Topic: Two recorders - can't keep tracks synchronized - what to do?

[ScotK]

Hi

I'm relatively new to recording although i do have several shows under my belt at this point. Twice recently I've been
able to record on two units at the same time: a Neuros DAC and a HiMD recorder.  I load both recordings into audacity
and pick an obvious place near the beginning to sync things up and lo and behold, after a few minutes, the tracks aren't
in sync any more!  This last time, I synced about 6m in, but by the end of the recording (~100m), they're about 7s off again.
A sync holds reasonably true to the ear for a track or so, but not much longer. So a few questions:

1) Is one of my units keeping bad time?
2) Is there any way to fix the problem and/or find out which is right?
3) Is there any way to stretch or shrink time in one recording so I can get them
     synced throughout the entire recording.

In one case, I had a SBD and an AUD recording and I wanted to mix the two. This last time I had two audience
recordings, but still with different mics and positions, so I wanted to mix the tracks together to try to get the
optimal recording. What can I do?

Thanks,

scot

[69mako]

If it a gradual drift what you need to do it align the tracks at the beginnig and then stretch or shrink one of the recording to match the other.  I've done this a few times with my two JB3's.  This is exactly the reason I'm jumping on the r-4 team.  It might be a lot of work, but the results can be quite stunning.

Good luck.
Mako

[aegert]

What recorders?

Timecode?

A

[SmokinJoe]

Here is what I do... given then I'm using Linux and different software than you are, I'm not sure how you do it with your software, but the process is probably the same.

- Don't worry about which is "right" and which is "wrong".  Just pick one, and sync the other one to it.
- Assume the creep is linear... you are dealing with modern digital clocks, not analog tape.
- Pick a point at the beginning, align them...
- Before you get into the micro details, make sure the big picture is correct... you have both LEFTs pointing left, and you don't have a phase inversion...
- skip to the end and find out how much they are off... example: 7.123 seconds over the course of 101 minutes, 18 seconds...
- do the math: 7.123/(101*60 + 18) = 0.00117193 (my R-09 to my UA-5 used to run a difference of about .0004)
- In my case I would actually issue a linux command to fix the creep, in your software you probably have a feature for stretch, speed, etc.
- import the new file, align at the beginning, and then check it at the end.  My rule of thumb (no flame wars please), is that if they align within .007 seconds that's "close enough".
- If you are off by a tad more than the above tolerance, rather than repeat the entire stretch process, it may be easiest to go in the middle and snip out a few milliseconds on the long one... that way it aligns at beginning, middle, and end, and drifts a little in between. It's not an exact science, but if you have to do this at 10 spots, that gets tedious.

After doing this a dozen times... I bought an R-4. 

[goodcooker]

The only way I've been able to deal with non linear drift from two different digital sources is to sync them up song by song. It's a total PITA but it works.

I also bought an R4 to rid myself of matrixing problems.

[Church-Audio]

Hi

I'm relatively new to recording although i do have several shows under my belt at this point. Twice recently I've been
able to record on two units at the same time: a Neuros DAC and a HiMD recorder.  I load both recordings into audacity
and pick an obvious place near the beginning to sync things up and lo and behold, after a few minutes, the tracks aren't
in sync any more!  This last time, I synced about 6m in, but by the end of the recording (~100m), they're about 7s off again.
A sync holds reasonably true to the ear for a track or so, but not much longer. So a few questions:

1) Is one of my units keeping bad time?
2) Is there any way to fix the problem and/or find out which is right?
3) Is there any way to stretch or shrink time in one recording so I can get them
     synced throughout the entire recording.

In one case, I had a SBD and an AUD recording and I wanted to mix the two. This last time I had two audience
recordings, but still with different mics and positions, so I wanted to mix the tracks together to try to get the
optimal recording. What can I do?

Thanks,

scot

The Neuros has horrible drift I would not be surprised if it was more then 3% I would stick to two of the same recorders if I was going to try this. I dont think your going to be able to get it perfect. No mater what software you use.

[SparkE!]

The easiest way to do it is to play back one of the recordings from the unit that originally made the recording and record it again on the other recorder.  That way, you have essentially recorded both versions on the same recorder.  I know it's not exactly great re-recording a recording, but you generally don't lose more than about 3 dB in S/N and it's a HELL of a lot easier than stretching portions of one recording to match the other. 

I've done the selective stretching thing before in order to match a PCM-M1 recording to a DA-20 recording and I had to break the recording up into 6 second intervals, stretching about 3.15 seconds out of every 6 second interval in order to get the recordings to align with each other within 5 ms over a 2 hour show.  What a pain that was!  It took me nearly 20 hours to do all of the required editing.  Never again!  From now on, I'll just re-record one of the recordings.

