Modular Composition: Learning to Make “Sounds” to Accompany “Notes”

I’ve been playing ‘structured’ music for most of my life – I grew up playing trumpet in various school bands and learning songs on guitar, I took music theory classes in college, and in recent years I started playing synths to learn more about theory and synthesis, and I recently (in the last month) dipped my toe into eurorack. After spending years locked into songs and structure, I’m finding that much of the synth music I like the best combines “notes” – a melodic sequence of some kind, vaguely chord-based structure – with several layers of “sounds” – a backdrop of textural sounds, recurring little beeps and boops, the occasional percussive strike, little sampled whatnots that fill in the blanks and make the “notes” seem less repetitive and more interesting. Sometimes it’s all sounds and no notes. I’m so inspired by what I hear on this thread, but as someone who has only ever really played more “structured” music, I’m having a hard time understanding this method of composition.

I guess my main question is – where do those come from (in your head, not literally)? What are the elements of those sounds? When do they enter the track-building process? Do you start with the notes and build the sounds around them? Are they happy accidents or carefully planned or both? So much of what I hear in the Latest Tracks thread and elsewhere on this site illustrates this method of composition (also cc @stripes amazing Water Memory), and I want to learn more about the approach.

I have a pretty small setup and a 1-channel audio interface, so I’m thinking it’ll be about multitracking for me. I’m a definite newbie to recording in general and also multitracking/mixing down multiple tracks, so any tips in that arena would be majorly appreciated as well!

If you’re wondering, here’s what i’m working with with plus an 0-Coast, SQ-1, and Keystep.


Highly recommend listening to the @ioflow Sound + Process episode for the breakdown of his process (recording tracks on top of each other without listening to them come together until the end).


when i came to modular synthesis my musical background was pretty much only playing and recording very structured music. (however my approach to that has always also been to “press record and see what happens”, just with a lot more editing). when i discovered the ability to essentially compose a complete piece of music for many voices within one instrument, that really changed the way i thought about composition. that, and sequencers, and trying to find all kinds of new ways to clock them.

as i learned synthesis through exploring my ever growing case of modules, i also learned exactly how i like to compose and record this new kind of music for myself. in my case this was separating each patch out into voices. this usually begins with three (the drawing on the cover of water memory): bass, middle melody, and high sparkles. i usually work on the sound of one voice at a time as soon as i know which module will be my sound source, what will sequence it, how i will modulate it (if at all), and finally what effects (delay, reverb, grit, etc…). then sometimes i add a wash of ambience, a field recording, a harmony. really all these things just depend on how i am feeling and how the sounds of what i’ve started with inspires me to add.

in terms of the actual notes, that depends. sometimes they are complete happy accidents, less often they are already in my head when i sit down to the synth, but mostly the melodies come after plugging the v/oct into my sound source, listening to what immediately happens, and then make changes when i hear a note that i don’t like, or when i hear a note in my head that i want to come next. i guess another thing to note would be that before i even start patching i always tune all my sound sources to the same root note to start with. if i want to transpose i do that later, but starting this way is inspiring because you will get something musical right off the bat that you can edit as you really form the patch.

making things that are generative or non linear is also something that i really love to do and i think can really give a patch very natural and human qualities. i have never been interested in making music sound snapped to the grid. even in my band we try to record without click tracks whenever possible. when you have one sequence that is odd or not quantized to a clock, it will dance around whatever else you have going, creating an ever changing track that gives the impression of newness. i also live record all of my tracks in stereo, and usually will be just slightly changing things. subtle changes really go a long way.

i started a thread last summer about making an album of modular synth music, and so many people gave really great insights to their process in regards to the technicals of recording, the philosophy of composing, workspace arrangements, and just all around really great discussions about some of what you’re going through now. high recommend giving it a read if you haven’t :slight_smile:

just keep exploring your rack. you’ll find what works for you when you find what doesn’t. don’t be afraid to make simple patches – they can be just as, if not sometimes more beautiful as crazy complex ones, and somehow get some reverb on it all :smile:


Thanks so much for the perspective! This is perfect.

Also, I TOTALLY agree about the “snapped to the grid” feeling. Whenever I fire up the SQ-1 and build a sequence I feel like i’m making music for Blade Runner. Which is great sometimes, but I also want to relax a bit and space out. Layering slow quantized and unquantized sequences is a great idea.

Also, thanks for the tip on your other thread!


From working with various kinds of hardware music creation its lead me to the approach of treating any musical device or system as a collaborator in a composition. So with modular yeah its pre patch ur ideas, slowly sending ur patch into a programmed complex mess of chaos and noise and then from there is where some real magic can spring forth…tuning, dialing back, scaling things… exploring intuitively with the machine as it generates infinitely varying and manipulatable sonic phenomena.

