The Next Station – ‘if only I had’

Tomorrow sees the launch of The Next Station, a project by Cities and Memory to reimagine the sounds of the London Underground. My contribution to this project is an audio work called if only I had, constructed entirely from a 3’42 recording of a train arriving and departing from Pimlico station.

The title is taken from Spike Milligan’s ‘Adolf Hitler: My Part in his Downfall’:

‘Edgington and I promenaded the decks. Harry stopped: “If only I had a tube.”
“It’s quicker by tube.”

… an inconsequential pun that has, for some reason, always stuck in mind!

I made this piece as a personal study into the possibility of using livecoding techniques in SuperCollider to develop a fixed piece. In recent months I have been very active in exploring coding in this way, particularly in the context of algorave: if only I had leverages these techinques. Here’s some of the code I used, with explanation:

Pbindef.defaultQuant = 4;
t =;
~path = "/Users/jsimon/Music/SuperCollider Recordings/pimlicoloops/";

This is a remnant of what turned out to be a bit of a false start to the project. My initial idea was to look through the file for shortish sections, in the region of 2-3 seconds long that, when looped, had some sort of rhythmic interest. This was done offline, using Audacity. I thought it might be interesting to develop the piece by using these fragments almost in the manner of drum loops, and wrote some code to juxatpose them in various ways at different tempi. This didn’t really produce anything very effective however: the material is rather dense and noisy, and when looped together the rhythmic interested was lost in broadband mush of sound.

Instead, I revisited a synth from an earlier project that slices a buffer into 16 pieces for playback:

~bufs = (~path ++ "*.aiff").pathMatch.collect({ |i|, i)});
SynthDef(\slbuf, { |out, buf, slices=16, slice=16, freq=440, sustain=0.8|
var myenv, env, start, len, basefreq = 440, rate, sig, sus;
rate = freq / basefreq;
len =;
start = (len / slices * slice);
sus = * sustain * 1.1;
myenv = Env.linen(attackTime: 0.01, sustainTime: sus, releaseTime: 0.1);
sig =, buf, * rate, startPos: start, loop: 1);
env =, 1, doneAction: 2);, sig * env)

As well as experimenting with reverb, I also had a delay effect in here at one point. Again, the nature of the already fairly resonant material meant that this was not that useful. In the end, I only used the reverb at the very end of the piece as a closing gesture.

~rbus =, 2);
SynthDef(\verb, {|out = 0, room = 1, mix = 1|
var sig =, 2), room:room, mix:mix);, sig)

At some point in developing the project, it occured to me to try playing together the sliced material with the orignal file. This seemed to effective, and gave me a clear trajectory for the work: I decided that the finished piece would be the same pop-song length as the original recording. In experimenting with this approach – playing sliced loops in SC at the same time as playing back the whole file in Audacity – I found myself gently fading the original in and out. This is modelled in the synth below: I used an explicit random seed together with interpolated low frequency noise to produce a replicable gesture:

~file = "/Users/jsimon/Documents/ Simon's music/pimlico the next station/Pimlico 140516.wav";
~pimbuf =, ~file);
SynthDef(\pim, { |out=0, start=0, amp = 1|
var sig, startframe, env;
startframe = start * 44100;,0);
env = ~pimbuf.duration - 9, releaseTime:9));
sig =, ~pimbuf, startPos:startframe, doneAction:2) *, 1.0);, sig * amp * env);

There was a nice moment in the original where the accelerating electronic motors of the departing train created a seried of overlapping upward glissandi, sounding very like Shepard tones, or rather, the sliding Risset variation. Looking to enhance this gesture, I tried a couple of my own hacks before giving up and turning to a nice class from Alberto de Campo’s adclib:

~shep = {
var slope =, 0.2, 60);
var shift =,2,60);
var b = ~bufs[8];
var intvs, amps;
var env =, releaseTime:7),1,doneAction:2);
#intvs, amps =, slope, 12, shift);
(, b, intvs.midiratio, loop: 1, startPos:3*44100) * amps).sum * 0.2

All of the above is essentially setup material. The gist of the composition was in iterative experimentation with Pbindefs, as can be seen below: trying out different slicing patterns and durations, working with the various segments I’d prepared beforehand in Audacity.

Pbindef(\a, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/2, \buf, ~bufs[1], \note, 0);
Pbindef(\b, \instrument, \slbuf, \slice, Pser((8..15).pyramid(1), 32), \dur, 1/4, \buf, ~bufs[1], \note, 0);
Pbindef(\c, \instrument, \slbuf, \slice, Pser((2..5).pyramid(1), 32), \dur, 1/4, \buf, ~bufs[0], \note, 0);
Pbindef(\d, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/4, \buf, ~bufs[3], \note, 0);
Pbindef(\e, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/4, \buf, ~bufs[3], \note, 12);
Pbindef(\f, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/4, \buf, ~bufs[3], \note, [-12,12,24,36]);
Pbindef(\g, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/4.5, \buf, ~bufs[3], \note, [12,24,36]);
Pbindef(\h, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/5, \buf, ~bufs[3], \note, [-12,12,24,36]);
Pbindef(\i, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(12), 1), \dur, 1/6, \buf, ~bufs[3], \note, [-24,-12,12,24,36]);
Pbindef(\j, \instrument, \slbuf, \slice, Pseq((1..8).pyramid(1), 1), \dur, 1/7, \buf, ~bufs[3], \note, [-24,-12,12,24,36], \amp, 0.3);
Pdef(\k, (instrument: \slbuf, buf: ~bufs[3], slice: 2, note: [-24,-12,12,24,36], amp: 0.5, out:~rbus));


The final composition was produced by running and recording the code below, which uses the handy Psym as a way to sequence the gestures developed above. The code by this point is entirely deterministic, and would produce the same piece on every run. No further editing was done, apart from normalising in Audacity.

