February 8, 2008


by Chris Randall

That's "What would Flava Flav do?" if you were curious. There's an interesting article here wherein tests were made of the relative accuracy of various drum machines, time-wise. Not surprisingly, the MPC3000 is tight as a gnat's ass. Equally unsurprising is that the MachineDrum fares quite poorly. Interestingly, their tests showed almost exactly the same results as my own (~100 samples via MIDI, and the dreaded 128 sample hiccup via the internal clock.)

Anyway, some interesting info there, if you're the sort that worries about such things. I'm not, personally. That's what "Quantize Audio" was made for.



Page 1 of 3

Feb.08.2008 @ 7:39 AM
using my meat matrix and flailing limbs I cannot play with an accuracy of 2ms, so I feel bad about criticizing the robots for sloppy timing.

some real pearls of Wisdom from Mr Flav
link [en.wikiquote.or...]">link [en.wikiquote.or...]

timeless gems


Feb.08.2008 @ 8:54 AM
oddly enough I was thinking about how to translate "Cold Lampin With Flava Flav" into Japanese instead of working today...

Feb.08.2008 @ 10:07 AM
holy "text-align:center;" batman.

Feb.08.2008 @ 11:32 AM
Adam Schabtach
It's kind of amusing that the TR-808 and the MachineDrum get the same overall rating, and yet I don't recall hearing many people complain about the 808's timing.



Feb.08.2008 @ 11:55 AM
yeah, i have no idea what to think about super-anal midi blah blah timing pseudoissues anymore. however, that sync box he makes is quite a handsome device.

Feb.08.2008 @ 1:02 PM
Midi jitter is the new loop-back latency.

Feb.08.2008 @ 1:08 PM
That dude's test is completely flawled.. a lab assignment with that kind of methodology in university engineering classes would have got my a nice "F". Not only is his test completely whack, but he incorrectly uses the term "variance" (wiki it) and then proceeds to use the absolute offset as the point of comparison. He should take a Stats course before he post his "scientific" tests.

The test is fundamentally wrong because he's measuring the accuracy of a clock against a moving target.. itself. Assume the first beat is in perfect time on a grid (because he makes it that way.) The first interval to the next beat will be accurate, whether it's late or early, but the third interval will be between two points which may or may not be in time. If one is early and the other is late, the interval is artificially too small. Vice versa, is one is late and the second is early, the interval is too large.

At the very least he should be comparing the difference between each measured interval and the expected result (120bpm at 44.1khz.. 22050 samples.) In the case of the MD he has numbers like 22018, 22050 and 22114 samples. The maximum deviation is 22114-22050 or 64 samples (1.4 ms). If you take the mean from his data you get 22058 samples and on AVERAGE the timing is 22058-22050 or only 8 samples early! If you shift the audio file over so that the average is "zeroed out" and aligned with the grid then the "true" max value is 64-8=56 samples or 1.2 ms.. almost half of what he said it was.

You can do the same thing by recording the audio, measuring the distance from each beat to a grid line (alignment with the grid is irrelevant) and then subtracting the average value from each data point.

The MD is obviously a little wonky but I think it's unfair to say it's as bad as they make it out to be. One interesting thing is that it is extremely "precise" in it's inaccurate timing. Meaning the inaccuracies seem to occur in a repeating pattern. Humans are not adverse to weird timing as long as it's predictable.

Incidentally my own test gave me a max value of 48 samples.. very close to the 56 samples I calculated from his data.

I also don't really care because the MD is full of techno.


Feb.08.2008 @ 1:14 PM
"I also don't really care because the MD is full of techno."

Wow, then i suppose i must have a broken machine. It seems to be making electro, hiphop, and it even accompanied some rock recently. probably should send it back for techno repairs!



Feb.08.2008 @ 2:32 PM
But where's the "boat anchor" category winner?? The machines still have better timing then I could ever do manually.
I think the fact that my MD can do 32 various parameter locks for each instrument on each step is a fair trade for slightly "off" timing. I don't even think my brain can keep that kind of accuracy and allow me to walk at an even clip for any extendend amount of time.

Feb.08.2008 @ 2:50 PM
Chris Randall
I find that the MD's timing (or lack thereof) can annoy me if I'm trying to add an MD part to an existing song. Then, I usually end up quantizing the audio. But if I build a song _starting_ with the MD, or if I'm using the MD and MM standalone, which is something I've been doing more often, I don't notice the timing issues at all. In actual fact, the drift makes it sound less metronomic to me.

Now, that said, one of the things I find incredibly annoying is that if I'm running the MD off MIDI clock, and I've got 32nd note stuff in the very last grid position, sometimes it will get cut off or not play because the sequencer skips to the beginning of the pattern early, or the pattern is playing long relative to Cubase or Live. This is the sort of thing that can really get under your skin, and I find it really irritating. Jimmy Chamberlain's magic 31/32 time signature, live and in person, living in my MachineDrum. (I suppose the same thing is why I could never stand the Smashing Pumpkins.)

Obviously, it depends on the tempo you're working at. At 100 it's not noticeable. At 180, it's huge. But that said, sorry, Gibbon, but it sounds to me like you're desperately trying to justify your $1400 purchase by taking it out on the rest of us. Almost all boutique gear has "quirks," and these are the MD's. Everyone is aware of it, and either it bothers you or it doesn't. It doesn't particularly bother me, because I can work around it. And the synth engines in the MD more than make up for any kind of oddities with the sequencer.



Page 1 of 3



Sorry, commenting is closed for this blog entry.