays", 248

{var delay1, delay2, delay3; delay1 = AllpassN.ar(In.ar(37, 2), 2.5, // input from 37 [LFNoise1.kr(2, 1.5, 1.6), LFNoise1.kr(2, 1.5, 1.6)], 3, mul: 0.8); // input from 38 delay2 = CombC.ar(In.ar(38, 2), 0.5, [Rand(0.2, 0.5), Rand(0.2, 0.5)], 3); // input from 39 delay3 = Mix.arFill(3, { CombL.ar(In.ar(39, 2), 1.0, LFNoise1.kr(Rand([0.1, 0.1], 0.3), 0.4, 0.5), 15) }); Out.ar(0, [delay1 + delay2 + delay3]); }).load; SynthDescLib.global[\Delays].makeWindow.bounds_(Rect(20, 600, 440, 120)).name_("Delays").front; ) // Then start as many bells as you like. // Use the sends to route each bell to a delay. ( SynthDef("Bells", {arg send1 = 0.6, send2 = 0, send3 = 0, rate = 0.3; var mix, delay, freq = 100; mix = SinOsc.ar(freq, mul: 0.5)*EnvGen.kr(Env.perc(0, 0.01), gate: Dust.kr(rate)); mix = Pan2.ar(Klank.ar(`[Array.fill(10, {Rand(500, 5000)}), Array.fill(10, {Rand(0.01, 0.1)}), Array.fill(10, {Rand(1.0, 6.0)})], mix), Rand(-1.0, 1.0)); Out.ar(0, mix*1); //send dry signal to main out Out.ar(37, mix*send1); //auxiliary send to bus 37 Out.ar(38, mix*send2); //to 38 Out.ar(39, mix*send3); //to 39 }).load; SynthDescLib.global[\Bells].makeWindow.name_("Bells: start after Delays!").front; ) ::

249

• Exercises 23:1 Continue to develop the last example by adding additional control sources. Create additional synths with output routed to control busses, then connect those controls using In.kr in the destination synth. 23:2 Combine the two examples by routing the output of the playBuff synth to auxiliary busses connected to the reverb and delay synths.

250

24.

On Random Numbers We used to use the I Ching, before computers.

•

Computers Cannot be Random

They only do what you tell them, nothing more, nothing less. With the power and sophistication of voice recognition and speech, it's hard not to characterize an unexpected and perplexing crash as random. But data is just 0s and 1s. It's very simple. Every single bit is one or the other. They only change if you change them. If something's wrong it's because someone gave bad instructions; 0s where 1s should be. Many times during a semester I say "the computer picks a random number." For nearly fifteen years no one objected. Then one smart aleck pointed out the paradox in my statement: "How," he asked, "can a machine do anything random?" It's true. A CPU can no more act randomly than a chair, or more aptly, a printing press. (Computers are, in a sense, a versatile printing press; more flexible, but an even simpler platform nonetheless.) The only way to get random numbers from a printing press would be to write down a long list of random numbers on paper, typeset them, and print them. Odd as that sounds, it's been done, by the RAND Corporation back in the 50s. The book is called A Million Random Digits and it contains, no kidding, a million random digits. Why? The forward explains: For cryptography, game playing, sociological surveys and various scientific calculations, people often need a series of numbers that are devoid of pattern. And that is a tall order. Generating true randomness is one of computer science's most difficult challenges. —George Johnson, "Connoisseurs of Chaos Offer a Valuable Product: Randomness," New York Times, June 12, 2001. The reviews at Amazon underline the absurdity of such a text. Some offer corrections, others give spoiler alerts, one is written entirely in digits, there are suggestions that it be translated into German or the original binary or made into an audio book, that the numbers would be easier to find if sorted, complaining that there are only 10 random digits (0 through 9), and my favorite dismisses all the previous reviewers, claiming they have missed the point; that the entire volume is allegorical. Entertaining, but serious, and essential for its time and yes, it is a 600-page hardbound, printed-on-real-paper book filled with random numbers, used by scientists to do random studies. (Before that we, and by we I mean the Chinese, and by before that I mean 3 million years BC by some accounts, used the I Ching or Book of Changes.) Fig. 24-1 A Million Random Digits (Excerpt)

I have both hard and soft copies of A Million Random Digits. Here are a few numbers from the first page: [23209, 02560, 15953, 34764, 35080, 33606, 99019, 02529, 09376, 70715, 38311, 31165, 88676, 74397, 04436, 27659, 12807, 99970, 80157, 36147, 64032, 36653, 98951, 82956] They do indeed seem random, and when graphed look very 251

similar to a noise wave (Figure 24.1). There is no apparent pattern. Random, maybe, but useful? Not any more. Once I've shown this series they must be discarded. If I ever use them again they become perfectly predictable. We learned this with our experiments on periodic waves. Any set of numbers becomes music when repeated. Repetition negates randomness. Figure 24.2 shows the same numbers repeated six times. The pattern is clear. The wave is periodic, and would produce pitch. Fig. 24-2 When Repeated, Not Random

The obvious solution: turn to a different page. But the conundrum continues. Those numbers can never be used in that order again. The truth is A Million Random Digits is not at all random. If I memorize the first number of page 431 I can predict it every time. I can predict what your copy of the book says at page 431. It's the same every time and every copy of the text is exactly the same. As one Amazon reviewer remarks, it's the opposite of random, it's exactly the same. To be truly random the text should show different numbers on the first page each time it's opened. The title should be A Million of the Same Digits but That Seem Random if You Don't Memorize Them. I'm sure they considered it, but it wouldn't fit on the spine. Randomness is elusive. Like subatomic particles, it evaporates once observed. The solution is a random method for choosing a page from the random number book. How about this: get two books and use the first to decide which page to use in the second. Or maybe riffle through the pages and have a colleague point. But then you would bias toward the middle of the text. Smart alecky student is still waiting for an answer. You could build a big rolodex machine that continuously riffles through all the pages without beginning or end and at the moment you need a random number turn it off and that's where you start. Less cumbersome, same results; use a stopwatch showing tenths of a second, resetting every minute. When you need to start a random process, hit lap and start on that page: 45.1" = page 451. A Million Random Digits is more than a curious side note. Each time you boot up a computer it creates a virtual rendition of A Million Random Digits using a … •

Random Number Generator

Like the text, it is always the same, every boot, every time it is invoked it produces the same series of numbers every time. Furthermore, computers with the same operating system use the same algorithm. In a very literal sense your computer generates a copy of a text containing 34,359,738,352 random digits when it starts up. (That's 34 billion.) Take that RAND.  It has the same problems. It's not random. If you start at the beginning of the sequence it will return precisely the same set of numbers. Yes, you can start in the middle, but if you pick the same spot every time the numbers will be the same. As with the text, it's the opposite of random. It is absolutely predictable—if you know the algorithm. To illustrate what I mean by "knowing the algorithm," imagine this method for shuffling: start with a fresh deck of cards, split in half and alternate exactly one from each pile in turn, repeat four times. The final shuffled deck will have a perfectly predictable order. The top five 252

cards will be a great poker hand: four aces and a king of spades. The only difference between a real shuffle and a carefully orchestrated one is a human's grasp of the process. If I were able to use this method without others knowing it, they would think the deck was randomized, when I would know it is not. What random really means is mixed up to the point where a particular human can't see the pattern nor predict the outcome. Randomness then is a matter of context and perception. We decide a number is random not if it corresponds with the numbers that came before or after but rather whether we can see any correspondence. To ensure the shuffle was not orchestrated fellow poker players cut the deck, just as we might rolodex through A Million Random Digits. This is called a … •

Random Seed

In Example 24.1 ten arrays are filled with random numbers. Just before they are filled, the seed is set to 0. This means the numbers are chosen from the very beginning of the random number generator. (The actual numbers are larger than 1000. The code 1000.rand performs a math operation to bring them below 1000.) In the second example (and all our examples to this point) there is no seed, and each array is filled with a different set of numbers. When no seed is provided, the OS picks one for you. It uses the clock, just like our example earlier. Each of the numbers from the clock are a sequence, thus predictable. But if used as a seed, they will point to a non-predictable number from the random number series. In practical application, you could say that each clock number is matched with a random number from the series. It's as if the random number series is spinning by along with the clock, like the riffling pages or cards. When you execute any random function it grabs a card from the constantly riffling deck. Below, just as an illustration, are sample clock values that would be matched to random number choices. Clock 22562.36 22562.37 22562.38 22562.39 22562.40 22562.41 22562.42 22562.43 22562.44 22562.45 22562.46 22562.47