By the way, you have to match the ambient temperature to the temperature in the original recording environment.  Crystal oscillators are (slightly) temperature dependent and it actually matters when you're re-recording a 1+ hour show.

[taper420]

I don't think anyone's outright said why his recordings are out of sync in the first place, which by the look of his post, he doesn't understand.

Your two different recorders have two separate ADC's (analog to digital converters) that run on two different clocks. No two clocks can be exactly the same because we're talking about taking over 44000 "snapshots" per second (and that's only cd quality). With that amount of processing there is bound to be a randomness factor, even if you have two models of the same recorder (but as Chris says, having two of the same models will certainly help cut down on drift....but not eliminate it)

For this reason most pro equipment comes with a means of sending and/or receiving a signal from an external clock source. I believe (correct me if I'm wrong) wordclock is the most accurate type of sync. But syncing can be done through spid/f and AES/EBU digital signals as well. In the studio it is common to have a clock unit, which has the sole purpose of keeping time. But for us in the field we will usually use one unit as the master clock and slave the other unit to it.

Example: I have 4 mics... two running into a V3, two running into an MOTU Traveler. I send the digital spid/f signal from the V3 into one MT recorder, and I send the AES/EBU digital signal to the Traveler. I slave the Traveler to the V3, and then send the mics going into the traveler out of the spid/f to a second MT recorder. This should produce a separate recording on each, which can be synced later.

Of course this is all in theory... I haven't tried it. As much in this thread have commented... there's an easier way around all the headache. A multitrack recorder. Syncing in post can give okay results, but it takes a lot of tweaking, and you will invariably end up with some phase cancelations here and there. Also keep in mind that certain digital recorders resample their digital input. They won't slave to an external clock. One of these such units is a Hi/Md recorder.

Good luck. For future recording I would recommend honing your stereo skills, and when you can afford it, go with a multitrack solution. Allot of the time, a stereo recording will sound as good if not better than anything you will get by mixing two sources.

Edit: and ... Welcome to Tapers Section! Feel free to ask more questions as they come to you. There is also a huge library of information already here. Many of what has been said can easily be found with a quick search. You've chosen a great hobby, and it sounds like you might even have a bit of the bug already. I hope you have a good job

[ScotK]

Thanks everyone. These responses help a lot.

To answer some questions/ask a few more.

1) The recorders are a Neuros DAC and a Sony MZ-M100.
2) It sounds like drift between   different recorders is pretty common, although maybe is
     less problematic with better units. Would I have similar trouble, say with two R-09s, for
     example?
3) I am using Linux, so if I could find out what tool you use for the time stretch, SmokinJoe, that
    would be great.  I'm currently using audacity and don't see any way to time stretch, just time shift,
    but it's possible I'm not finding the right option.  -[edit: Ahh, change tempo? trying that now...]
4) The main reason I sometimes use two recorders is I often find one gets screwed up for some
     reason or another.  Two gives me some better odds of getting something useful.  Furthermore, when
     I am lucky enough to get a soundboard feed, having an audience recording is nice to add to the mix
     when there's good audience participation or something.
5) I like the Neuros because I don't have to worry about storage space, but since I do have a reasonably
    good job, I have been looking at the R-09 and that new Tascam unit as possibilities.  Although, I also suspect
    I may benefit more from better mics than a better recorder - not sure yet.
6) By R4  - you mean the Edirol R4 which is being proposed since it is a 4-track device? That still wouldn't
    solve my redundancy aim or help at SBDs if I can't also put my mics nearby, though.

thanks again everyone!
scot

[DSatz]

> 1) Is one of my units keeping bad time?

All recorders "keep bad time" to some degree. The only question is whether their "bad time" is good enough to use or not.

When you're merely listening to a recording that you've made, its absolute timebase accuracy is a problem only if someone actually hears a problem (i.e. if it seems off-pitch or off-speed). You could probably tolerate, say, 0.1% error or more. But to sync up two independent recordings, you'd need a closer tolerance than that, particularly if the music continues without interruption for some time.

When multiple digital audio recordings are made for the purpose of being synced together later on, the recorders themselves should be synced while the recording is being made. That is, they should both be driven by one common clock signal, which can either come from one of the recorders or from an external (third) source. Most recorders that have digital inputs will track (sync themselves automatically to) the timing and sampling rate of the incoming data stream. Then your "only" problem is how to line up the start of the two recordings in your editing software.


