Machine swing and odd meters

I’ve been thinking about machine swing (or shuffle) and how to deal with odd meters and I thought I might ask about it here in case anyone has any opinions.

I’m working on a sequencer. In each Pattern, each Track can be 1 to 8 bars long and each Bar can be 1 to 32 steps long.
Global swing can be introduced by turning the Swing knob.
Per-Track swing can be introduced by holding a “track” pad (or a number of “track” pads) and turning the Swing knob.

As is traditional, I’m conceptualising swing as an idea relating to pairs of steps. As the swing value increases, the first step in each pair gets a little longer and the second step in each pair gets a little shorter to compensate. The upshot of this is that each pair of steps (and thus the bar) retains the same total length.

This works fine when the bar lengths are even but if a bar is 9 steps long for example, that means it consists of 4 pairs of steps and one (lonely) unpaired step (see the diagram below). The 4 pairs can be swung as normal but my question is “what should happen to the length of the ninth step?”

Each track can potentially have a different length, so in order for them to stay in sync, it’s not possible to increase the length of this step, so at present I’m leaving it untouched. But maybe there’s another way to think about it?

Any thoughts appreciated!
(PS. I’ve done the programmer thing of numbering the steps from 0 in the diagram so hopefully that isn’t too confusing)


One way that could be useful is doing some sort of euclidean calculation based on the step length, then limit the calculation to have only groups of two or three steps per “chunk” of swing. You’ve already figured out some way of calculating swing based on groups of two steps, so you could work out a way to calculate swing for a group of three steps, then use a euclidean algorithm to fill in how to sequence based on groups of two and three.

So you could force patterns divisible by 2 to be made up of only two group swing, patterns divisible by 3 to be made up of only 3 group swing, then something like a 13 step pattern could be swing groups like 3+2+3+2+3 or 3+3+2+3+2, or however you’d set up that calculation to be made.


Thanks kbit. This is a useful way to think about it. I will definitely explore this.
I get the maths of this but I’m also wondering whether there is a basis for this in human-played music?
Might you have any examples of how a three step swing should/would feel?
In other words, which steps should be lengthened and which should be shortened?
Many thanks!

1 Like

I personally like the feeling of a three beat swing where beats two and three are a little late, then the one of the next group of three hits right on time for a quick snap.

Music that has a strong 3/2 polyrhythmic feel would be my listening suggestion from a less machine oriented perspective i.e. Afro-Cuban, Gnawa, Highlife, other West African music from Senegal, Nigeria, Ghana. This is one of my personal favorites:


This is good food for thought when it comes to subdivision.

Though I feel like most modern “mathy” music like Neely’s stuff is often felt pretty straight when it comes to breaking rhythm down into whole integers. Like his videos on quintuplet swing and what not, it’s a very useful way of conceptualizing how to displace beats but it still feels rigid, like the music is “off” in a very precise way, which is not usually how swing makes me move as a listener. Not meant as a criticism at all, just something I’ve thought about.


Oh wow. That’s some groove. Perfect. Thanks.


This is all super-useful Gahlord. I’m Irish, so I got the 9/8 ref in the Neely video. :wink:
I really appreciate these other references.

It seems to me that these kinds of rhythmic nuances can get lost in the machine context and I’m interested (if at all possible) in giving the user some way to enagage with them via an interface that doesn’t require them to fully understand all of this theory. For example, maybe there could be a way of scrolling through the “logical” 2 pulse/3 pulse combinations for a particular bar length?

I also have to admit I’m starting to see why this kind of stuff is not usually dealt with in (hardware) sequencer-land! Or maybe they are and I’ve just been looking in the wrong places?


This is low key blowing my mind.
My first inclination was “if there’s a pair, swing it. If not, keep it straight.”
That negates the option of a three-note grouping, which is doesn’t seem like a good solution, only a quick one.
What platform/UI are you planning to use? There are some really interesting possibilities if you allow the user to select different groupings; suddenly I’m thinking “what does 3+3+3+2 sound like against 2+3+2+2+2,” especially if it’s swung?? I’d love a simple tool to try that out, especially if it’s a front-facing feature, or not menu-divey.

1 Like

Another thing I’d suggest in this realm is listening to how live bands subgroup odd time signatures. A very common way is to break up an odd phrase in several nearly equal pieces and put the smallest piece last.

For example:
5 felt as 3+2
7 felt as 4+3
11 felt as 4+4+3 or 3+3+3+2

Regarding @wheelersounds thought of multiple swing patterns superimposed, something simple could be a “rotate” parameter like many euclidean pattern generator eurorack modules have, but in my experiments with this kind of thing using Teletype, the variations are quite subtle unless very heavy swing is used, which can quickly become “non-musical.”


in the karşılama meter/rhythm-family in Turkey—a ubiquitous “9/8” style meter that’s the basis of a lot of dance music—there will typically be four prominent accents per measure, and what is swung are 2 or three beat fills that precede those beats. In the basic form:
S - S - S - L - - (1 - 3 - 5 - 7 - - )
are “the beat”

S- xxS- xxS- xx-L- x- x- (one and-a two and-a three and-a four slap slap)
is the most simple embellishment, and for the “xx” pickups to the beat, the first of those “x” would begin just a bit later than you’d expect from a metronome. It might work out something like this, where numbers represent “percentages” of beat subdivisions:

S… -… x … x … S
27. 27. 23. 23. –

A big problem I’ve long seen with electronic instruments is with the mechanical sound of ratchet fills; they’re nothing like the flams and paradiddles (etc) you find in human-performed drum rhythms.


