I'm pretty lucky because the UA-5's that I currently use have crystal oscillators that are pretty close to each other as long as they are in good thermal contact with each other while the recordings are being made.
How can one tell how close the oscillators are? Do you just record the same source and see how out of skew they are, or is there another way of calculating this?
Yes, just record the digital output of both boxes for an hour, line them up with each other at the beginning of the recording in your favorite editor (I use Audacity) and see how much they've drifted apart at the end of the recording. That way you know how many milliseconds per hour of timing skew you will see.
I don't like to ever see more than about 4 ms of skew between two sources in the mix at any given spot. Some people claim you can't hear the skew until it's over 20 ms, but I don't believe that. At 4 ms of skew, you can get notches in your spectrum at odd multiples of 125 Hz, so even that can be noticeable. At 20 ms of skew, you can get notches in your spectrum at odd multiples of 25 Hz, so the notches are much more tightly packed. If you don't want the potential for any notches below 15 kHz, you'd have to line them up to within about 1 1/2 samples at 44100 kHz or 3 samples at 96 kHz. So, if you have any skew at all, there is the potential for it to be audible, especially if the same sound is present in both recordings at the same level and the mix is equal parts from each source.
I got really burned one time when I tried to mix a PCM-M1 recording (from mics) and a DA-20 recording (from SBD) in post. There was so much timing skew that I had to make edits every 6 seconds or so through a 2+ hour show. The result was a mix with the sources within about 4 ms of each other, but it took hours and hours to do the editing. I think I had almost 25 hours in the project.
If you do crack this i'd love to see your solution although it sounds a wee bit harder than a digimod.
It sounds like you have run two UA-5's before, are you going into 1 laptop? and if so what software or are you using two recorders?
If I ever get this done, I'll definitely start a thread about it.
When I run two UA-5's I'm just recording to NJB3's, one per UA-5. Then I line up the recordings in post. I don't currently have a way to do 24/96 in the field, although I've done it into my desktop machine at home. I've never tried two UA-5's over USB at once, but I wouldn't be surprised if the drivers will only support one at a time per computer. I'm betting that you'd need two computers if you really want to do 24/96.