User Manual

CHAPTER 31. MIDI FACT SHEET 504
31.5 Tests and Results
Our procedure for testing the timing of incoming MIDI events is represented in the following
diagram:
MIDI Source
MIDI
Splitter
MIDI-to-Audio
Converter
Live
MIDI Clip
Audio Clip
MIDI Input Test
Conguration.
The output of a MIDI Source (a keyboard or other DAW playing long sequences of random
MIDI events) is fed to a zero-latency hardware MIDI Splitter. One portion of the splitter's
output is recorded into a new MIDI clip in Live. The other portion is fed to a MIDI-to-Audio
Converter. This device converts the electrical signal from the MIDI source into simple audio
noise. Because the device does not interpret the MIDI data, it performs this conversion with
zero-latency. The converter's output is then recorded into a new audio clip in Live. In an
ideal system, each event in the MIDI clip would occur simultaneously with the corresponding
event in the audio clip. Thus the difference in timing between the MIDI and audio events in
the two clips can be measured to determine Live's accuracy.
In order to assess MIDI performance under a variety of conditions, we ran the tests with
three different audio/MIDI combo interfaces at different price points, all from well-known
manufacturers. We will refer to these interfaces as A, B and C. All tests were performed with
a CPU load of approximately 50% on both OS X and Windows machines, at both 44.1 and
96 kHz and at three different audio buffer sizes, for a total of 36 discrete test congurations.
Windows:
 Interface A: The maximum jitter was +/- 4 ms, with the majority of the jitter occurring
at +/- 1 ms.
 Interface B: For most of the tests, the maximum jitter was +/- 3 or 4 ms. At 96 kHz
and 1024 sample buffer, there were a small number of events with +/- 5 ms of jitter.