Timing of MIDI events - wavering accuracy

DuskDave

Hello all, first post here... and hoping someone can help.
I am having varied results using midi@mm:ss.ttt commands and I need to know if there is a better way to fire events at EXACTLY the right time.
I use a lot of (MIDI-triggered) looping in my act and before I bought ST3 I used a sequencer for drum beats and to fire MIDI commands: this meant that the looper kicked in and out precisely on the downbeat of the bar or whenever I wanted it to.
So, I created a calculator to turn the bar:beat:tick values from the sequencer into a timestamp for a given BPM and it seems to work OK for short loops here and there. However, for songs where a short chord sequence is repeated throughout the song (e.g. Mavericks - Dance the Night Away; Happy Mondays - Kinky Afro) I have found the timing of the loop begins to drift but, worse than that, the drifting is not consistent: I can play the song again straight after and where it lagged before, it now starts to lead... or it can stay useably in-time for 4 minutes on one play only to drift noticeably by the first chorus the next time.
At this point the best I can do is to stop and restart the loop much more frequently within the song but this is already making my lyrics file a mess to read and is more of a sticking plaster rather than a solution

Can anyone explain how the timing is tied to the MP3 files? It seems like the clock sometimes starts a little before the file starts to play which causes everything to warp.
Is there a better way using MIDI clock? But how could that sync with the MP3 player any better than the system clock?

I searched long and hard to find an app to do this and was told over and over that Apple devices had better sync between clock and media player than Android does. I don't mind if it is always wrong by the same amount because I can correct for that but this random behaviour is of little use.

Any advice would be gratefully received.

peter

The way MIDI is implemented in Stage Traxx leads to a bit a variability when sending MIDI events. It usually drifts around +/- 5 milliseconds.

DuskDave

peter
Hi Peter,
OK, thanks. If it is a design fact rather than a design flaw then I'll figure out some way to work around it.

DuskDave

peter One more question about the wavering. Does this mean that MIDI events can be triggered out of sequence? If I have...
[midi@00:10.500: ... event 'A' ]
[midi@00:10.505: ... event 'B' ]
Given the +/- 5ms could these two events be processed such that event 'B' happens before event 'A'?

Also, while we are on the subject when I write...
[midi@mm:ss.ttt: A, B, C, D ]
am I right to assume that the commands will be processed in order A, B, C, D?

Thanks.

DuskDave

So, essentially, I've had to rework the song so that I now record a 32-bar loop rather than a 2-bar loop but that seems to mitigate most of the effects of the time drifting.

I did try stopping and restarting the shorter loop multiple times during the song but that just caused the playback of the loop to not fire at all on some occasions. If only I had more time to investigate but I just have to work around it for now.

peter

No, all events are fired in the order you define via timestamps. What is possible though is if the timer for firing events is a bit late, it might fire B immediately after A.

And yes, events concatenated in one midi command are sent in the order you define them.