This is a journal and a feed of sonic debris generated by my practice / practise / practising. I used to post these directly to but I’m going to post them here now and point to this page from elsewhere. These posts will be a little transient but I want to leave the ones I like up, so it should coalesce into a collection of sorts. I will hopefully do some backfilling of old posts.

Feel free to download these for your own listening. If you’d like to use them in your own work, please get in touch, I’m open to that in principle. (I might put a CC something or other on everything posted to this page but I have to examine and think about the licenses yet, so it’s case-by-case for now.)


Muted-feeling jabs with dirt. Pleasing level of dynamism.

Tried to get as far as I could with only a few modules, in the end all of the movement comes from one module. This is a dual envelope generator with a few extra utilities (slew limiter, clock divider, 4-quadrant VCA, OR comparator between the envelopes) and some nice controls (volt-per-octave inputs in addition to rise & fall times, also control over shape and attenuverters on some outputs). (IYKYK.)

(If you’re not into analog synth: volt-per-octave is a standard for pitch control which specifies a logarithmic interpretation of the incoming value, and a range of one octave for each volt. For example, if you send an oscillator’s volt-per-octave input a straight slope from zero to one volt, you would hear a smooth glissando over one octave. This is particularly handy on envelope generators because it means they can double as oscillators.)

The patch is:

There’s also a delay with some filtering on it.

0:05:55 / flac (44M) / mp3 (5.5M)


Another revision of that granular ring modulated feedback patch. This time with the envelope-following limiter and more careful fader surfing. Still some clipping but nowhere near as dirty as yesterday’s.

Yesterday’s fixes in grains~ are now live.

0:05:15 / flac (44M) / mp3 (4.9M)


A modification of yesterday’s patch after chasing out some gremlins (and possibly finding some other ones). This one uses three modulation tones (yesterday’s 40 Hz plus 400 and 1200 Hz) which I manually mix in & out. Also some tweaking of other parameters like grain rate, position / pan / size variation, and pre-ring-mod feedback. Somehow it ended up much noisier 🤷

The gremlins in grains~ were hiding in two spots: first, the per-grain DSP toggle was turning off too early and truncating some grains. A fix for this will go live tomorrow. Second, I was reading over the loop-around point in the circular buffer. grains~ can protect against this when using a static buffer, but to use a live one (as in a delay), the delay value is sent to the circ-buff~ object, which means it’s on the user to account for all possible values which might push a grain over the edge. (In my case, delay, position variation, grain size, grain size variation.) I’m going to make a note about this in the README (also tomorrow). It’s a lot for a user to think about and it would be nice to hide away, but to do that grains~ would need to include the buffer, which means a pretty substantial modification. I might rather do this in an eventual Rust rewrite – that’s for another day / month / year.

0:05:15 / flac (68M) / mp3 (4.9M)


Tried the Jaap Vink ring mod feedback patch totally digital, with [grains~] as the delay and [round-clip~] from dried-utils as a limiter. Lots of crunchy noise and digital glitchiness. I think the gremlins from a few weeks back really love feedback. (What I really think is happening is some performance improvements I made in [grains~] to turn off DSP in inactive voices cause glitches which get amplified by the patch. I’ll have to fix them eventually.)

For the modulator signal I used a trick to get an analog-sounding sine wave by sending a (band-limited) triangle through tanh with a gain of 2-3x. tanh rounds out the corners to get it nearly to a sine with some of the triangle harmonics still coming through. (I think this might be similar to how analog triangle-core oscillators generate sine waves IRL, but I haven’t looked into it yet.)

(First 10 sec are silent for no good reason.)

0:04:35 / flac (49M) / mp3 (4.2M)


Quiet and sometimes less quiet feedback in regular, ragged-edge swells. Sporadic clicks. Noise floor.

Similar patch with different modulation and a new feedback path through a delay and a (different) filter. The ring mod modulator tone is folded a bit but steady, no modulation on the folder.

0:07:32 / flac (86M) / mp3 (6.9M)


More heavily modulated feedback.

Similar feedback patch at its core, except the ring modulator affects the pre-filter signal. There are also some LFOs animating things: the wave folder offers two different timbral outputs, each have a different LFO bringing them up in a mix that goes to the ring modulator; a sine tone modulates the fold level of the signal modulating the ring modulator, and an LFO brings it up or down as well.

