Chel van Gennip wrote:

The topic has been here before, but syncing audio and video is an ongoing
battle. Recently I made some progress. Normally my targets are piano
video's. Piano video's are extremely sensitive for sync problems.

I recently replaced my Digital8 camcorder by a JVC Everio GZ-MC500
compact flash camcorder. I just don't like tapes.
The output of this camcorder is a mpeg file that is DVD compatible. What
is mo when played on a DVD player, the audio and video are in sync.
I've had a lot of problems with sync in the conversion to mpeg. Even
recordings with the original soudtrack showed drift after about 10
minutes. Those problems seems to be fixed.

The soudtrack of a camcorder often is not very usable, so I replace it
with a better recording of the event. Currently I am using a minidisk.

This week I made a recording of a Beethoven Sonata, about 25 minutes. As
the mpeg was in sync, I dumped the audio track of the mpeg file, and used
it as a reference track in my DAW. The minidisk file had a drift of about
20ms in this 25 minute recording, so I did not have to make changes.
In my DAW it was easy to synchonise the minidisk file with the dumped
audiotrack of the camcorder. I added the newly created audiotrack to the
mpeg file in Womble's video wizard and muted the original audio track.
The video track was not changed. I burned a DVD from this mpeg file and
the result was nicely in sync.


It was in synch over 25 minutes? I could be wrong, but I
think you're just lucky. There are two independent,
relatively cheap crystals in play there. If this is not
true, then please disregard the rest of this post - I
don't know how the minidisk could be synched to the camera.

FWIW, I've found the "chase synch" capability of the
little Fostex VF-160 class recorders very useful. It's
full 16 bit PCM, not ATRACS. It also actually
adjusts its speed to match the synch signal - varispeed
on an analog tape recorder causes it to shift pitch.

It's only good to +/- 5%, but that ougtha be plenty.

If the camera will emit SMPTE, you can convert that
the MTC. The Fostex will chase that nicely.

If it will not emit SMPTE, I'd try two
SMPTE-MTC/MTC-SMPTE converters ( like
the JLCooper PPS-2 ) .

Have one of the synch converters stripe to the audio
of the camera, and have the other "wye" cabled
to "listen" to this, driving the Fostex. The
Fostex takes a couple seconds ( up to five )
to lock, but the clock on the Fostex is correct
with the incoming synch stream after
that:

[pps2] +- [camera input]
|
+- [pps2] - [Fostex]

It'll be a twister game to start things up , but
it'll be in synch, once you get all the details
ironed out.

On playback, you'll need one PPS2 to convert the
SMPTE leaving the audio track on the camera for
the Fostex.


--
Les Cargill

Les Cargill wrote:
It was in synch over 25 minutes? I could be wrong, but I
think you're just lucky. There are two independent,
relatively cheap crystals in play there.

Les, I don't know why you're so surprised. It is very reasonable for two
"cheap crystals" to remain in sync for 25 minutes or longer. It doesn't
always happen, but it happens often enough to be almost expected.

I have been using digital watches for thirty years or so. They contain
cheap crystals, and usually are accurate to a minute or less per year.
That translates to about 3 milliseconds over half an hour. Considering
worst case of opposite drift, they should still be within 6 ms of each
other after half an hour. That's close enough for me.

I routinely shoot a half-hour concert with two DV cameras and a separate
audio recorder. In post, I sync them all up at the start, and check the
end. So far, I have never had to make any adjustments for drift.

Ed Anson wrote:

Les Cargill wrote:

It was in synch over 25 minutes? I could be wrong, but I
think you're just lucky. There are two independent,
relatively cheap crystals in play there.


Les, I don't know why you're so surprised. It is very reasonable for two
"cheap crystals" to remain in sync for 25 minutes or longer. It doesn't
always happen, but it happens often enough to be almost expected.


I'm not all that surprised. But this is happening by accident,
not by design. These clocks aren't that good.

I have been using digital watches for thirty years or so. They contain
cheap crystals, and usually are accurate to a minute or less per year.
That translates to about 3 milliseconds over half an hour. Considering
worst case of opposite drift, they should still be within 6 ms of each
other after half an hour. That's close enough for me.


I'd have to see what you're talking about to know - 6
milliseconds is an eternity in music. Then again, I
compensate down to half a millisecond for rendering
MIDI tracks. It's relatively easy to do.

Then again, I consistently see TV material that's that
far off or more. I've even seen movies released on DVD
that were way off - the lips and the voice had little
to do with each other. They were as much as two
frames apart.