// start from beginning
10.wait; "off again".postln;

Overall, I’m happy with the piece, and glad to have been able to contribute to this very interesting project.

First public livecode

Last night I stumbled into my first public outing of some livecoding I’ve been working on in SuperCollider. The context was an improvisation night called In Tandem run by Bruce Wallace at the Academy of Music and Sound in Glasgow. I hadn’t intended to play, as I really don’t feel I’m ready yet, but I had my laptop and cables with me, they had a projector, so…!

I was jamming along with three other people, on bass, guitar and analog synth. It all went by in a blur, but everyone there seemed to think what I was doing was ok – mostly making grooves out of a random collection of drum samples, but running some algorithmically chosen chords as well.

The code is below: this is my screen exactly as I left it at the end of the night, mistakes and all. Although Toplap say ‘show us your screens’, they don’t say ‘show us your code’, but… it seems the right thing to do.

// the end!
// they still going
// if you're curious, this is SuperCollider
// musci programming language
// writing code live is called, er, livecoding
// i'm just starting out
"/Users/jsimon/Music/SuperCollider Recordings/hitzamples/".openOS;

Pdef.all.clear; // clear things out
~hitzpath="/Users/jsimon/Music/SuperCollider Recordings/hitzamples/"; // a folder of samples
~hbufs = (~hitzpath ++ "*.aiff").pathMatch.collect({ |i|, i)}); // samples into an array of buffers
t = TempoClock(140/60).permanent_(true); // tempo 140 bpm
u = TempoClock(140/60 * 2/3).permanent_(true); // tempo 140 bpm * 2/3
SynthDef(\bf, {|out=0 buf=0 amp=0.1 freq=261.6255653006|
var sig =, buf, * freq/60.midicps, doneAction:2);, sig * amp)
}).add; // this whole chunk defines a synth patch that plays samples

// Pdef.all.clear;
//"/Users/jsimon/Music/SuperCollider Recordings/".openOS;
// t.sync(140/60, 16);

(instrument: \bf, \buf: ~hbufs.choose).play; // play an event using the synth called \bf
// pick a randoms sample from the array
(instrument: \bf, \buf: ~z).play;
~z = ~hbufs.choose;

t.sync(140/60, 32); // gradual tempo changes possible
u.sync(140/60 * 2/3, 16);
v.sync(140/60 * 5/3, 16);

Pbindef(\x, \instrument, \bf, \buf, ~hbufs.choose).play(t).quant_(4);
Pbindef(\y, \instrument, \bf, \buf, ~hbufs.choose).play(u).quant_(4);
Pbindef(\z, \instrument, \bf, \buf, ~hbufs.choose).play(v).quant_(4);
Pbindef(\z, \instrument, \bf, \buf, ~hbufs.choose).play(v).quant_(4);
~g1 = {~hbufs.choose}!16; // choose sixteen samples at random = one bar full
~g2 = {~hbufs.choose}!16;
Pbindef(\x, \buf, Pseq(~g1, inf)); // play those sixteen samples chosen
Pbindef(\x, \buf, Pseq(~g2, inf)); // different sixteen, so, a variation.
Pbindef(\x, \dur, 0.5);
~d1 = {2.rand/10}!16;
~d2 = {2.0.rand/10}!16;
Pbindef(\x, \amp, Pseq(~d1, inf));
Pbindef(\x, \amp, 0.2);
Pbindef(\x, \note, Prand((-36..0), inf));
Pbindef(\x, \note, Pseq({(-24..0).choose}!16, inf)); // pitch each sample down by random amount
Pbindef(\x, \note, nil);

// hmm. blx diminished, that's just C major!
// was using \degree instead of \note, better sounds a bit more like messiaen now :)
~c = {var x = Scale.diminished2.degrees.scramble.keep(4).sort; x.insert(1,(x.removeAt(1)-12))};
// hexMajor thing also works beautifully now!
~c = {var x = Scale.hexMajor6.degrees.scramble.keep(4).sort; x.insert(1,(x.removeAt(1)-12))};

// next question might be changing \note, \dur and \root in a coordinated way
Pbindef(\k, \note, Pstutter(Prand([5,7,9,11,13]*2, inf), Pfunc(~c)),
\dur, 0.5,
\root, 3, // best option for feeling of key change
\amp, Prand((2..5)/70,inf)

Ball of Sardines @ Red Note ‘Noisy Nights’

excerpt from score

Just for fun, I did an arrangement of Ball of Sardines for one of Red Note’s ‘Noisy Nights’: flute violin, trombone, and the conductor playing kethuk. (This is instead of the piece I was going to write, a masterpiece of algorithmic pointilism to be entitled ‘Moment of Indecision’. I may finish that one day…)

Monday 1 February 2000-2200 – Free
1 Summerhall Place
Edinburgh – EH9 1QH