0:04:25 / flac (49M) / mp3 (4.1M)


More drone feedback with more folded ring modulation and more noise floor / noise on the floor. After wrapping up yesterday’s session I noticed I could shift between the resonating harmonics in the feedback by manipulating the spectrum with the fold level of the ring modulator modulator signal. Worked some more with that today.

0:09:49 / mp3 (9.0M)


More sustained midrange feedback on the Haptic Box. No scanning this time, working on exploring just these nodes as well as ring modulation with them. Today’s fun trick is wavefolding the modulator signal for the ring mod.

(Lots of noise floor from the mic on this one. I don’t want to be super picky about things like that, but boy do I not like it.)

0:05:17 / flac (29M) / mp3 (4.9M)


Long feedback tones in the midrange of the Haptic Box. Mostly just dialling in filter frequencies, but there’s some ring mod in there too, which sounds great. There’s also some clipping / buzzing (not sure which).

0:03:49 / flac (30M) / mp3 (3.5M)


More drone. A similar sound and granular delayed signal. I built an interface around yesterday’s patch and fiddled with the feedback. Ended up getting some reverb-like sounds, which I’m happy about because I don’t know a thing about reverb algorithms.

0:06:05 / flac (52M) / mp3 (5.6M)


A drone bit. A ring modulated sound is mixed with its granular delayed signal. Grain overlap slowly increases over 2 minutes, little digital gremlins hang out, then grain overlap decreases over 2 minutes. Lots of fun pulses in the low end.

Granular delay done by grains~. I made some performance gains in grains~’s core a few weeks ago, and just tonight I discovered I could eke out more performance by manually clocking grains~ and also circ-buff~.

0:06:02 / flac (46M) / mp3 (5.6M)


A more standard version of the Jaap Vink patch. I’d never got it to bloom the way I’ve heard it in recordings until tonight, so I’m pretty happy about that. I think it needs more gain than I was giving it, and the active limiter (that is, using an envelope follower inverted and applied to the gain, rather than something that just saturates like I used yesterday) probably contributes to the slow changes.

Here the same guitar pedal is used, passed through a tanh saturator for gain, through a wavefolder (not modulated) for a bit of noise, ring modulated against a steady sine, passed through a filter acting like a low-pass gate, then back / out.

0:08:00 / flac (45M) / mp3 (3.7M)


A tweaked Jaap Vink patch. The original is more or less a feedback loop through a tape delay, ring modulator, and ad hoc limiter. Here I use a wavefolder instead of a ring modulator, modulating both the fold level and the “shape” input. tanh clipping takes the place of the limiter, and an analog delay guitar pedal takes the place of the tape delay. I also feed back the output of the wavefolder to the FM of the modulating oscillator, and pass the whole thing through a completely open low-pass filter to trim some of the ultra-highs.

The result is nothing like the smooth reverb sounds of the Vink patch but quite tasty in a different way. There’s a ton of crunchy noise with a steady low fundamental and squirrelly rhythms in the noisy midrange and chirpy highs.

3’ 26” / flac (22M) / mp3 (7.9M)


Some quieter noise for you. Several feedback paths through a wavefolder moving through quiet clicks and crackles and noise floor.

5’ 15” / flac (19M) / mp3 (4.9M)


More or less the same configuration as yesterday’s second recording, the main difference is that every 20 gates from the lower voice triggers a new random ASR feedback distribution.

1’ 51” / flac (7.4M) / mp3 (871K)


Two experiments following up on yesterday’s ASR adventures.

The first is an attempt to hear the differences that the various feedback configurations make. I used a melodic hocket because I though it would be relatively easy to hear the deviations between the two voices. The lower voice stays a simple sample-and-hold throughout and the top voice is affected by the shift register, which is sampling an LFO. Both are quantized to a major scale mode. I moved through the same configurations as yesterday’s, increased the clocking tempo, then backwards through them. It’s clear when the shift register starts because the top voice no longer hockets at the octave, choosing instead different pitches, but again the difference between the different feedback amounts is ambiguous (and again I do think there are differences, but I’m not able to point to them exactly).

