the problem is with the accuracy of the clocks. Apogee is known for their rock solid clocks, the reason for this is they do not drift and are locked to a crystal. Less expensive a/d's like the one in the JB3 and M1/D100, etc. don't have crystal lock and might run at 44100hz for a while and then drift to 44106hz and then back the other way. This is why apogee sells studios units such as the "Big Ben" so that all their gear (with w/c in) runs exactly the same on each track
That's very interesting. I've experienced this as well but had a different theory...
We generally like to synch sources on some kind of sharp peak. Many of these peaks are direct sound, reflected sound (delayed) and p.a. sound (direct and reflected)
Keep in mind, when you're talking about a few milliseconds, you're talking about a few feet difference in location. The drummer being 6 feet behind the guitar - there's 6 ms. Both come out of the p.a. at the same time but the direct sound from the drums/guitar amp could offer different 'match points' for 2 sources. I think many house p.a. systems do introduce a slight delay so the drummer's natural sound (behind the p.a.) has a chance to 'catch up' to the p.a.
Reflected sound could also explain it - I've noticed on some drum stuff that you can actually see a 'double image' in the wave form - the direct sound followed by the lower amplitude reflected sound (you have to be seriously zoomed in to see this). Again, knowing that every extra foot of travel will result in a 1ms delay, it doesn't take much to create a few ms gap.
I wonder what would happen if we only looked at the same instrument for time alignment i.e. check the hi-hat at several points in the recording.
So anyways...that was my theory. I haven't done the above where you'd only look at the same instrument at various points in the recording. Didn't really think of it until I was writing this.
I didn't obsess over the variations because my ears seemed to tell me that +/- 5ms was undetectable when mixing 2
different sources. Instead, I just choose a happy middle ground and as long as I wasn't out by more than a couple of milliseconds, all was good. The reason I say
different sources - for fun - try multitracking the
same source twice and introduce a 5 ms delay - it gets really phasey and weird.
I suppose the other way to find out is to have 2 devices that can sync (either master/slave or both to another clock) and then see if the same thing happens - if so, then the above explanation seems reasonable.
Anyone have any real live experience with synched clocks? I was actually gonna ask about this anyways.