> 2) Is there any way to fix the problem and/or find out which is right?

Yes to both--but curiously, the second part may be more difficult and less useful than the first.


> 3) Is there any way to stretch or shrink time in one recording so I can get them synced throughout the entire recording.

Sure--you can resample a recording in most editing software to adjust its sampling rate slightly, so that the two recordings have equal length. But that is a very tiresome repair/cleanup type of job, and it will be effective only to the degree that both recorders' internal clocks kept a constant rate throughout the entire recording (e.g. not speeding up or slowing down as they warmed up--which isn't normally a problem, but I'm mentioning it just to clarify the concept).

If you're going to do this sort of thing on a regular basis, you will almost certainly want to find a better way--either syncing the two recorders to a common word clock while recording, or using a multi-track recording arrangement of some kind. For four-channel recording I use a laptop computer and an outboard preamp and adapter, but that's not a "stealthable" setup unless you're Rubeus Hagrid.

--best regards

[SmokinJoe]

If you are using linux... Now I can help you.
sox  original.wav  newfile.wav  speed  1.00117193

Here are my notes from when I figured this out.

to answer your other questions...
- I once synced 2 recordings made with 2 separate Edirol UA-5's.  They we right on.  I suspect any 2 matching A/Ds work well.
- I have long cables with my R-4, so I can run my mics FOB and keep the recorder close to the SBD.
- BTW, even with an R-4 the aud recording will be a few milliseconds late, because it takes about 1ms/foot for the sound to travel from the stacks to the mic (speed of sound is on the order of 1000ft/second).  But this is simple... whack about .015 seconds off the front of the AUD and you are done.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Linux Matrix micro-HOWTO

Trim the front of both in audacity:
- command: "audacity sbd.wav &"
- Once that loads "project > Import Audio", select aud.wav, and let it load.
- Find some particular feature on both tapes, like a particular drum beat when they are warming up.
- Zoom right in on this feature, and click right where they peak
- Delete everything before that... "Edit > Select > Start to Cursor" <delete key>
- repeat for the other source.
- Pan Upper source Left, Lower source Right and play some of it.  It  should sound in sync, at least at the beginning.
- Pan upper and lower back to middle (if you don't when you save them it will save them like that).
- Save the two sources seperately  "File > Export Selected as Wav..."
- fast forward to the end.
- find some identical feature again.  Are they lined up?  No??? Damn!
- Figure out how much they are off... carefully select the peaks.
  Hint: To select and area in the upper and lower tracks, click in the upper half, drag down to the lower half, select the end.
- How big is the gap?  In my case is was 0.3 seconds at 1 hour 55 minutes, 9.5 seconds.
- Whichever source is longer... i.e. the drum beat extends to the right, close that one.

Shrink the longer track:
- command: "xcalc" (or use a hand calculator)
- math: 1 hour 55 minutes, 9.5 seconds = 115 minutes * 60 sec/min + 9.5 seconds = 6909.5 seconds
-       Since tracks are off by 0.3 seconds over that time... ratio= 1 + 0.3/6909.5 = 1.0000434
- command: "sox <trimmed filename> <shortened trimmed filename> speed 1.0000434"
  (wait a few minutes)
  This shrinks the whole thing a little bit in linear fashion.
  The clocks on my unsynced recorders aren't tthe same, but I assume   they are steady, unlike old analog tape.
- command: "shntool len *.wav"  look at the old and new file lengthes.
  The newer one should be shorter. Don't worry that it's 0.5 seconds shorter instead of 0.3... that's round off error in shntool, and at least it is in the right direction.

Back to Audacity:
- "project > Import Audio", select new shortened file, and let it load.
- Pan upper left, lower right, and listen. If you did everything right they will align perfectly through the whole file.
- Put the Pan Left/Right back in the middle.
- Adjust levels with "Effect > Normalize" and/or "Effect > Amplify".
  You don't want both sources going to peak zero, or you will get severe clipping when combined.
  Ideally normalize to -3 db would be perfect, because -3 db x 2 sources combined = 0. 
  Trust me on that. :-)
- "File > Export as wav...", it warns "I'm going to mix these down into one..."  Excellent! 
   Just what we want.

- Listen to it via XMMS or whatever.  Track out out like you would anything else.

[ScotK]

Thanks, Smokin' some great tips there!
I've already had a few stabs at this particular case and yeah, it's a bit of a pain, but it's coming along.

-s

[SmokinJoe]

It's one of those things where after you do it a couple of times it's not a big deal.