Random number series 323285.46 701351.76 478870.03 970076.8 793066.74 608190.06 // number chosen when you execute code at 22562.41? 703109.74 229301.21 107420.56 178472.04 539228.68 511396.65

The third example creates a window that shows four number boxes. The first is the clock, refreshed every 100th of a second. Below that is the same clock, but showing the current value at the press of any key. To the left of the clock is the random numbers that match up to each of the clock positions if each were chosen as a seed. In reality the number generator isn't constantly spinning by, but in practicality it is, when matched to the clock seed. The last is the number that happened to match up to the clock when a new seed is chosen. I cheat a little with the code to keep it simple, but it gives an accurate visual illustration of the process. The last example chooses, sorts, then plots 10000 random choices. You can use this code to test distribution of biased random choices, covered next. 253

24.1::

Random Seed

// Set seed to 0 before picking 10 random numbers 10.do({thisThread.randSeed = 0; ({1000.rand} !10).postln }) 10.do({ ({1000.rand} !10).postln }) // no seed ( var rWin; rWin = Window.new("Random Numbers", Rect(20, 600, 400, 80)).front; rWin.view.decorator = FlowLayout(rWin.view.bounds); a = EZNumber(rWin, 150@20, "Clock ", [1.0, 10000000.0, \lin, 0.000001, 0].asSpec); b = EZNumber(rWin, 150@20, "#@Clock ", [1.0, 10000000.0, \lin, 0.000001, 0].asSpec); c = EZNumber(rWin, 150@20, "Seed ", [1.0, 10000000.0, \lin, 0.000001, 0].asSpec); d = EZNumber(rWin, 150@20, "#@Seed ", [1.0, 10000000.0, \lin, 0.000001, 0].asSpec); StaticText(rWin, 350@24).string = "Press any key to choose random number."; AppClock.sched(0.0, {|time| a.value_(time); b.value_(1000000.0.rand); 0.01}); rWin.view.keyDownAction = { d.value_(1000000.0.rand); rWin.view.background = Color.rand; c.value_(AppClock.sched(0.0, {|time| c.value_(time)})); } //AppClock.sched(0.0, {|time| c.value_(time); // d.value_(1000000.0.rand); // rWin.view.background = Color.rand; // rrand(2.0, 5.0)}) ) // Linear distribution ({10.0.rand} ! 10000).sort.plot // test distribution ::

The line thisThread.randSeed seeds on the client side (where you type and evaluate code). This means we can control the choice of random values for the initial construction of the patch, but once sent to the server, they cannot be changed or reset. RandSeed allows you to send a seed to the server, resetting in real time during playback. This is useful to manipulate events as they happen. In the examples below, the first starts an LFNoise control without any seed. Each time it's run you should get different values. The second seeds the random number then chooses a rate. That rate is sent to the server as part of the synth. It will always be the same. (It will be 254

8.9959564208984—How do I know? Because it's not really random.) But the values generated by the LFNoise, which is a random process on the server side, are different every time. The second example uses RandSeed, which sends a seed to the server each time it receives a trigger (down arrow). In that example the rate is different every time (because the client side seed has been removed), but the values generated by LFNoise are now not only the same each time you run it (103.724, 238.201, 317.462, 782.221, 113.28, 562.014.), but reset every time you press the down arrow (at the trigger). 24.2::

Client/Server Seed

{SinOsc.ar(LFNoise0.ar(rrand(2.0, 6.0).postln, 100, 500))}.play; ( { // Client random seed. Run several times to confirm rate is the same. var seed = 1768236, rate; thisThread.randSeed = seed; // comment out for clock seed "*****rate: ".post; rate = rrand(6.0, 10.0).postln; SinOsc.ar(LFNoise0.ar(rate, 100, 500).poll(rate, "freq")) }.play; ) ( { // Server var seed = 1768236, rate; "*****rate: ".post; rate = rrand(6.0, 10.0).postln; RandSeed.kr(KeyState.kr(125), 1956); // down arrow seeds server. Press several times. SinOsc.ar(LFNoise0.ar(rate, 400, 500).poll(rate, "freq")) }.play; ) ( { // Both: the same every time var seed = 1768236, rate; thisThread.randSeed = 1956; "*****rate: ".post; rate = rrand(6.0, 10.0).postln; RandSeed.kr(Impulse.kr(1), 1956); SinOsc.ar(LFNoise0.ar(rate, 400, 500).poll(rate, "freq")) }.play; ) ( { // Both: different every time var seed = 1768236, rate; thisThread.randSeed = Date.seed.postln; rate = rrand(6.0, 10.0).postln; SinOsc.ar(LFNoise0.ar(rate, 400, 500).poll(rate, "freq")) }.play; 255

) ( // Even something as complex as PinkNoise can be seeded // Change 4 to 10, 100. { RandSeed.kr(Impulse.kr(4), 1956); PinkNoise.ar; }.play ) ( { var trig; trig = Impulse.kr(4); SinOsc.ar( TIRand.kr(60, 72, trig).midicps, mul: EnvGen.kr(Env.perc(0.01, 1/8), Impulse.kr(4)) ) }.play ) a = {|rate = 1.5, num = 1916| RandSeed.kr(Impulse.kr(1), num);}.play a.set(\num, 1912); a.set(\num, 1883); a.set(\num, 1901); a.set(\num, 1937); a.set(\num, 1936); ::

The last example creates two processes; the random function, and the random seed. The sine wave is being driven by the TIRand which is reading values from the random number generator. The next lines sets up a RandSeed in a separate synth. Since we declared the seed (num) as an argument, it can be used with a.set to change periodically. The loop comes about from the trigger (Impulse). The number indicates where in the RNG the values are taken.  A rationalist might argue that nothing is random. Given enough information about any system the results are predictable. Card tricks work on the illusion that a shuffling has not been orchestrated when it has, and that the order of cards is not being managed and controlled when they are. Even though I can predict the frequencies generated by an LFNoise when seeded with 1956 on the server side of SC running on Mac OS 10.x, if you can't then it's random. We arrive at the improbable conclusion that the random number generator is completely random, if we want it to be, and absolutely predictable, when we need it to be—the best of both. Earlier we described aperiodicity as waves with no pattern. We now have to adjust that definition slightly. In SC, aperiodic waves are streams of numbers with patterns, but so complex that humans can't grasp their periodicity. {PinkNoise.ar([0.3, 0.3])}.play, essentially streams the random number generator through the speakers. It will repeat after a mere 108 hours. By comparison, A Million Random Digits, if converted to an aiff file, will only play for four seconds. Trust me, I've tried it. That's using the groupings of five for values in the range of 0 to 256

10000. But if you took each digit alone that's still just 20 seconds. A serial composition using random number generators (playing MIDI equipment, assuming 10 decisions per note and an average length of 1 second per event) will repeat after 153,351 years. But it will repeat.  This puts a rather existential spin on computer assisted generative composition. It is not random, since any performance can be duplicated with the same seed. It seems random because we haven’t heard that series of numbers before and even if we have, chances are we aren't going to remember it. But each seed (billons of them) represents a specific repeatable generation of the sequence described by the composition. So instead of a random process, it can be thought of as a billion possible variations, each identified by that seed. The code you write then is the DNA for billions of variations, of which one is chosen at each performance, depending on the exact time it is run. Would it strain the metaphor to imagine the clock spinning through not random seeds, thus random numbers, but complete performances, each played in a parallel universe, waiting to be chosen and experienced at the touch of the enter key. Probably.  How is all this information useful to a composer? The reasons for using the seed again include repetition during a performance, repeating a performance verbatim even though the design of the performance includes aleatoric events, or finally to consistently reproduce a crash for debugging purposes. Example 24.3 illustrates how to manage the best of both: Seed both the client and the server using the date or a random choice, but print that date-seed for reference. In essence; shuffle the deck to use for the random choices, but remember the exact sequence for later use. (This is a common feature in gaming software. You can retrieve a game "number" that will reproduce, for example, the precise ordering of a bridge shuffle, in case you want to play duplicate with yourself.) Using control ugens such as if this example could be expanded into a system that chooses it's own random server seeds for, say, 30 seconds, compiling them into an array, then sometimes choosing a repeated gesture, sometimes choosing a new gesture. 24.3::

Pick a Card, but Show Me

( // Start a process { var trig; thisThread.randSeed = Date.seed.postln; // Pick seed, but print for future use trig = Impulse.kr(rrand(4.0, 9.0)); SinOsc.ar( TIRand.kr(60, 72, trig).midicps, mul: EnvGen.kr(Env.perc(0.01, 1/8), trig) ) }.play ) // Run the first line several times, note the numbers posted. a = {|rate = 1.5, num| RandSeed.kr(Impulse.kr(1), 100000.rand.postln);}.play; a.set(\num, 33653); // repeat any of those by replacing 33653 with the posted number 257

