Patternable tempo changes via `cps`

[yaxu]

There is cpm in core but it assumes a scheduler running at 1 cps.

It would be great if cps could be patternable, e.g.

cps(saw.range(0.9,1.8).slow(16))

or

sound("bd [bd bd]").cps(1,1.5)

[felixroos]

I wonder if the assumption of 1 cycle / second has any disadvantage? If not, we could implement cps like cpm

[yaxu]

Hm, I think having a global cycle rate that can be changed is pretty important. Personally I don't really like 1cps, the default in tidal is 0.5625.

[felixroos]

But anything against implementing it with slow, assuming a hard coded cps? I kind of like 1s because it's normalized

[yaxu]

slow doesn't hold state, so slow (saw.segment(8).range(1,2)) would miss a lot of events.

With something like: "a b".cps(1.5).every(3,rev)

The rev would not work on cycle boundaries.

More generally, when it comes to syncing between systems, it would be difficult to do that without a global clock.

[felixroos]

makes sense. so we need a global clock that dictates how fast the scheduler will run. also thought syncing would be great. .cps would then have a side effect to talk to the scheduler.

[yaxu]

SuperDirt likes to have some metadata - the current cps, the duration (as delta) of the event in seconds, and the event onset (wholeorpart.begin) in cycles (as cycle). So it could be good for the scheduler to provide this info (or a way to get it) to the callback.

[felixroos]

i made some improvements to the scheduler + sketched cps option DEMO

I am not sure if the behavior is right when I change the tempo: when pressing minus, the scheduling might repeat notes, if pressing plus, the scheduler might jump over notes. it feels like the scheduler has to use a time phase? is this how tidal does it too? this is the code: https://github.com/tidalcycles/strudel/blob/ea1ffca4ae430cc64a0be08dfc062204021c351a/packages/webaudio/scheduler.mjs

[felixroos]

tempo changes are now working smoothly (the above link will still show the old behavior). Updated demo

Interestingly, the implementation is quite similar to a digital oscillator, which also requires keeping the phase in state to avoid phase jumps / cracks when repitching: https://github.com/tidalcycles/strudel/blob/optimize-scheduler/packages/webaudio/scheduler.mjs#L17

The next step is changing cps from a hap. I think it will only be possible to change the cps once per slice (in the above example it's 10Hz)

[felixroos]

The next step is changing cps from a hap. I think it will only be possible to change the cps once per slice (in the above example it's 10Hz)

That's also working now not sure if this is similar how it works in tidal though

[yaxu]

This got lost along the way..

[felixroos]

moved from https://github.com/tidalcycles/strudel/issues/494:

global clock https://github.com/tidalcycles/strudel/issues/470 is now implemented in https://github.com/tidalcycles/strudel/pull/493 . example: https://strudel.tidalcycles.org/?9xvgxvOjkBR1

now the next step is making it work with patterns

[yaxu]

Good stuff. FWIW I needed to do a hard refresh for that to work.

[felixroos]

Good stuff. FWIW I needed to do a hard refresh for that to work.

sometimes the old version may be stuck in the cache.. ideally, the pwa service worker will update itself and reload the page when it's ready

[jarmitage]

small request: once this is fixed, if possible find a way to make a separate function that does what the current behaviour does, because it's quite cool!

[yaxu]

I noticed that passing a pattern to setcps("0.6") stops all sound and needs a hard refresh. No errors/warnings are printed to console.