MIDI Timing

[liveinthemix]

quake2 napisat ću ti ne?to ?to sam pokazao na jednom od svojih predavanja...check this:In the past i've posted a few results of some midi timing tests i'd ran using different sequencers. I came to the conclusion that sequencing on a Mac using a Unitor 8 could be pretty good. However i've recently been further testing, but this time checking how multi midi out ports are addressed.
I had Hinton Instruments make me up a neat little box where i can plug two midi cables in, and then run two audio cables out so as to record midi data as audio pulses. As the unit is just wires working at the speed of light, no conversion time/latency is added. Just the same as making a midi plug to audio jack cable.
I reran my Unitor 8 test, sending from 8 midi tracks of 16ths in Logic 9, to the eight ports on the Unitor 8. I took audio recordings from ports 1 and 8, and then checked the difference between each track in the sample editor. The timing between port 1 and 8, was that port 8 was behind port 1 by around 7ms. If i muted tracks 2 to 7, thus no longer sending to ports 2 to 7 on the Unitor, the delay between ports 1 and 8 dropped to around 1ms. Therefore you could not compensate for the delay with a track parameter.
The Unitor 8 was crap then....
I then ran the same test with a Motu Express 128 USB interface, which I bought brand new on a gamble. I installed the latest Motu drivers, and ran the exact same test again. This time there was NO DELAY between midi ports 1 and 8 on the Motu. The audio recordings start waveform for midi note on data lined up perfectly, as did the second byte for the note off data. The same test was carried out in Live 8, and again the same results. Now i've established that the Motu sends midi note on/off data simultaneously across its midi ports, i'll be checking jitter. But i can say that so far early checks seem to indicate a value of around 1.5ms - 2ms worst case. Most of the time it's actually smaller than this.

To be fair i also ran the same test, checking for parallel transmission across midi ports on my Notator + Atari/ Log3/ Unitor2 setup. There two was delay between ports Atari main out A, and Unitor2 port F of around 2.5ms - 3ms. I checked other ports and found jitter to be around 1ms, while other ports would jitter by worst case 2.5ms.

I believe the result i've so far found for the Motu Express 128 are all down to the MTS drivers (Midi time stamping drivers) OSX core midi is meant to support this at OS level, but unless the Midi interface itself has this capability it'll just act like any other dumb MIDI device, and midi events will be sent out as quickly as they arrive and can be serviced.