I routinely shoot a half-hour concert with two DV cameras and a separate
audio recorder. In post, I sync them all up at the start, and check the
end. So far, I have never had to make any adjustments for drift.


That you know of

--
Les Cargill

Chel van Gennip wrote:

On Sat, 10 Sep 2005 21:06:12 +0200, Les Cargill wrote:


Chel van Gennip wrote:

..

This week I made a recording of a Beethoven Sonata, about 25 minutes.
As the mpeg was in sync, I dumped the audio track of the mpeg file, and
used it as a reference track in my DAW. The minidisk file had a drift
of about 20ms in this 25 minute recording, so I did not have to make


It was in synch over 25 minutes? I could be wrong, but I think you're
just lucky. There are two independent, relatively cheap crystals in play
there. If this is not true, then please disregard the rest of this post
- I don't know how the minidisk could be synched to the camera.


25 minutes is 1.5e06 ms, crystals with 1e05 accuracy would give less than
20ms drift.

That 1.5e06 figure is likely a mean for the spectrum of
the device, not a scalar. You could get two standard
deviations worse case in typical operation.

If it goes to accumulating error, it can add up quickly. Is
this likely? I do not know....

Compared with a watch crystal this would mean a watch being
off by 25 seconds a month. This seems not to be an extreme specification
to me. I found out mpeg is quite sensitive for audio/video frequencies. In
DVB forums I read about receivers getting out of sync when tuned on the
same station for a long time. So I expect the (1K$, 3CCD) mpeg camcorder
will have decent crystals.


No, I think you pretty much nailed it - they're
mostly 30ish ppm clocks.


FWIW, I've found the "chase synch" capability of the little Fostex
VF-160 class recorders very useful. It's full 16 bit PCM, not ATRACS. It
also actually adjusts its speed to match the synch signal - varispeed on
an analog tape recorder causes it to shift pitch.


I have more trust in a fixed good clock, than in an adjustable clock. The
camcorder does not have a synchonisation signal.

I figured.

In my DAW I can make small adjustments. With a reference track with good
AV sync it is possible to make these adjustments. I found out it is less
difficult to align audio to audio in the DAW, than to align video to
audio.

That would most likely work pretty well, but it's gonna drift now
and again. Have you put the minidisk audio in one ear, and the
camera audio in another, and listened for the entire program?

I dunno - maybe this isn't a problem on a subjective level. I
do know that national broadcast program material is *routinely*
off by quite a bit.

With piano the keystroke rate often is higher than the framerate.

That is not what we're after here. The thing being managed is the
drift bewteen audio and video. The intervals between keystrokes
can be well represented by something as slow as MIDI. It's the
gross level drift that can be a problem.

In my case, with long (25 minutes or more), shots, at least a good
synnchronisation between the video and one (the camcorder) 48khz channel
is a good starting point.

I am still looking for a good device to replace the minidisk. The M-Audio
Microtrack could be a nice replacement. It will do 48khz and 24 bits.


That thing sure is cute. And it's *PROFESSIONAL*!
Just doesn't chase synch. I respectfully submit that
chasing synch isn't that hard - it's just not considered
to be important in this case.

I suspect you like it better because it's going to be
considerably more convenient. And that is fine. That
may be more important in this case.

Respectfully, all them hurtz and bitz don't affect the
subjective entertainment product consumption experience
like things being off synch

But perhaps this ia personal gripe on my part. Bugs hell
out of me when things are off synch.

--
Les Cargill

remeber 6 ms is about the time it takes sound to travel 6 feet.

If you are watching a performance out in the audience, you are easily
more than 6 ms off anyway.

Mark

Ed Anson wrote:

I routinely shoot a half-hour concert with two DV cameras and a separate
audio recorder. In post, I sync them all up at the start, and check the
end. So far, I have never had to make any adjustments for drift.

Don't forget that milliseconds are a much smaller deal with sound/pic
sync than they are for pure sound recording.

What do I mean? In audio 20ms means you get comb filtering -- no
laybacks or "bounces" next to other tracks that have any leakage. Since
most of us do audio, every millisecond counts.

But, as a video editor once pointed out to me, that kind of accuracy is
futile when doing sound to picture. Why? Well, let's take video at 30
frames per second. The video is basically taking a still frame once
every 33 ms. There's a chance that a video of a snare drum hit might not
even capture the very moment the stick hits the head. Instead there will
be two pictures -- one just before it hits the head and one just after.
(And the stick will probably be a blur in those pictures anyway.) So, a
few ms slop one way or the other don't hurt that much. You get away with
it in a way you wouldn't if you broke up the aural components.