2’ 54” / flac (21M) / mp3 (2.7M)

The second is more of a fun excursion. The code is changed so there are two ASRs. These are clocked by the zero-crossings of the opposite’s data input (when A’s data input moves from negative to positive, B processes a sample from its own input, and vice versa). Whenever they are clocked, a new “note” is played by a dedicated voice (A has its own voice, B has its own voice, which are incidentally quantized to a more interesting scale, some JI something or other). The data inputs are fed by two LFOs. The bottom’s frequency remains constant and the top’s is modulated by the bottom’s output and also by the bottom’s ASR output. (The bottom voice doesn’t sound rhythmically steady, I’m not sure why that is.) I moved through different feedback value configurations – I think they were more clearly different in this recording, but again I’m not sure how.

4’ 45” / flac (38M) / mp3 (4.4M)


A little experiment with a shift register and wave interference.

The kick sound is an old trick that uses a wavefolder on a low oscillator as a VCA, so the sound loses harmonics as its amplitude envelope decreases. (Like a low-pass gate but with a different sound. Additional crunchiness from an analog delay pedal. Occasional transient clicks courtesy of my envelope generator.) The audio from the kick sound is the input for the shift register. The shift register is clocked by a comparison between the kick audio and another oscillator: when they cross, the register takes a new value from the input. The first output of the shift register feeds the frequency control of both oscillators.

The fun part is that because the analog shift register is digital (analog as in not binary gates, digital as in code), its behaviour is very mutable. I’m specifically interested in how feedback affects the output of the register values. Feedback here meaning that values from the shift register are mixed in with the new value to become the shift register input. I’ve been imagining all kinds of cool structures of interfering matrices of shift registers, but I’m not sure how it actually plays out. Part of the experiments I want to do this #noisevember revolve around listening to these changes.

This short recording moves through a few configurations of a 6-step register (even though only the first value is taken):

  1. No new values
  2. No feedback (effectively a sample & hold)
  3. 50% feedback from the final value
  4. “Memory fading” feedback (more recent values have more presence in the feedback than more distant ones, ranging from 0.3 to 0.01)
  5. Reverse “memory fading” (more distant values have more presence than more recent ones)
  6. Return to no new values

It’s really difficult to hear what the differences are between the sample and hold and the feedback configurations, but it does seem like there are differences. More experimenting required.

2’ 22” / flac (11M) / mp3 (2.2M)


More box feedback meandering at a more even volume than yesterday’s. Also line out from the box, but somehow that 50Hz hum is gnarlier. Some pretty fun stereo effects.

8’ 55” / flac (68M) / mp3 (8.2M)


Feedback whistle duet on Haptic Box. (First minute is much quieter than what follows.)

4’ 19” / flac (35M) / mp3 (4.0M)


Shimmers from Haptic Box.

1’ 41” / flac (3.0MB) / mp3 (1.6MB)


It is still ochdtober, here are eight unsynched LFOs chewing their way through a wavefolder and some rectifiers.

3’ 31” / flac (17MB) / mp3 (1.7MB)

Here’s another one, closer to the world I think I was first aiming for with this analog synth stuff, which I take as a good thing. It’s all performed, no feedback-self-modulating-generative nonsense. I didn’t wire up any controllers, and turning small knobs in a big nest of cables to get short chirpy sounds is not very intuitive, as it turns out. I think I’m going to keep this patch up and see how it interacts with my current controllers.

7’ 00” / flac (39MB) / mp3 (4MB)


Taps and tones on the Haptic Box.

flac (21.51MB) / mp3 (2.11MB)


It is still Septimbre.

This one is another feedback patch on my main system featuring more high pitched chirps. Most of this was changing the balance between three different routes through the feedback.

Wavefolder -> quad filter -> mix -> delay -> limiter -> wavefolder
           -> ring mod    -> mix
Complex LFO nonsense modulates filter pitch, oscillator pitch, and pings wavefolder fold level

flac (76.76MB) / mp3 (7.32MB)


It is Septimbre.

This is a feedback patch using a semi-modular monosynth and an analog delay pedal (and also using some analog clipping at the mixer as a dirty limiter). Part one: mix in the delayed signal with the oscillator before its filter and tweak levels a bit. Part two: add complex audio rate modulation to filter resonance and tweak levels a bit.

flac (33MB) / mp3 (5.6MB)


I performed that last patch with composer Mia Windsor on the 25th, you can watch it on YouTube (or YewTube if that’s more your speed).