::

To further dispel the myth of random events, not only is everything a computer does absolutely predictable, but even if we concede that a clock seeded random number generator is sufficient obfuscation to be unpredictable by me, it's still not random because of built in … •

Random Bias

 Though I can't predict the exact number, I can predict it will be a number. To be truly random, rand might occasionally produce a sperm whale and a bowl of petunias in mid air, both coming to terms with their identity. 1000.rand is biased in at least four ways: it returns a number, a positive number, a positive number between 0 and 999, and a positive integer between 0 and 999. These biases have come from the human who wrote the code for rand. In chapter 22 I proposed generating random frequencies using a deck of cards. Though at the time this system might have made sense, it was also very biased toward whole numbers and lower frequencies. This is because of the rule using three numbers per frequency and royalty and tens adding 1000. To produce a frequency above 1000, you need several royalty in a row. Odds are against this. Furthermore, if several royalty are in a row, you do get a higher pitch, but that removes those cards from the line up. While it would be possible to produce a frequency in the 14000 range (fourteen royalty grouped with three non-royalty), it is very improbable, and even if you did, that would mean only two other frequencies, maybe only one, above 1000. There is 0 chance for frequencies above 17k. But bias is normal. All music circumscribes choices. Describing bias amounts to teaching. For example, in common practice style the leading tone resolves up to tonic 90% of the time. It occasionally moves down a third in the tenor or alto, and it occasionally remains the same as a common tone to the seventh of the next chord. That is a bias that we teach first year music students. If your goal is singable, tonal, counterpoint, in common practice style then you follow that bias. In previous code we have used exprand, which biases choices toward lower numbers; [0, 2, 4, 7, 9].choose, which biases to only numbers in the array, rrand(2, 10) which biases to integers between 2 and 10, and so on. The next example shows a variety of bias choices. I use histo to track the random choices. The last method is wchoose, which picks from the array with weighted probabilities. The probabilities are given in the array as the first argument. 24.4::

Random Bias

// A histogram returns the total number of occurrences // of integers within a range. 0 0s, 4 1s, 2 2s, 0 3s, 1 4, 4 7s, etc. [1, 1, 1, 1, 2, 2, 4, 7, 7, 7, 7, 9, 9, 9, 9].histo(10, 0, 10) ({exprand(1.0, 100.0).round(1)}!20000).histo(100, 0, 100).plot ({[0, 20, 40, 50, 70, 90].choose}!100).histo(100, 0, 100).plot ({rrand(30.0, 66.0).round(1)}!20000).histo(100, 0, 100).plot // Bart's head ({min(100.rand, 100.rand).round(1)}!20000).histo(100, 0, 100).plot 258

({max(100.rand, 100.rand).round(1)}!20000).histo(100, 0, 100).plot ({div(100.rand+100.rand, 2).round(1)}!20000).histo(100, 0, 100).plot ({[0, 20, 40, 50, 70, 90] .wchoose([0.1, 0.1, 0.3, 0.3, 0.17, 0.03])}!100).histo(100, 0, 100).plot (Array.fill(20000, {if(0.4.coin, {20.rand * 2},{20.rand*2 + 1})})).histo(100, 0, 20).plot ::

•

Aleatory and Generative Composition

… provide the performer (in this case, the computer) with a general set of parameters and freedom of choice for others. All live music has varying degrees of generative elements. Many early works have no instructions for instrumentation, leaving it up to the performers. But even works that we think of as being well defined, that have been performed and refined over hundreds of years, even recorded, cured in our minds as definitive realizations can come to life with subtle differences in interpretation, perhaps a slightly faster tempo, more fluid tempo, wider dynamic interpretation (hey, sort of like physically modeled instruments). Improvisation, of course, is a defining characteristic of jazz. The instructions for performance may be as general as "let's jam." Even so, there is still an implied bias; the jam will probably be blues, tonal, traditional, 16 bars, trading solos. "What key," someone might ask. Specifying a key is a bias. A Day in the Life, Beatles, has a famous generative segment between two sections. At first Paul wanted the orchestra to play randomly. George Martin objected, so Paul refined the general instructions; have them start on a low note then slowly play to the top of their range.  The more general the instructions, the less control the composer has over the outcome. In this style of composition the focus moves from each event to groups of events, and often the entire work. Rather than obsess over the duration of this particular series of pitches, this phrase, this dynamic, the composer's vision is global; the overall level of dissonance, density of events, amount of overlap, tempo, or instrument change. And though you can always describe generally what the results will be, you can never predict any specific event at a specific time. Like physical models, generative scores set a process in motion. The biases described help maintain the character of the work, but the random choices embrace billions of possible variations. I'll balance my enthusiasm for aleatory with this: It can spawn indifference. I've seen many works where I feel the composer is just lazy, intent on imitating the carefully controlled styles popular a century ago (!) by writing a score with vague graphics because they know the performers will conform by playing pseudo random serialist/atonal "bee blonk" events. If asked to perform, I do my best to hang them with their own rope, and have been uninvited from several aleatoric performances because I realized the score in ways that most annoyed the composer while still within the bounds of the instructions. That is, in fact, an advantage to working with a computer (see chapter 25). It keeps you honest. It doesn't know what you "mean" when you're coping. It will do exactly what you say, so if you want bee blonk (carefully crafted distribution of elements in a total control style), you better understand bee blonk style. Computers may not be intentionally subversive, but the results are often different from what you expect. The next few examples create a random walk. Students often complain that it 259

doesn't sound very random. That's because sequences emerge that fall into our realm of experience. The pitch sequence C, D, E, C, (Frere Jacques) is a plausible result of a random walk. This set of pitches would sound ordered even to those who had never heard Frere Jacque because of it's tonal structure. But even a seemingly random set of pitches, say, C E D B B E F F , would be recognized by nearly anyone who has lived anywhere but under a rock in the last half of this century as the Twilight Zone theme, and therefore not random. (This raises yet another interesting conundrum; is random just lack of exposure?) Being random and sounding random are very different propositions. The pitch collection can't belong to any structure I know nor can it call on any work I know, even if the work was conceived using random, pseudo random, complexly ordered systems. Though the distinction is blurring, I divide computer music into two broad categories; the first half of this text covered synthesis or sound design. The second half covers the organization of events and structure design. Orchestrating events deals with pitch, duration, next event scheduling, instrument choice, and so on. For me at least, when working on this style of composition instrumentation is a secondary consideration because I am focused on pitch, duration, etc. (For the same reason, many of Bach’s works have no instrumentation, and even those that do are transcribed for other instruments. The essence of his music is in the notes the notes.) For that reason, these examples use … •

MIDI Out Once you design a process you simply send it to the MIDI synth.

MIDI is complicated by local variation. The example below shows how to connect MIDI to the IAC driver, which should be automatically routed to other software synths. I usually use SimpleSynth because it consistently works as expected. While Logic has wonderful instruments they don't respond to MIDI program changes or (as far as I can determine) events sent to different channels. Similar steps should work with Sibelius, virtual synths, and outboard synths. Each event in MIDI requires a note on and note off command (which creates a gate). The rest of the MIDI examples in this text assume you have initialized MIDI, assigned an out, and that it is called m. First review the MIDI out section and follow the instructions for enabling the IAC driver. Launch SimpleSynth, choose an instrument, then try the lines below. 24.5::

MIDI Init

( MIDIClient.init; m = MIDIOut(0, MIDIClient.destinations.at(0).uid); ) // Test m.noteOn(1, 60, 100); //channel, MIDI note, velocity (max 127) m.noteOff(1, 60); //channel, MIDI note, velocity (max 127) ::

Now we can try a simple … 260

•

Random Walk

Here is the seed example with a twist. First, when MIDI is used to generate notes the server never comes into play, so to speak, just the client which sends messages to a synth. For that reason we can seed on the client side. A Task is used as in a previous example, with an infinite do with a two second wait. Be sure to save your work before inviting an infinite do on board. It's a sure fire lock up if it isn't done correctly. (Sure, I could have used 100000.do, but this is, after all, an infinite monkey experiment.) Inside the first do is a second that plays 8 randomly chose notes between 60 and 72 with a 0.25 wait time. Each note is played with by sending a noteOn message to the IAC driver, which will play on anything that is connected to the driver. The duration is executed by scheduling the clock for a length of 0.25, which then sends the note off. Before playing the eight notes an if statement (read below) checks to see if a seed other than 0 was given. If not, the thread is seeded by the clock and that value is posted to the data window so you can use it again. If you hear something you recognize, hopefully the Twilight Zone theme, note the seed, replace 0 with it, then email it to me. Seriously. I'll also take anything from Gilligan's Island. I pronounce this my latest composition, in which you are a participant monkey; the missing twilight zone episode where Marius Constant dreams of infinite monkeys plunking at the keyboard. 24.6::

