Further to my review of the Yamaha Motif Rack earlier this year [see SOS June 2003], I have, as promised, some more findings regarding the latest version of the Motif Rack operating system.
As you may remember, I experienced poor timing when using the Motif Rack multitimbrally, finding that even quite simple multitimbral song sequences of five or six parts were being played back with inconsistent timing. As you'd expect, the problem was not as noticeable when using the Motif Rack in isolation, but if you gave the Motif Rack some parts of a song to play and had others played by different modules, the timing discrepencies between the Motif Rack's parts and those played by the other synths was all too obvious. The problem persisted, even when SOS arranged for the original review unit to be exchanged for a new one, and when Yamaha-Kemble in Milton Keynes found that they were able to reproduce what I'd found on their units, we had to conclude that there was indeed a problem.
As I said in my original review, this significantly marred what was for me an otherwise great product, and I was delighted to hear that Yamaha were trying to find a solution implementable in software for inclusion in the next OS upgrade of the Motif Rack. Thanks to Yamaha-Kemble, I was able to hang on to the review Motif Rack, so that I could re-test it when the upgrade arrived. This took a little longer than anticipated, but when it shipped in late July I loaded the upgrade and tested the Motif Rack in the same way as before.
It's worth recapping how I did this, as I used the same method this time so that the results could be fairly compared. Using a six-part sequence in Sonar (five fairly basic song parts and one rhythm part), I sent the MIDI parts to both the Motif Rack and another synth, then recorded the output of both synths directly back into Sonar as audio. Comparing the waveforms of the drum parts I had recorded as audio at a high zoom factor in Sonar enabled me to determine precisely how much later the audio from the Motif Rack was triggering. As explained in the review, I obtained a variety of results from pretty good to audibly out of time depending on whether I gave precedence to the rhythmic part in the test sequence by placing it on the first track (sequencers usually prioritise tracks in numerical order, outputting the data they find in the topmost track first). Furthermore, the delay was not consistent, meaning that the timing problem couldn't simply be solved by recording the parts and then offsetting them by a fixed amount to bring them back into time.
Of course, no MIDI device responds instantaneously, and all have a built-in latency of some kind, but this is usually well below the perception threshold of most of us. My previous tests with the Motif Rack and other test synths, however, had produced average delays of between 400 and 1400 samples. Even at its shortest, this delay translates to nine milliseconds' delay at 44.1kHz, which is certainly perceptible, and at the other extreme of 1400 samples, the delay translates to 40ms at 44.1kHz, which is more than audible, and is, of course, how I noticed the problem in the first place.
So how did it fare this time? Simply playing back the test sequence with the new OS installed, I immediately noticed an audible improvement. On closer examination, however, I found that the timing issues were still present, just to a lesser extent. With priority given to the drum track as described above, the relative delay on the Motif Rack's drum part when compared with that of other test synths averaged 100 samples, or just over two milliseconds at 44.1kHz, which is so low as to make no audible difference (on some test runs, the delay was as low as 40 samples, or under a millisecond at 44.1kHz!). However, if I did not prioritise the rhythm parts as described, and left them on the default Track 10 (set up as the default to match MIDI channel 10, to which Motif Rack rhythm parts default) the results were not as encouraging. On the better passes, the delay was around 200 samples (or 4.5ms), which is again acceptable, and imperceptible to all but the most rhythm-attuned ears. However, on the worst passes, I measured the delay at 1200ms, or 27ms, which is clearly audible. The average delay I measured for rhythm parts on Track 10 in this way came to 1000ms, or nearly 23ms at 44.1kHz — the kind of delay you'd baulk at if it was down to driver latency.
Moreover, it's worth restating that these were average delays — once again, the delay was not consistent throughout, thus preventing you from correcting the problem with a timing offset in your sequencer after recording. As previously, this was particularly noticeable when assigning some parts in a song to the Motif Rack, and others to different synths. If some of your sounds were coming from virtual instruments, the delay might prove particularly noticeable, as of course the response of software instruments to incoming MIDI is virtually instantaneous.
The bottom line, then, is that as before, the timing issues on the Motif Rack still produce noticeable effects, though the problem is better than it was at the time of my original review. You can mitigate these effects by prioritising rhythm-sensitive parts and placing them on the lowest-numbered tracks of your sequencer, and relegating less timing-specific parts such as pads to the lower tracks (arguably good sequencing practice anyway). If you do this, and particularly if the Motif Rack is the only sound source playing back your compositions, the chances are that the effects of the timing problem will be so minimised as to be imperceptible. However, given that a big part of the Motif Rack's appeal is as a keyboardless synth expander module, and that it is therefore likely to be used in conjuction with other sound sources, I would say that the Motif Rack's timing problems are still significant. I therefore stand by the caveats I gave to prospective buyers in my review.
Managing Editor Matt Bell adds: According to Yamaha representatives, no further work is currently planned on the MIDI timing aspects of the Motif Rack's synth engine, so this matter may never be fully resolved. Should this situation change, we will of course let everyone know through the News pages of SOS.