[LAD] MIDI jitter - was: automation on Linux (modular approach)

[Ralf Mardorf]

fons at kokkinizita.net wrote:
> On Wed, Mar 24, 2010 at 11:06:43AM +0100, Ralf Mardorf wrote:
>
>   
>> At the 
>> moment Linux on my computer and on computers of around 30 other people I 
>> know can't use hardware MIDI equipment because of MIDI jitter. On the 
>> same machines there is less jitter for Windows, so using Windows would 
>> solve this problem for most of them.
>>     
>
> This made me curious, and as I rarely use MIDI (just
> to play piano using Linuxsampler) I wrote two trivial
> test programs usin MIDI over Jack.
>
> The first will output a note on or off every 10 ms.
> The second just receives midi a prints the time
> (number of frames) since the last event for each
> one received.
>
> When connected directly via Jack the result is a
> boring series of '480', one event each 10 ms.
>
> When connected via a loopback on a HW interface
> I expected the worst case to be events quantised
> to Jack period (256 frames). Actually it's 10 times
> worse - events are bunched into groups, one for
> every 10 periods. That 53 ms of jitter or if you
> are optimistic, +/- 26 ms. 
>
> The interface used is PCI based, no USB problems.
>
> What is going one here ? 
>
> Ciao,
>   

Hi Fons :)

could you please send me the test programs off-list. I made tests 
sending MIDI to a DX7. The DX7 generated an extremely short sinus 
impulse. I recorded it and examined the waveforms by Audacity. Btw. the 
graphics has access to the main memory, unfortunately it's a shared RAM, 
OTOH I used HPET so unwanted interrupts because of a shared RAM 
shouldn't be the cause, if I do understand the workings of HR timers 
correctly.

Cheers!
Ralf