Constant Monkeys

// MIDIClient.init; m = MIDIOut(0, MIDIClient.destinations.at(0).uid); ( r = Task({ inf.do({ var seed; seed = 0; // replace with a seed once you recognize a phrase if(seed == 0, // Read about "if" below { "Do you recognize this? -- ".post; thisThread.randSeed = Date.seed.postln;}, { thisThread.randSeed = seed;}); 8.do({ var note; note = rrand(60, 72); m.noteOn(1, note, 100); thisThread.clock.sched(0.25, {m.noteOff(1, note); nil}); 0.25.wait }); 2.wait; }); }); r.start; ) // Stop all and send MIDI note off on all channels r.stop; 127.do({arg i; m.noteOff(1, i, 0)})

261

::

These examples do not respond to ⌘-period to stop playback. Rather, run the last line, which stops the routine and sends noteOff commands to all channels. This next example expands our walk to include octave, amplitude, and duration. The dur value determines both the length of the current event, and the time of next event. 24.7::

Random Walk

// MIDIClient.init; m = MIDIOut(0, MIDIClient.destinations.at(0).uid); ( r = Task({ 50.do({ var note, dur, oct, amp; note = 12.rand; oct = (4.rand + 4)*12; note = note + oct; amp = rrand(50, 100); dur = 8.rand * 0.125; m.noteOn(1, note, amp); thisThread.clock.sched(dur, {m.noteOff(1, note); nil}); (dur).wait }); }); r.start; ) r.stop; 127.do({arg i; m.noteOff(1, i, 0)}) ::

Though any trained musician would characterize this as "random" it is biased in many ways. Pitches are biased to discrete equal tempered values between 48 and 84. Try 48.0 for microtonal pitches. Durations are multiples of 0.125, resulting in a pulse. Try rrand(0.125, 2.0) for a free meter. Each event is sequential without overlap. This is because the duration of the current event and the time until next event are the same. Read further for variations. This model is expanded to include instrument choice and articulation next. Articulation essentially treats duration and next event differently allowing events to overlap or die away before the next is played. For instrumentation we can use both program changes or several channels with different instruments assigned. SimpleSynth uses the Apple DLS sound set, which conforms (with slight naming variation) to the general MIDI sound set. Here is a list. This really should have been two chapters. Sorry.

262

Keyboards 0 Acoustic Grand Piano 1 Bright Acoustic Piano 2 Electric Grand Piano 3 Honky-tonk Piano 4 Rhodes Piano 5 Chorused Piano 6 Harpsichord 7 Clavinet Tuned Percussion 8 Celesta 9 Glockenspiel 10 Music Box 11 Vibraphone 12 Marimba 13 Xylophone 14 Tubular Bells 15 Dulcimer Organs 16 Hammond Organ 17 Percussive Organ 18 Rock Organ 19 Church Organ 20 Reed Organ 21 Accordion 22 Harmonica 23 Tango Accordion Guitars 24 Acoustic Guitar (nylon) 25 Acoustic Guitar (steel) 26 Electric Guitar (jazz) 27 Electric Guitar (clean) 28 Electric Guitar (muted) 29 Overdriven Guitar 30 Distortion Guitar 31 Guitar Harmonics

Basses 32 Acoustic Bass 33 Electric Bass (finger) 34 Electric Bass (pick) 35 Fretless Bass 36 Slap Bass 1 37 Slap Bass 2 38 Synth Bass 1 39 Synth Bass 2 Strings 40 Violin 41 Viola 42 Cello 43 Contrabass 44 Tremolo Strings 45 Pizzicato Strings 46 Orchestral Harp 47 Timpani Strings 2 48 String Ensemble 1 49 String Ensemble 2 50 SynthStrings 1 51 SynthStrings 2 52 Choir Aahs 53 Voice Oohs 54 Synth Voice 55 Orchestra Hit Brass 56 Trumpet 57 Trombone 58 Tuba 59 Muted Trumpet 60 French Horn 61 Brass Section 62 Synth Brass 1 63 Synth Brass 2

Reeds 64 Soprano Sax 65 Alto Sax 66 Tenor Sax 67 Baritone Sax 68 Oboe 69 English Horn 70 Bassoon 71 Clarinet 72 Piccolo Winds 73 Flute 74 Recorder 75 Pan Flute 76 Bottle Blow 77 Shakuhachi 78 Whistle 79 Ocarina Synth Leads 80 Lead 1 (square) 81 Lead 2 (sawtooth) 82 Lead 3 (calliope lead) 83 Lead 4 (chiff lead) 84 Lead 5 (charang) 85 Lead 6 (voice) 86 Lead 7 (fifths) 87 Lead 8 (bass + lead) Synth Pads 88 Pad 1 (new age) 89 Pad 2 (warm) 90 Pad 3 (polysynth) 91 Pad 4 (choir) 92 Pad 5 (bowed) 93 Pad 6 (metallic) 94 Pad 7 (halo) 95 Pad 8 (sweep)

Synth Effects 96 FX 1 (rain) 97 FX 2 (soundtrack) 98 FX 3 (crystal) 99 FX 4 (atmosphere) 100 FX 5 (brightness) 101 FX 6 (goblins) 102 FX 7 (echoes) 103 FX 8 (sci-fi) Folk Instruments 104 Sitar 105 Banjo 106 Shamisen 107 Koto 108 Kalimba 109 Bagpipe 110 Fiddle 111 Shanai Percussion 112 Tinkle Bell 113 Agogo 114 Steel Drums 115 Woodblock 116 Taiko Drum 117 Melodic Tom 118 Synth Drum 119 Reverse Cymbal 120 Guitar Fret Noise Sound FX 121 Breath Noise 122 Seashore 123 Bird Tweet 124 Telephone Ring 125 Helicopter 126 Applause 127 Gunshot

• Total Randomization … randomizes all musical parameters: duration, next event, instrument, pitch, articulation, amplitude, etc. 24.8::

Random Walk

MIDIClient.init; m = MIDIOut(0, MIDIClient.destinations.at(0).uid); // Set synth channels to different programs m.program(0, 6); m.program(1, 12); m.program(2, 8); m.program(3, 24); // Pan them, 0 is left, 127 is right. m.control(0, 10, 0); m.control(1, 10, 42); m.control(2, 10, 84); m.control(3, 10, 127)

( ~note = {rrand(36, 84)}; ~dur = {rrand(0.125, 1.0)}; 263

~inst = {rrand(0, 3)}; ~amp = {rrand(33, 120)}; ~nextE = {[0.125, 0.25, 0.325, 0.5, 0, 0].choose}; r = Task({ "I\tPC\t\tAmp\t\tDur\t\tNxt\t\t".postln; 100.do({ var note, inst; inst = ~inst.value.post; "\t".post; note = ~note.value.post; "\t\t".post; m.noteOn(inst, note, ~amp.value.post); " \t\t".post; thisThread.clock.sched(~dur.value.round(0.1).post, {m.noteOff(inst, note); nil}); " \t\t".post; (~nextE.value.postln).wait }); }); r.start; ) // Execute these during playback to swap functions. ~note = {[0, 2, 4, 7, 9].choose + [36, 48, 60, 72].choose}; ~note = {[0, 2, 4, 5, 7, 9, 11].choose + [48, 60, 72].choose}; ~note = {(0, 2..36).choose + 48}; // 0.125 is the smallest division of the tempo ~dur = [1, 2, 3, 4, 5, 6].choose * 0.125 ~amp = 120; // constant amplitude ~inst = 0; // single instrument // Rhythms come about through "play" (50, 70, 100) // or "don't play" (0) choice. ~dur = 0.125; amp = {[0, 50, 70, 100].choose}; r.stop; 4.do({|e| 127.do({|i| m.noteOff(e, i, 0)})}) ::

You may not find these examples very interesting because they are not. Randomness rarely is. (Bee blonk is interesting because it is not random. It's carefully crafted and balanced.) Even in the natural environment randomness is rare. I can listen to birds, crickets, rain, and wind much longer than anything by Wagner, and I stare at the branches of trees long enough to concern neighbors, but none of those are random. They are complex—the results of biased generative processes. Complex, biased, chaotic processes are beautiful. Another way to bias a choice is to create a condition. For example, if there are already 3 notes in the chord don't pick another one, if we are about halfway through the work increase the density, or if the random seed is 0, generate a seed using the clock and print it to the data window. The if message is an example of …