New Daedelus beat tape has some interesting polyrhythms done with computers and/or drum machines. It’s borderline experimental music. I was listening to it while cooking last night and my housemate was like “your sautee pan is making some funny noises tonight.”


So based on the original post and the responses, a few other questions/thoughts!

My understanding is that thinking of timing in terms of the kinds of rhythmic groupings discussed above is often about figuring out where to put accents i.e. where to play harder/louder for emphasis. If that’s the case, then are rhythmic cell note-groupings equivalent to swing? I guess that it might be that what we’re talking about relates to the larger concept of ‘feel’ maybe? Like ‘swing’ is a subset of ‘feel’?

As someone who is used to working with machines, I’m kind of haunted by @Gahlord 's comment:

It’s obviously the case that it’s a norm in sequencer design that all rhythmic events are generated in relation to a regular pulse. Is it the case that the notion of “the step” as a rhythmic unit is somehow problematic in that it suggests non-fractal rhythmic relationships where local variations are not related to global variations? I think that part of the reason that something like the Pulsar 23 feels like such a breath of fresh air is down to its potential to play with or ignore this (the regular pulse).

The sequencer I’m working on will stick with the regular pulse for now. This is partly because of the decision to allow the music-maker to pick arbitrary bar lengths/counts. Keeping to a regular pulse seems to me to be the only way to ensure that things remain synchronised. The sequencer will also implement a custom permutational algorithm, so that brings its own complexities. It’s also because I personally like that machine-swing sound.

One thought that does recur for me in relation to this though is captured by @kbit 's comment about ‘musicality’:

Regarding @wheelersounds thought of multiple swing patterns superimposed, something simple could be a “rotate” parameter like many euclidean pattern generator eurorack modules have, but in my experiments with this kind of thing using Teletype, the variations are quite subtle unless very heavy swing is used, which can quickly become “non-musical.”

@wheelersounds idea of “beating” (in the sense of beat-frequency) swing amounts or schemes against each other is something that really appeals to me and relates to @Gahlord 's comment about the way that Jazz musicians play off each other’s timings. I think those kinds of subtle temporal slippages and resolutions can be super-powerful from a compositional/feel perspective.

I know that this is deeply subjective but in the process of designing this device, I’ve been thinking about the penchant within the modular community for music that consists of multiple rhythmic/melodic loops of different step lengths and what the implications of introducing the (potentially complex) rhythmic offsets we’re discussing here into that mix might be. I wonder about the degree to which is makes sense ‘musically’?

I know that this risks opening a huge can of worms on a very subjective topic! I’m interested in music that grooves (i.e. that notion of entrainment) and music that does things to time perception. Ideal music for me exists in the overlapping of those two ideas.

That said, maybe there’s something to be said for just creating the tools to allow these ideas to be explored and letting the musicians figure it out for themselves?


Just to answer this separately:
the sequencer (called STREAMER) is a eurorack ecosystem. The paradigm is similar to teletype in that the core module acts as a USB host for a MIDI grid, in this case a Push 2 controller. The first protytpes arrived a couple of weeks ago, so we’re working on getting those up and running. We’re still a little way away from having something commercially available but I’m hoping that STREAMER addresses a whole bunch of things that I feel are important in terms of how to maximise flow-state for music-makers working with sequencers and modular. I’d like it to be something that allows people to make new kinds of music. We’ll see!


I don’t have one but the Eloquencer apparently allows some editing of ratchets I think? That said, it just allows you to mute certain ratchet elements. Is it the case that the mechanical sound you’re feeling is at least partly to do with a lack of per-note volume control? Playing all ratcheted notes at the same volume is not something human players typically do I guess?

i wouldn’t over think it?

break compound meters into couplets and triplets.

in each couplet / triplet, delay the weak beat(s)

in a triplet, may want to delay the 2nd beat only (or different amounts for 2 and 3)


Yes, volume can be one aspect - some of the “human ratchets” I’m thinking of might increase in volume over the course of the ratchet. Some stay the same as close as possible. But the timing this I was suggesting is more aligned with what @Gahlord said about laid back / in the pocket grooves in jazz. If instead of dividing a quarter note into four “16ths” represented by percentages:
23 23 27 27
as I suggested (with the played pickup just a bit quicker, giving a feel that the tempo of the whole part is faster than it actually is), you play it
27 27 23 23 . [or 27 26 24 23, good to experiment here]
you get a more laid back sense that the piece is more lazy/relaxed and slower tempo than it is. Both are musically meaningful… For “some” hypothetical rhythm sequencer it’d be awesome to spit a cv change and switch between ahead/on/behind the beat.

1 Like

Here’s another thought, by no means strictly technical but more of a feel thing.
Take a 1/4, 1/2 or 1/8 note clock however you chose to generate that and reset another faster square LFO to generate the subdivisions. Combining those clocks together can create all kinds of cool grooves. Another oldie but goodie and personal favorite of mine is the Pittsburgh Modular VILFO or more specifically the Chain Reactor (2 VILFO & 2 LFO) The “voltage influence” gives all sorts of weirdly organic feels. Subtle shifting of the wave shape gives even more variation of push and pull.