Gotta keep an open mind as well as use the machine in an explorative and exploitive fashion to discover the real hidden niches and grooves :wink:


For me, personally, they come from happy accidents - I experiment with different sound processing techniques and something cool will come out. Starting with found sounds and field recordings always makes something interesting.

(the MIT open courseware on making music with computers gives great info -

I have experimented a bit with sort of reversed spectralism ( and by taking found sounds and using EQ to only use the frequencies that matched the pitches I wanted…starting with that might give you something halfway between sounds and pitches like you are describing…


One thing I’ve done is make long generative pieces on the modular and then cut them up into loops in Ableton. I will layer the loops, sometimes up to 4 layers at a time (but usually just 2 or 3), and set different time signatures for each one. This creates poly rhythms and if you can get it to sound good you’ll end up with a melody that is constantly changing. All kinds of surprise harmonies and rhythms can happen from this. I may then record all of the layers onto one track and further edit it down from there. It’s just one way I come up with melody ideas. Sometimes it works out nicely and when it doesn’t I just move on and try something else. There are loads of videos about this on youtube.

As far as using sounds goes, what I’ve been doing recently is multing my QCV from the modular into my ES8 and then using the Silent Way CV to midi plugin to trigger a Kontakt instrument or Sampler. Then I route that audio from the computer back into my modular and blend it in with my oscillator. I usually use some kind of acoustic instrument or atonal sound. You can get really amazing timbres and textures this way.


one note on structuring ‘sound’ rather than ‘notes’: you might want to play with sequencing things other than V/Oct inputs. that way you get structure, but not necessarily in the manner of musical notes. For instance, I really like running a Turing Machine into the CV in on a filter or wavefolder; it gives movement to a texture, but also can give it repetition or evolution. You could do similar patching sequencing or semirandom sources into a VCA CV input. It’s particularly interesting with things that shape harmonics, though (hence my ‘filters/folders’ suggestion) - it sounds as if something musical is happening, though you can’t quite pin down what.


Great suggestion!!! I find that I do this a lot within the framework of “notes”. Like everything I didn’t plan it, I was interested and stumbled upon something. This is the first time I’ve actually reflected upon it and written something down, so apologies if it’s a bit less than clear. Often there is polymeter, the note-sequence may be a 7/8 loop, while the timbre sequence is 5/8 or 9/8. (Or many other choices!) Sometimes a S/H is used to latch the timbre sequence to the note sequence, but it’s more interesting not to do this.

An important step (for me) in getting this to work is to establish a multidimensional timbre space in some way that is also independent of filtering. Even if we’re talking about sequencing timbre along one dimension, it still helps to have another dimension in reserve to further shape the sound and place it appropriately in the mix.

I found two particularly effective ways to create a multidimensional timbre space, one for “East Coast” setups (in this case the Aries 300), the other for “West Coast” (in this case Serge). This happens with the oscillators before any filtering. With percussive envelopes both are in the general domain of metallic sounds: bells, guitars, scrap metal… that which Inayat Khan has associated with “earth”:

(link: “Chapter I: The Silent Life”, in The Mysticism of Music, Sound and Word;

For the “East Coast” setup, I have three VCO’s, two (VCO2 and VCO3) are synced to the first. All track each other. VCO3 is synced to VCO1 using hard sync; VCO2 is synced to VCO1 using either hard or soft sync (soft is more interesting if available). Basically the role of VCO2 is to modulate the frequency of VCO3. Triangle or saw modulation seems to work best. Usually the mix is VCO1 (sine or triangle) + VCO3 (square). Basically this can be thought of as a hard-sync formant sound (VCO3->VCO1) with some very weird and interesting sidebands caused by the frequency modulation from VCO2. The mix of VCO1 is just to provide solidity and depth, but it’s important. The two useful dimensions of timbral control are: VCO2 and VCO3 frequency (in parallel) – this places the spectral peak, then the difference between VCO3 and VCO2 – this gives a different “flavor” of the sidebands.

maybe it’s better to just do this with math, let CV1 be the common (pitch) control, CV2 be the
VCO2: CV1 + CV2
VCO3: CV1 + CV2 + CV3 (+k*FM from VCO2)

if it’s difficult to mix four inputs (the Aries system allows explicitly for this) you can do
VCO2: CV1 + CV2 - CV3
VCO3: CV1 + CV2 (+k*FM from VCO2)

On the West Coast setup, which favors timbral variation through waveshaping (and in this case I’m referring to Serge because I haven’t had experience on Buchla), the basic idea is two synced triangle oscillators in some mix, say VCO2 synced to VCO1. Generally the higher-frequency oscillator (VCO2) has a higher gain in the mix. The mix (VCO1+VCO2) is sent through a waveshaper. In the case of Serge this is a wave multiplier. Usually the middle wave multiplier alone (odd harmonics) but also sometimes a mix of the middle and bottom (odd+even). Just like the Aries patch it helps to make a more “natural” sound, by mixing in the VCO1 sine/triangle with the output of (VCO1+VCO2->waveshaper). This is a silly and mundane detail but can make all the difference whether or not you like the sound! The two dimensions of timbral control are 1) the difference between VCO2 and VCO1 (be careful to keep harmonic relationships) and 2) the waveshaping amount. Incidentally, I had Kevin modify my NTO/PCO to establish a stronger sync (at his suggestion) perhaps according to the earlier (Serge 73-75) designs.