264

•

Conditional Control

 The if function takes three arguments: an expression to be evaluated, a function to be executed if the expression is true, and a function for false. It evaluates the expression, returns the results of the first function if true, the second if false. if(expression, {true function}, {false function}) The true or false often results form a comparison of two values separated by operators such as < or > for greater/lesser than, = for greater/lesser than or equal to, != not equal to (In computer parlance the exclamation point, or bang means not. The inside joke is that it lends a more accurate representation of advertising, such as "The must see movie of the year!"), and == for equals. (Note the difference between =, which means store this number in the variable and == which means "is it equal to?") The message or combines two statements, returning true if either are correct; or(a > 20, b > 100). The message and combines two statements, returning true only if both are correct; and(a > 20, b > 100). The word true is true and false is false. 24.9::

Control Using If

( Task({ 40.do({ |a| if(a%10==9, {"ten repetitions".postln}, {a.postln}); 0.3.wait; }) }).play; ) if((1 == 1).and(5 < 7), {"both are true"},{"maybe only one is true";}) if((1 == 20).and(5 < 7), {"both are true";},{"one or both are false";}) if((1 == 20).and(24 < 7), {"both are true";},{"one or both are false";}) if((1 == 4).or(true), {"true is always true";},{"1 does not equal 4";}) if(false.or(true), {"true is always true";},{"true wins with or";}) if(false.and(true), {"true is always true";},{"but false wins with and";}) // Can you predict these two before running? if(or(10 > 0, 10 < 0), {34},{78}) if((1 == 1).and((10 > 0).or((5 < 0).or(100 < 200))), {78},{88}) ::

The if function is usually used in combination with some iterative process such as do. Example 24.10 first shows how it can be used to manage the range of MIDI interval choices. It begins with 60, then adds each new choice to m and returns that new value. I've biased the 265

choices so that there are more intervals up than down. Eventually the MIDI values exceed a reasonable range for most instruments. Even if I carefully balanced the choices it is conceivable that a positive value is chosen 20 times in a row. (There are better schemes for managing interval choice. For example, you could always choose positive values then reduce that result to a single octave using modulo, so that all values are between 0 to 12, then choose the octave separately.) The if statement below checks the value during each iteration and reduces it by two octaves if it exceeds 72 and increases it by two octaves if below 48. 24.10::

Range Control with If

( Task({ m = 60; 50.do( { var next; next = [6, 17, 14, 2, 11, 8, -12, -16, -1, -3].choose; "next interval is : ".post; next.postln; m = m + next; "before being fixed: ".post; m.post; if(m > 72, {m = m - 24}); if(m < 48, {m = m + 24}); // Same thing // m = m.wrap(48, 72); " after being fixed: ".post; m.postln; 1.wait; }; ) }).play ) ::

It is worth poking around in SC to see if a function already exists that will do what you want. There is wrap, which wraps a value around, but in this example we are buffering, not wrapping. So in this case we really do need the extra lines of code. The situations where I most use if, do, while, is parsing data files (chapter 27). Below shows a do iteration over an array of pitch classes with an if test to look for C, E, or G. The computer doesn't understand these as actual pitches (see the discussion on strings below), but just text. Even so, it does know how to compare to see if they are equal. The second example uses a series of if statements to define a region on the screen. When the mouse enters the region with an x greater than 0.3, but less than 0.5, and a y that is greater than 0.3, but less than 0.7 (all conditions are "true") it generates a positive value, or a trigger. In this example the * is equivalent to and. There are no true or false functions, just the values 1 (on) and 0 (off). The last example controls when to postln rather than post. This formats the output more concisely. 266

24.11::

More If

( ["C", "C#", "D", "Eb", "E", "F", "F#", "G", "Ab", "A", "Bb", "B"].do( {arg item, count; if((item == "C").or(item == "E").or(item == "G"), //Boolean test {item.post; " is part of a C chord.".postln;}, //True function {item.post; " is not part of a C chord".postln;} //False function ) } ) ) ( { // Use Mouse to trigger event var aenv, fenv, mgate, mx, my; mx = MouseX.kr(0, 1); my = MouseY.kr(0, 1); mgate = if((mx>0.3) * (mx0.3) * (my= 0, {(ea + k.wrapAt(ct)%25 + 65).asAscii.post}, {(ea + 65).asAscii.post})}); "" d = "XMWI D WLNTOF UPXDUMRO, FEFI OFXWFV PJ ULH EOQLDCIW MV VRUEWFH CD B JJAHO DNSXOX. E SSW 1 GSU IBBQSMI, BLMP FRFPHH XKF OFXWFVV EED DT EDH. E SSW2 AJPO VHUYUO FEI. XKFWH EUF HBWC XR GUBGN YVJRJ THSGHOXDHIV FHDEXTI FEFI OFXWFV XMOM KBAH XKF VBQH WXCWWJXXUMRO. B YFKHOIUF FJTKFV JW B WATH SI VRUEWJSQ, EVX FEFI QFB MIWUIU MV VRUEWFH B GJJIFVHOX BQRVRW, DDGRSHLOK US B NFD XSUE. JJ ULH OHA APVG MV FHE, ULHO WII GMUTX MIWUIU BLMP CI FRFPHHE AJXK VRU 1 (E), ULH RHYX JW SSW 4 (I), WIIQ VRU 3 (G), BRG WR SQ. CPY DEQ UXJGNMD DLHDO B YJKHOIUF EA HYEPJRLOK BPO PHUXHS D'W JR ULH SUJKLOEO. WIIC BLMP DSUSIVQSQE WP OFXWFVV MQ XKF NFDAPVG. DBEDBEDBEDB AJPO IQDSGF DT EFHEFHEFHEFH, MQ WRNI WEUJEWJSQ".ascii - 65; // To decode, simply subtract d.do({|ea, ct| if(ea >= 0, {(ea - k.wrapAt(ct)%25 + 65).asAscii.post}, {(ea + 65).asAscii.post})}); ""

303

// Gives the illusion of "thinking" about it. ( r=Task({ d.do({|ea| (ea + 65).asAscii.post; 0.01.wait}); "\n\nTranslation:\n\n".post; d.do({|ea, ct| if(ea >= 0, {(ea - k.wrapAt(ct)%25 + 65).asAscii.post}, {(ea + 65).asAscii.post}); 0.03.wait}) }).start; ) r.stop; ::

As a final example of working with files and characters, here is what I use to extract all the code examples from this document, after saving as a text file. 26.13::

Extracting Code

( var input = 1, fpr, count = 0, lnct = 0, nm; nm = Date.getDate.bootSeconds; fpr = File("/Users/blah/Desktop/ SCDoc.txt", "r"); ~fp = File("/Users/blah/Desktop/" ++ nm ++ ".scd", "w"); // read and discard most of the beginning of the document while({lnct < 500}, {fpr.getLine.postln; lnct = lnct + 1;}); while({input.notNil}, { input = fpr.getLine; lnct = lnct + 1; // [input, input.isNil].postln; if(input.asString.find("----") != nil, {count = count + 1; ~fp.write("\n" ++ lnct ++ "\n")}); // count turns the write process on and off if(count%2 == 1, {~fp.write(input)}); }); ~fp.close; ) ::

304

• Exercises 26:1 Realize two of the Kinderstuck variations. 26:2 Find a sound effect for a single typewriter key press. Add playback of this effect to the last example of 26.12. 26:3 Write your own Kinderstuck variation.

305

27.

Maps, Masks, and Markovs Probability tables, external sources.

