Line input differences - are there any, or is it just A/D and level matching?

[Gutbucket]

This is correct.  It's the relative difference in any two clocks that determines how long it will take for the difference to become noticable.

Interstingly, if you playback two sources simultaneously independantly, using the same machines the recordings were made on, and manually sync the two at the begining, they will stay in that relative sync for a very, very long time. In that case the difference in timing between the two clocks is self-compensated for the most part since the same clock is being used for recording and playback, leaving only the absolute variation within each clock to cause things to drift, and that is miniscule.

[Jema]

Relative difference - lesson learned. Though I suppose you could talk about the stability of clocks in comparison to a perfect one. If the two imperfect sources both have clocks that are close to this, then it should sync a bit longer.

[morst]

Interstingly, if you playback two sources simultaneously independantly, using the same machines the recordings were made on, and manually sync the two at the begining, they will stay in that relative sync for a very, very long time. In that case the difference in timing between the two clocks is self-compensated for the most part since the same clock is being used for recording and playback, leaving only the absolute variation within each clock to cause things to drift, and that is miniscule.

If only the PCM-M10's analog output didn't sound like butt!?! 

[runonce]

This is correct.  It's the relative difference in any two clocks that determines how long it will take for the difference to become noticable.

Interstingly, if you playback two sources simultaneously independantly, using the same machines the recordings were made on, and manually sync the two at the begining, they will stay in that relative sync for a very, very long time. In that case the difference in timing between the two clocks is self-compensated for the most part since the same clock is being used for recording and playback, leaving only the absolute variation within each clock to cause things to drift, and that is miniscule.

I have to ask - how does one achieve that feat?

[Gutbucket]

I have to ask - how does one achieve that feat?

It's really for playback listening only. I don't use it for mixing the sources, but stretch and sync on the computer for that.

Make sure you are playing back the files on the same recorders that were used to record them. Start by picking a region with spoken word near the begining of the recording. Count-offs, sparce music like drum stick hits can also work, but spoken word is by far the best as it is easier to hear subtle delay differences as well as much easier to hear gross absolute position errors. Pause one recorder at the start of that.  Start the other playing and when that section rolls past un-pause the other.  Determine which is slightly ahead in time.  Use fast double clicks on the pause/play button of the machine farther ahead in time to delay that source by fraction of a second increments until the two are in close sync with no echo or smearing.  You'll hear it and learn to get good at it.  You may go a bit too far and have to switch to delaying the other, or rewind and re-try if you run out of speaking.  You'll get a feel for what the minimal play-pause-play delay time length is and will begin to know if you are close enough to hit the sync or if you'll need to try again.  If the two different sources are playing back on seperate speakers, it can help to get up and put your head between the two to help hear which is ahead and fine tune the sync.  It may help to unplug the feeds to the other speakers.  You may have one machine you can play-pause-play faster, with less incremental delay, so aim to begin with that one further ahead in time for better accuracy control.

Doing it requires concentration and a good ear, looks crazy and makes people smile to watch.  I've sucessfully sync'd 3 seperate playback decks this way for multi-channel surround playback which is a PITA, but syncing just two machines is pretty easy once you get the hang of it.  The trick is learning to hear minimal millisecond delays in the Hass time range that aren't long enough to be echos, and perfecting the very fast double jab on the play/pause button. Once sync'd that way the machines will remain in sinc until the file split, which usually seems to throw one deck off slightly even if playback across splits is supossedly 'seamless'.