The code I used to get the drawing tablet to talk to the modular synth is called inscribe (source). I’ve made quite a few updates to it that aren’t yet released (and the naming will eventually change, so watch out). The live version of the script will simply spit out CV for X, Y, and pressure, and a gate for contact, but the new version decouples that data by calling the public API of monome crow so parameters can be used arbitrarily in crow scripts. I used it with the zone math I mentioned a few weeks back. I’ve also written a simpler, similar script to use any HID interface as a controller by converting its evdev events to OSC. I might merge these, but I’m not sure.


Short snippet of long droney sounds.

Oscillator through different wavefolders (left and right), slowly modulated, then through two pairs of filters (left and right) whose resonance is occasionally boosted, then mixed with oscillator core before going out. One wavefolder is additionally rectified, passed through a relatively high bandpass filter, and back to the folder’s fold control for audio rate feedback (the squealy / chirpy sounds).

flac (30MB) / mp3 (3.9MB)


Another one using more or less the same patch, also quite bright. Learning to love those sustained 10+ kHz whistles – hide your pets. Opens with some similar sounds from yesterday, chugging, full spectrum but quite bright, then eventually settles into something sparser / quieter with brain tickling chirps and some low growly interruptions. This quieter section basically played itself and I didn’t tweak too much, which I was really into. Finding those evolving, self-changing systems is part of the reason why my patches get so tangled.

flac (69.45 MB) / mp3 (12.33 MB)

So about those patch notes…

I’ll try to explain it in generic terms, but there are some specifics that can’t be generalized. If you’d rather look at a diagram, here’s the drawing I came up with.

As mentioned in yesterday’s post, this thing centres on wavefolder feedback. At audio rates, wavefolders add harmonics to a sound, making it sound richer (and kind of buzzy). This sort of works like clipping, except instead of cutting off the top portion of the waveform, it’s mirrored down. Wavefolders add gain to push the incoming signal against this limit so that it’s folded back in on itself, adding the harmonic. When the folded signal reaches the opposite limit, it’s folded back. Wavefolders often have symmetry controls as well, which offset the incoming signal so that it folds more on one side than on the other.

I discovered while making zonder that analog inputs that directly affect the sound can be treated like mix inputs, for example modulating the fold intensity of a wavefolder or the resonance of a filter with an audio-rate signal will result in that signal being present in the output signal. That means that these control inputs can be used in a feedback path. This discovery is important because the second wavefolder in my system is part of an oscillator, so the oscillator is the only thing going into its audio input. A feedback loop is completed by looping the audio back in through the oscillator-folder’s fold input (at least, I think that’s what’s going on).

The basic feedback loop goes like this:

The folder-folder mix also goes to

Other smaller feedback loops:


Now for that particularity I mentioned. The mixer, the half-wave rectifiers, and the envelope follower are all part of the same module, Cold Mac, which has an output that mixes together all of their inputs. That output is the other channel of audio in the recording. Another peculiarity of Cold Mac is its macro control, which in this case controls both baselines of the half-wave rectifiers.

That’s the basis of the audio-rate rat’s nest. There is a small amount of modulation:

There’s a ton of area to explore in this patch. To perform it, I was mainly tweaking the controls of the dedicated wavefolder (fold, shape, and symmetry) as well as the Cold Mac macro. Changing the rates & cross-modulating FM depth of the semi-random voltage generating LFOs had a good deal of influence on the overall rate of things as well. The resonance controls of the filters were fairly predictable performance controls as well.



So that album-shaped thing zonder is basically all wavefolder sounds. I got a new oscillator recently that both a) has a wavefolder baked in and b) goes into LFO rates, so I wanted to see if I could get it to do something similar. The trick is that it only has one dedicated input for the wavefolder intensity, and the rest has to be done by manipulating the oscillator core. Anyway, like always happens with these feedback patches, things got out of hand, and here’s the result. I might try to draw out the patch, I have a copy of a patch by Jason Lim of Instruo and it looks like a good method (better than writing a play about it). If I do, I’ll update this post with a link or something.

Fair warning: it’s noisy and quite bright. If you’re not into high frequencies and kzzht crhsshhht hshhh, this might not be the one for you. There will be others, I don’t exclusively do this noisy stuff!

flac (72.6 MB) / mp3 (11.8 MB)


I just posted an album-shaped sequence of audio recordings here, which is useful if you like listening to albums. I might write more about it soon, but it’s very much in the spirit of this page.


Two versions of the same multi-layered feedback patch.