•

Music from Non-Musical Ideas

Nearly all music is influenced by some extra-musical criteria. In traditional styles it may be a text, a painting, geometric patterns, or philosophical ideas. In the 20th Century composers have expanded this influence to include natural data such as electrical impulses from plants, position of stars, a mountain silhouette, or rain and snow patterns. Given the freedom of styles in modern music (not bound, for example, with the rules of tonality), the connection can be and often is much more transparent. These data sources could include human structures such as other music, text, bus timetables, or masked input such as an audience whose motion through a performance space is being mapped to pitch, amplitude, tone, etc. The appeal of external sources is that the patterns and structure of the subject can be transferred from the extra-musical source to the composition. A star map, for example, would provide x and y locations that would generate a fairly even distribution of events, while a mountain silhouette would provide a smooth contiguous stream of events. There is also a philosophical connection that can bring additional meaning to the composition, e.g. pitch choices taken from the voice print or favorite poem of one of the performers. With text, musicians may paint the character of a single word, but also generate the exact replica of the word in musical symbols, as in Bach's double fugue which begins with the four pitches Bb, A, C, B (in German musicianship: B, A, C, H). Any data stream, whether read from an existing file or from a real time input source such as a mouse or video link, can be used in influencing musical parameters. SC provides a number of outside control sources such as MouseX and Y, Wacom pad controls, MIDI input, a data file, or Open Sound Control (which can be linked to iPhones, iPads, etc.). Text has many advantages, and is a good place to start. The ASCII values for 'a' through 'z' are 97 to 122, and 'A' through 'Z' are 65 through 90. This conveniently corresponds with MIDI values, though a bit high. They could easily be offset to a lower range, but… •

Mapping

… allows greater control over the correlation between notes and letters therefor better control over the nature of the composition while retaining the link to the patterns in the stream. In text, for example, we know that vowels occur about every other letter. Sometimes there are two in a row. They are separated by as many as four consonants. With this understanding we can map key pitches to vowels and achieve (for example) a sense of tonality; a = C4, e = E4, i = G4, o = C5 (a = 60, e = 64, i = 67, o = 72), etc. A map also allows you to confine data structures to characters in use, omitting (or including) characters such as spaces, punctuation, etc. In example 27.1 the variable ~pm is assigned the IdentityDictionary array. It contains pairs of elements; the original character and its associated value. The association is made with dash and greater than sign, such as $b -> 4, which reads "make the character b equal to the integer 4." The numbers associated with the characters are MIDI intervals or pitch classes. 306

27.1::

Pitch Map

~pm = IdentityDictionary[ $H -> 6, $x -> 6, $b -> 6, $T -> 6, $W -> 6, $e -> 11, $o -> 11, $c -> 11, $, -> 11, $. -> 11, $n -> 3, $y -> 3, $m -> 4, $p -> 8, $l -> 9 ]; ::

After using the IdentityDictionary for a few projects settled on a computationally less efficient, but more logical algorithm using a two dimensional array and a search. The first item in the enclosed arrays is the list of characters—a string—the second element is the value to be associated with all items in that string. The obvious advantage over a dictionary is that a group of letters can be mapped to one value more concisely. They are parsed using a do function which looks at each of the first elements and if a match is found (using includes) the mappedValue variable is returned. 27.2::

Mapping Array

~intervMap = [ ["ae", 2], ["io", 4], [" pstc", 5], ["Hrn", -2], ["Sxm", -1], ["lfg", -4], ["Th", -5], [".bdvuw", 1] ]; "Sample text, which can be mapped to intervals".do({|char| ~intervMap.do({|each| if(each.at(0).includes(char), {char.post; ": ".post; each.at(1).post; " ".post}) }) }); "" ::

Here is a patch which controls only pitched elements in an absolute relationship (that is, pitch classes rather than intervals). 27.3::

Extra-Musical Criteria, Pitch Only

MIDIClient.init; m = MIDIOut(0, MIDIClient.destinations.at(0).uid);

// Test m.noteOn(1, 60, 100); m.noteOff(1, 60); ( var noteFunc, blipInst, midiInst, channel = 0, port = 0, prog = 0, intervMap, count = 0, ifNilInt = 0, midin = 0, inputString; //The input stream. inputString = "Here is an example of mapping. The, them, there, these," 307

"there, then, that, should have similar musical interpretations." "Exact repetition; thatthatthatthatthatthat will also" "be similar."; //intervMap is filled with arrays containing a collection of //characters and a value. In the functions below the character //strings are associated with the numbers. intervMap = [ ["ae", 2], ["io", 4], [" pst", 5], ["Hrn", 7], ["xmp", 1], ["lfg", 3], ["Th", 6], [".bdvu", 11] ];

"// [Char, Interval, ifNilInt, midi interval, octave, midi]".postln; noteFunc = Pfunc({var parseInt, octave; //Each array in the intervMap is checked to see if the //character (inputString.wrapAt(count)) is included. If //it is then parseInt is set to the value at item.at(1) intervMap.do({arg item; if(item.at(0).includes(inputString.wrapAt(count)), {parseInt = item.at(1)}) }); //If parseInt is notNil, midin is set to that. //ifNilInt is for storing each parseInt to be used if //no match is found and parseInt is nil the next time around. if(parseInt.notNil, {midin = parseInt; ifNilInt = parseInt}, {midin = ifNilInt} ); octave = 60; "//".post; [inputString.wrapAt(count), parseInt, ifNilInt, midin, octave/12, midin + octave].postln; count = count + 1; midin + octave }); Pbind( \type, \midi, \midiout, m, // must provide the MIDI target here \midinote, noteFunc, \dur, 0.125, 308

\amp, Prand([0.4, 0.6, 0.8], 180); ).play ) m.allNotesOff(1); m.allNotesOff(0); ::

To achieve a particular balance of interval choices the intervMap should contain proportional interval values rather than absolute pitches. The numbers may look pretty much the same, but in the actual parsing use midin = parseInt + midin%12 (thus parseInt becomes an interval) rather than midin = parseInt. Interest can also be added by mapping pitches across several octaves, or mapping octave choices to characters. Notice in this next example the series "thatthatthatthat" results in an intervallic motive, rather than repeated pitches. 27.4::

Text Map, Intervals

( var noteFunc, blipInst, midiInst, channel = 0, port = 0, prog = 0, intervMap, count = 0, ifNilInt = 0, midin = 0, ifNilDur = 1, durMap, durFunc, ifNilSus = 1, susMap, susFunc, ifNilAmp = 0.5, curAmp = 0.5, ampMap, ampFunc, inputString; //The input stream. inputString = "Here is an example of mapping. The, them, there, these," "there, then, that, should have similar musical characters." "Exact repetition; that that that that that that will also" "be similar."; //intervMap is filled with arrays containing a collection of //characters and a value. In the functions below the character //strings are associated with the numbers. intervMap = [ ["ae", 6], ["io", 9], [" pst", 1], ["Hrn", -3], ["xmp", -1], ["lfg", -4], ["Th", -5], [".bdvu", 1] ]; durMap = [ ["aeiouHhrsnx", 0.125], ["mplf", 1.5], ["g.T,t", 0.25], ["dvc", 2], [" ", 0] ]; susMap = [ ["aei ", 1.0], ["ouHh", 2.0], ["rsnx", 0.5], ["mplf", 2.0], ["g.T,t", 4.0], ["dvc", 1.0] ]; ampMap = [ 309

["aeHhrsnx ", 0.8], ["ioumplfg.T,tdvc", 1.25] ]; noteFunc = Pfunc({var parseInt, octave; //Each array in the intervMap is checked to see if the //character (inputString.wrapAt(count)) is included. If //it is then parseInt is set to the value at item.at(1) intervMap.do({arg item; if(item.at(0).includes(inputString.wrapAt(count)), {parseInt = item.at(1)}) }); //If parseInt is notNil, midin is set to that plus previous //midin. ifNilInt is for storing each parseInt to be used if //no match is found and parseInt is nil. if(parseInt.notNil, {midin = parseInt + midin%12; ifNilInt = parseInt}, {midin = ifNilInt + midin%12} ); octave = [24, 36, 48, 60, 72].choose; [inputString.wrapAt(count)].post; ["pitch", parseInt, midin, octave/12, midin + octave].post; midin + octave }); durFunc = Pfunc({var parseDur, nextDur; durMap.do({arg item; if(item.at(0).includes(inputString.wrapAt(count)), {parseDur = item.at(1)}) }); if(parseDur.notNil, {nextDur = parseDur; ifNilDur = parseDur}, {nextDur = ifNilDur} ); ["dur", nextDur].post; nextDur }); susFunc = Pfunc({var parseSus, nextSus; susMap.do({arg item; if(item.at(0).includes(inputString.wrapAt(count)), {parseSus = item.at(1)}) }); 310

if(parseSus.notNil, {nextSus = parseSus; ifNilSus = parseSus}, {nextSus = ifNilSus} ); ["sustain", nextSus.round(0.01)].post; nextSus }); ampFunc = Pfunc({var parseAmp; ampMap.do({arg item; if(item.at(0).includes(inputString.wrapAt(count)), {parseAmp = item.at(1)}) }); if(parseAmp.notNil, {curAmp = curAmp*parseAmp; ifNilAmp = parseAmp}, {curAmp = curAmp*ifNilAmp} ); count = count + 1; if(0.5.coin, {curAmp = rrand(0.2, 0.9)}); ["amp", curAmp.round(0.01)].postln; curAmp.wrap(0.4, 0.9) });

a = Pbind( \type, \midi, \midiout, m, // must provide the MIDI target here \midinote, noteFunc, \dur, durFunc, \legato, susFunc, \amp, ampFunc ).play ) a.stop; m.allNotesOff(1); m.allNotesOff(0); ::