One thing I’ve never played with is through-zero FM! I think there would be tremendous possibilities here, and perhaps easier to use than the other two methods.

Anyway I didn’t set out to plan any of this, this just came by discovery and a lot of repetitions, it’s perhaps not the best approach to follow these ideas exactly but to see them as possibilities, you will probably find something better, something that comes fully out of your setup, your mind and heart… listen to it, let it out, just repeat it, habituate it, make it a part of yourself… I’m just describing what I have found so far…


This should be printed on leaflets and distributed to all modular synthesists trying to figure out how to get to level 2 modular mastery. It is super fantastic advice that’s totally obvious after you hear it.


This is something I’m also interested in. There’s only a couple of analog VCO’s on the common market that can do through-zero. Doepfer has one, the Intellijel Rubicon (and I’m assuming their soon to be release Rubicon 2) and a new one that’s been teased from hexinverter electronique, called the Mindphaser.

I love analog FM noise, I suspect adding a through-zero module is a next step, but…haven’t gone down that path yet. I asked a question about it here, but it didn’t kick off any conversations. Glad to hear someone else is wondering. Given the Rubicon 2 and the Mindphaser, maybe it’s a soon-to-be-trend?

Don’t forget the SSF Zero-Point VCO! That one looks like a whole new take on analog thru-zero. Particularly excited to give that one a spin.

1 Like

Junto project this week seems custom built for this thread, Disquiet Junto Project 0318: Linear Training

1 Like

Oh awesome, cool. Thanks.

1 Like

Several of the tracks on my “Ocean Music” were done this way. It can produce some very unexpected and rewarding results.

My bio says that I “explore the fractal edge of planning and coincidence”, which is a good summary of how I do stuff like this. Also known as the “what’s this do?” school of composition and performance.


I used the technique a bit for the mixing process of a record I’m (just about!) finished with. Basically listening to the track, taking down notes as to what is missing (maybe bass, maybe something that “glues” things together) building a very quick patch and then running the track through modular/pedals without listening to the other tracks and then coming back and mixing the various tracks together to make things sound fuller.

Definitely not the most ambitious use of the technique, but I think it allows the composition of stuff to build in a more natural way rather than it feeling more structured (i.e. bar 4 is when the delay starts, bar 12 is when this next part comes in, etc.)

1 Like

So essential, thanks.

The “what’s this do?” = the nucleus about which the crystal forms.

1 Like

I was taught this concept by @CarlMikaelBjork while watching his video on the Black Wavetable VCO.

Now I will often times have the VCO sit in C (or some other note) the entire time and will sequence the wavetable (like in Carl’s video), or the filter, or the resonance, or the envelope shape (using a make noise maths), or even the delay amount for each note. I absolutely love sequencing things other than the V/Oct and I feel it really lends a great variation of sound.

Good tip!

1 Like

I’m going to take you more literally than the other posts- hoping maybe this will give you an idea.
I have a VERY simple recipe for this, and I built every module involved from a kit(which doesn’t matter to anyone but me, but whatever!) .
I use 3 modules:
Bastl Skis (a dual VCA with an envelope gen)
MusicThing Modular RadioMusic(Look this module up, I think you’d love it)
MusicThing Modular Turing Machine(generates timed randomness)

I feed a clock signal into the Turing machine,
I then take the trigger output of the Turing machine and connect that to the trig input on Skis & the reset input on the RadioMusic
I take the random signal out of Turing machine and connect that to the Station input of the RadioMusic

The result is a percussive, random audio, selected from an SD card. I saw “random”, but it’s the result of years of curation. I typically mix this way low against drums in techno - JUST enough to flavor a kick drum, for instance. RadioMusic is a spectacular module for this. I have mine filled with Cassette recordings of a swamp I made as a 15 year old, books on tape, recordings of radio, sound FX records, acapellas, field recordings, etc etc. 16GB of stuff that I have been collecting since 1999!


To build on what @deltasleep suggests: If you want something less percussive than that, putting a Morphagene after the RadioMusic will do very similar stuff with a smoother texture if you want something smooshier.