argh: no powers of 2 are divisible by 3 huh.....

(ducks.)
m.

MATH.

http://home.comcast.net/~nspil/Mathnet1.jpg

Anyway, I am trying to realign tempo changes on 32 notes but this sucks if I have fucking triplets.

xpost

1) round the non-two to the nearest multiple of two. approximate. i bet this sounds like shit at lower tempos.

2) real-time shift the sequencer pattern

3) use a short circuit step count sort of thing... so instead of on the 32nd note going back to the 1st note, let them set the step to go back to the first... like on the 30th note, or the 17th. etc...

?
m.

how to fit 33 notes:

123456789012345678901234567890123

into 32 slots:

12345678901234567890123456789012
a..a..a..a..a..a..a..a..a..a..a.

and retain a.. without fudgin the count.

solution 2 is really tricky and requires real-time interaction...

a..a..a..a..a..a..a..a..a..a..a.

shifted right on the next to last note after "a." becomes:

.a..a..a..a..a..a..a..a..a..a..a

and the note played gives you the extra "." you need to space the notes. it's an drum machine trick i used to used to get really whacked out shit.

next time you have to again, wait after the last "a" and the first "." etc etc... hard to be consistent but... music ain't easy to play.

the 3rd solution makes more sense because:

12345678901234567890123456789012
a..a..a..a..a..a..a..a..a..a..a.

set to 30 notes length becomes

123456789012345678901234567890
a..a..a..a..a..a..a..a..a..a..

and there's no finger gymnastic insanity at all.

but perhaps i'm totally confused about the problem.
m.

ah... like you said Hurting... probably a wrong approach anyway.
m.

Yea, if I moved the funademental beat clock to like 128 or 64 notes I think this would be ok. When I am less hungover I will do a little math and see how far out of alignment this would be at various tempos

m.

Quarter notes are 2/3s of a second long. A measure is 8/3 seconds. A 3rd of a measure is 8/9 or ~889 ms long. 889 / 75 = 11 (with remainder).

11*75=825
12*75=900

hmmmm... I think I might want a higher resolution.... Maybe I need a 64th note clock and play notes divisble by 3 on the nearest 64th note.

75 * 10 = 750
75 * 11 = 825

75 * 21 = 1575
75 * 22 = 1650

So, triplet tones on the 11th and 21st note of the clock are only off by 25 ms, or one-third of a 32rd note. Going to 64 notes would make the resolution twice as good, with only 10 ms error or so, which should be fine.

Did I mess up that part?

Yea 10 ms error is no big deal since I'm sure there is at least that much error between my midi device and the sound on my computer

sounds like one of those things you just have to try. and then search for every optimization you can wriggle if you're coming up short.
m.

Then again, PD might not use interrupts on intervals so small that the context switch would be dangerous and instead poll a timer register while busywaiting.