Artificial Intelligence begins by describing human phenomena in a language the computer understands: numbers, probabilities, and formulae. Restricting choice in any way teaches bias because the CPU makes an informed choice. The random walk in Chapter 24 was biased in many ways. A map represents a set of ratios and probabilities. In a simple biased random choice there is only one level of probability; a probability for each possible choice in the set. This is known as a …

311

•

Markov Process

… with a zeroth order probability. Building a probability table is as simple as counting all pitches in a work and dividing by the total number in the work. It would take barely an afternoon to download Bach's entire opus from a MIDI site, read and parse each note (transposed to C) and calculate the probabilities that, say, C will appear at any given time. (Any theory student might take a stab at it without going to all that bother, say, 14%.) Writing a biased random walk with those probabilities would not recreate Bach because we perceive melody and musical progression in the relationship of the current pitch to the next pitch, and the last three pitches, and the pitches we heard a minute ago. To describe a melody, you have to describe the connections between pitches. For this we use a higher order Markov chain; first or second at least. A first order chain describes the connection between two notes. Given the current note, what are the probabilities for the next note?  For example, if the current pitch is G (in the key of C), what are the probabilities that the next pitch is C? Quite high (resulting in a V-I relationship) in tonal systems, higher than say F would follow G (a retrogressive V-IV). If the current pitch is F on the other hand, then there is a greater chance that the next pitch is E (resolution of the IV) than C (retrogressive). In non-tonal music you might likewise describe relationships by avoiding the connection G to C. If the current pitch is G and avoiding a tonal relationship is your goal, the probability that the next pitch is Dsharp, or A-flat would be higher than C. To illustrate, consider the first four bars of Frere Jacque. They contain only five pitches, so a chart listing only C through G is required. Current pitches are in the first column and next probabilities are in each subsequent column. C D E F G

C 0 0 0 0 0

D 0 0 0 0 0

E 0 0 0 0 0

F 0 0 0 0 0

G 0 0 0 0 0

To calculate each probability, count up all the times C is followed by D and enter that number (2) in the D column of the C row. Next we examine all other Cs and count the number of times C is followed by C, E, F, G, and enter each of those totals into the table. We do the same for the pitch D, then E, and so on. This is the resulting transition table: C D E F G

C 1 0 2 0 0

D 2 0 0 0 0

E 1 2 0 0 1

F 0 0 2 0 0

G 0 0 0 2 0

Total 4 2 4 2 1

The final probability is the number of actual occurrences in that column divided by the total possible outcomes in the final row. The C column row will be 1/4, 2/4, and 1/4. C D E F G

C .25 0 .5 0 0

D .5 0 0 0 0

E .25 1.0 0 0 1.0

F 0 0 .5 0 0

G 0 0 0 1.0 0 312

Total 4 2 4 2 1

 This is a first order transition table. A second order table adds a pitch to the chain. That is to say, given the last two pitches, what is the probability of the next pitch being C, D, etc. Here is the same chart expanded to include all of Frere Jacque in a second order of probability. Combinations that never occur (e.g., C-A) are left off, leaving 36 combinations.

C C-C C-D C-E C-G D-E E-C E-F F-E F-G G-C G-E G-F G-G G-A A-G Total

D 1

E

F

G 2

A

2 1 2 2 2

1 1

1 2

1

1

2 1 1 2 1 2 9

1

6

2 4

8

2

Total 3 2 1 3 2 4 1 2 2 1 1 2 1 2 2 30

Note that the total combinations are at the bottom. This is a quick way to check if you have the correct number of total connections. The total should equal the number of notes in the piece minus two (because the first two don't have a connection of three items—or second order). Also note the combination C-C has no probability for F. If there were one, then the combination C, C, F could result. But there are no probabilities for C-F. This broken link would return a nil value and crash (or get stuck in a loop). I don't have a quick or clever way to check to make sure you don't have any bad leads. You just have to proof carefully. Here is the chart with percentages. C C-C C-D C-E C-G D-E E-C E-F F-E F-G G-C G-E G-F G-G G-A A-G Total

D 33

E

F

G 66

A

100 100 66 100 5

33 25

25 1.0

5

5

100 100 100 100 100 100 9

1

6

100 4 313

8

2

Total 3 2 1 3 2 4 1 2 2 1 1 2 1 2 2 30

Charts can eat up memory. If, for example, you were to do a chart for the piano works of Webern (which wouldn't make sense because that's not how his music works—but that shouldn't stop you from trying it), assuming a four octave range, 12 pitches for each octave, second order probability would require a matrix of 110,592 references for pitch alone. If you expanded the model to include rhythm and instrument choice, dynamic and articulation, you could be in the billions in no time. So there needs to be efficient ways of describing the matrix. In the Foster example below I do a few space saving short cuts. 27.5::

Transition Table

//A collection of the pitches used ~legalPitches = [60, 62, 64, 65, 67, 69]; //An array of arrays, representing every possible previous pair. ~transTable = [ [0, 0], //C, C [0, 1], //C, D [0, 2], //C, E [0, 4], //C, G [1, 2], //D, E [2, 0], //E, C [2, 3], //E, F [3, 2], //F, E [3, 4], //F, G [4, 0], //G, C [4, 2], //G, E [4, 3], //G, F [4, 4], //G, G [4, 5], //G, A [5, 4] //A, G ]; // Parsing the transTable transTable.do({arg index, i; if(index == currentPair, {nextIndex = i; true;}, {false})}); ::

Since so many midi values are skipped in a tonal scheme, ~legalPitches is used to limit the number of pitches used. The actual code looks for and works around array positions, not midi values. The variable ~transTable describes the first column of my transition table. Each of the possible previous pitches are stored in a two dimensional array. The value used to compare and store the current two pitches is currentPair. It is a single array holding two items, the first and second pitch in the chain. At the beginning of the program they are set to 0, 0, or C, C.

314

The values in currentPair are matched with the array transTable. This is done with a do iteration. In this function each of the two position arrays will be compared to the variable currentPair, which is also a two position array. When a match is found the index of that match (or position in the array where it was found) is stored in nextIndex. In other words, I have found the index position of currenPair. This is necessary because the table has been pared down to include only existing combinations. Next the index for each previous pair is located. If, for example, the current pair was D, E. Their values in the transTable would be [1, 2], and the lines of code above would find a match at array position 4 (remember to count from 0). That means the probability at position 4 is used. In this chart it says there is a 100% chance that C follows the currentPair D, E. 27.6::

Probability chart

nPitchProb = [ //C D E F G A [0.00, 0.33, 0.00, 0.00, 0.66, 0.00], //C, C [0.00, 0.00, 1.00, 0.00, 0.00, 0.00], //C, D [0.00, 0.00, 0.00, 1.00, 0.00, 0.00], //C, E [0.66, 0.00, 0.00, 0.00, 0.00, 0.33], //C, G [1.00, 0.00, 0.00, 0.00, 0.00, 0.00], //D, E [0.50, 0.00, 0.25, 0.00, 0.25, 0.00], //E, C [0.00, 0.00, 0.00, 0.00, 1.00, 0.00], //E, F [1.00, 0.00, 0.00, 0.00, 0.00, 0.00], //F, E [0.00, 0.00, 0.50, 0.00, 0.50, 0.00], //F, G [1.00, 0.00, 0.00, 0.00, 0.00, 0.00], //G, C [0.00, 0.00, 0.00, 1.00, 0.00, 0.00], //G, E [0.00, 0.00, 1.00, 0.00, 0.00, 0.00], //G, F [0.00, 0.00, 0.00, 0.00, 0.00, 1.00], //G, G [0.00, 0.00, 0.00, 0.00, 1.00, 0.00], //G, A [0.00, 0.00, 0.00, 1.00, 0.00, 0.00] //A, G ]; ::

The choice is actually made using windex (weighted index), which takes an array of probabilities as its first argument. The array nPitchProb is a two dimensional array used for the next probability array, say index 4, identified as nPitchProb.at(4). Since it is a multi-dimensional array, there are two indexes, e.g. nPitchProb.at(4).at(5). Using variables this becomes: (nPitchProb.at(prevPitch).at(nextIndex). The array for windex has to total 1 which is accomplished with normalizeSum. The return from windex is an array position, which is stored in nextIndex. The variable nextPitch is an array position that can then be used in conjunction with legalPitches to return the midi value for the correct pitch: legalPitches.at(nextPitch). It is also used for some necessary bookkeeping; rearranging currentPair to reflect the new choice. The value in the second position of the array currentPair is moved to the first position, and the 315

nextPitch value is stored in the second position. (In other words, currentPair was D, E, or array positions 1, 2, C or 0 was just picked according to the table, so what was [1, 2] becomes [2, 0].) currentPair.put(0, currentPair.at(1)); currentPair.put(1, nextPitch);

A more complex example of a transition table and Markov process is demonstrated in Example 27.7. It uses a small section of Harry F. Olsen's transition table for the works of Stephen Foster. The system does not control duration using a Markov, and has no information about phrase structure or cadences, which is apparent. Predetermined phrases could be gathered in a series of Pseq arrays, which would give more structure, but would not account for placement of tonic vs. leading tones. The result? It is very Fosterish, F nearly always moves to G. 27.7::

Foster Markov

MIDIClient.init(0, 1); m = MIDIOut(0); m.program(0, 4); m.noteOn(0, 60); m.noteOff(0, 60); /* For tests a = Pbind(\tempo, 1, \amp, 0.8, \type, \midi, \midiout, m, \channel, 4, \dur, 0.1, \sustain, 0.05, \midinote, Pfunc({48.rand + 36})).play; a.stop; */ ( var wchoose, legalPitches, previousIndex, prevPitch, currentPitch, nextIndex, nextPitch, nPitchProb, pchoose, blipInst, envelope, pClass, count, resopluck, seed; seed = 0; // random seed // seed = = Int32Array[ -571935851, 1571969825, 1165557446 ]; // randData seed if(seed == 0, {thisThread.randData.postln}, {thisThread.randData_(seed)}); {1000.rand}.dup(10); prevPitch = 3; currentPitch = 1; count = 1; pClass = #["A3", "B3", "C4", "D4", "E4", "F4", "F#4", "G4", "A4", "B4", "C5", "D5"]; //pchoose is the mechanism for picking the next value. pchoose = 316

{ legalPitches = [57, 59, 60, 62, 64, 65, 66, 67, 69, 71, 72, 74]; //Both prevPitch and nextPitch are not pitches, but array positions. previousIndex = [ [2], //previous is 0 or A3 [2], //1 or B3 [0, 1, 2, 3, 4, 5, 7, 9, 10], //2: C4 [1, 2, 3, 4, 7, 10], //3: D4 [2, 3, 4, 5, 7, 8], //4: E4 [4, 5, 7, 8], //5: F4 [7], //6: F#4 [2, 4, 5, 6, 7, 8, 10], //7: G4 [2, 4, 5, 6, 7, 8, 10], //8: A4 [8, 10], //9: B5 [7, 8, 9, 10, 11], //10: C5 [7, 9] //11: D5 ]; previousIndex.at(prevPitch).do({arg index, i; if(index == currentPitch, {nextIndex = i; true;}, {false})}); nPitchProb = [ // [00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11] array position // A, B, C, D, E, F, F#, G, A, B, C, D [ //arrays for A3 [00, 00, 16, 00, 00, 00, 00, 00, 00, 00, 00, 00] // one array: C4 ], [ //arrays for B3 [00, 00, 05, 06, 00, 00, 00, 05, 00, 00, 00, 00] // C4 only ], [ //arrays for C4 [00, 00, 16, 00, 00, 00, 00, 00, 00, 00, 00, 00], // A3 [00, 00, 16, 00, 00, 00, 00, 00, 00, 00, 00, 00], // B3 // [00, 02, 02, 09, 02, 10, 00, 00, 00, 00, 00, 00], original C4 [00, 06, 02, 09, 02, 06, 00, 00, 00, 00, 00, 00], // C4 [00, 00, 03, 04, 08, 00, 00, 01, 00, 00, 00, 00], // D4 [00, 00, 00, 07, 03, 02, 00, 04, 00, 00, 00, 00], // E4 [00, 00, 00, 00, 11, 00, 00, 00, 05, 00, 00, 00], // F4 [00, 00, 00, 00, 04, 00, 00, 12, 00, 00, 00, 00], // G4 [00, 00, 00, 00, 00, 00, 00, 00, 00, 00, 16, 00], // A4 [00, 00, 00, 00, 00, 00, 00, 02, 11, 03, 00, 00] // C5 ], etc. ]; nextPitch = (nPitchProb.at(prevPitch).at(nextIndex).normalizeSum).windex; //current is set to previous, next is current for next run. The actual pitch 317

//is returned from legal pitch at nextPitch. [pClass.at(nextPitch), legalPitches.at(nextPitch)].post; // if((count%10) == 0, {"".postln};); count = count + 1; prevPitch = currentPitch; currentPitch = nextPitch; legalPitches.at(nextPitch) }; a = Pbind(\tempo, 1, \amp, 0.8, \type, \midi, \midiout, m, \channel, 4, \dur, Prand([ Pseq(#[1]), Pseq(#[0.5, 0.5]), Pseq(#[0.5, 0.5]), Pseq(#[0.25, 0.25, 0.25, 0.25]), Pseq(#[0.5, 0.25, 0.25]), Pseq(#[0.25, 0.25, 0.5]) ], inf), //\dur, 0.1, \art, 0.5, \midinote, Pfunc(pchoose)).play; ) a.stop // All notes off: m.allNotesOff(1); m.allNotesOff(0); ::

This concludes our discussion of control sources. Below is a summary. •

Control Source Review

Periodic Wave Forms Aperiodic Wave Forms Sequencers Sample and Hold External data Probabilities

Character Exact repetition, classic (cliché) Random, classic (cliché) Exact, but custom, complex repetition Evolving patterns, blend between variation/repetition Complex, meaning through context, natural patterns transfer Artificially intelligent

Examples Sine, Saw, Pulse LFNoise1, LFNoise0 Dseq, Ultrabeat Latch Data files Data tables, if()

• Music of the Spheres Music, art, and nature is made of varying degrees of repetition and variation. Humans are hard wired to recognize and interpret repetition and variation. Effective music strikes the correct balance (for the desired effect) between repetition and variation. Too much and it approaches noise, seems random. Too little results in trite, uninteresting events. It's all subjective and relative. It is subject to our ability to perceive. Our personal and collective sense for structure matures and evolves. Our taste for musical style likewise evolves. 318

Though inharmonic spectra are not heard as pitch, the constructive and destructive interference still blend into a recognizable pattern. We are just as capable of distinguishing a set of inharmonic frequencies, when repeated, as we do a melody. The harmonics blend into a unique timbre. The Solar System patch translates astrological data into sound. The first example transposes the rotation of each planet in our solar system to an audio frequency range. It is then couched in a bell patch. Run the "Just us" segment, listen for a while, then the second segment, which adds 8 other known planetary systems within our galaxy, and 4 outside our galaxy, each transposed to a range that humans can perceive. Put yourself in a virtual space ship, traveling between galaxies. You are lost, but your perception has evolved to include magnetic fields. You listen to the rotations of each system you pass. Could you find ours? Is it random? 27.8::

Solar System

( { // Just us: var scale, specs, freqs, amps, rings, numRes = 5, bells = 20, pan; scale = [60, 62, 64, 67, 69].midicps; Mix.fill(1, { freqs = 72.midicps * [175.94, 116.75, 1, 1.027491, 0.41011, 0.4264, 0.6125, 0.7667, 6.3863]; amps = Array.fill(9, {rrand(0.3, 0.9)}); rings = Array.fill(9, {rrand(1.0, 4.0)}); specs = [freqs, amps, rings].round(0.01); // specs.postln; pan = (LFNoise1.kr(rrand(3, 6))*2).softclip; Pan2.ar( Klank.ar(`specs, Dust.ar(1/6, 0.03)), pan) }) }.play; )

( // 12 other known planetary systems* { var scale, specs, freqs, amps, rings, numRes = 5, bells = 12, pan; scale = [60, 62, 64, 67, 69].midicps; Mix.fill(bells, {arg c; freqs = Array.fill(numRes, {rrand(1, 15)*(scale.choose)}); amps = Array.fill(numRes, {rrand(0.3, 0.9)}); rings = Array.fill(numRes, {rrand(1.0, 4.0)}); specs = [freqs, amps, rings].round(0.01); 319

// specs.postln; pan = (LFNoise1.kr(rrand(3, 6))*2).softclip; Pan2.ar( Klank.ar(`specs, Dust.ar(1/6, 0.03)), pan) }) }.play; ) // *Technically untrue ::

• Exercises 27:1 Generate a second order Markov Chain for the melody of Lullaby of Birdland. 27:2 Invert the probabilities of the Foster Markov patch. Use only the existing values, but add code so that they are the compliment of 16; 2 becomes 14, 10 becomes 6. 27:3 Generate pi to a few thousand places. Use it for the data component of one of the examples in this chapter.

320