HD Audio Drift
 

Hi!

I just recorded a session on the HD10 in HD and also recorded it onto DAT to get better A to D converters and no AGC on the audio. When I went to conform the MPEG-2 TS file from the HD10 and the sound that I captured into the computer from the DAT, I found that the MPEG stream was drifting forward about one frame every four minutes. After a half hour the MPEG-TS stream was a full 8 frames ahead of the sound file! I know it's not the DAT drifting since I've used this unit for years and it's always been spot on down to the millisecond.

Anybody else run into this? Any solutions or do you just have to record the sound straight to the HD10 to keep the sound and picture synced? Or is this an artifact of the bundled MPEG Pro LE software that comes with the HD10 not keeping its sync straight? Or could it possibly be the bundled capture utility is drifting?

Interesting problem--wonder if it's because of the 30p...

I will say one thing, good job doing audio on DAT! The HD10 is still brutal going through an unbalanced jack into the camera.

heath

Heath,
What are the problems you had with the unbalanced type of jack ?
-Les

<<<-- Originally posted by Heath McKnight : Interesting problem--wonder if it's because of the 30p...

I will say one thing, good job doing audio on DAT! The HD10 is still brutal going through an unbalanced jack into the camera.

heath -->>>

The more the background noise, the more it sounds metallic.

Also, I noticed, 8 months after it was shot, in a small doc about the making of my film (shot in DV mode on the HD10--couldn't get an XL-1 that day, seriously), I'm playing putt putt golf for my interview (don't ask). I'm standing next to a small waterfall, and when I shut up, the water flows are louder, then drop when I talk. That's probably not so much the unbalanced mic jack as it's the auto audio...Sigh...

heath

ps-I need to market the HD mode more, because so far, I've made money three times with the DV mode, loaned the camera to the Palm Beach International Film Festival on SD mode and made nothing in HD.

Well, I tried CapDVHS to capture today and had the same problem. Apparently it is a clock issue. The DAT is running in real time and the HD10 is running in 29.97 drop frame so it's dropping 8 frames per hour. Now I can't use the DAT audio in MPEG Pro LE because it has zilch audio abilities and if I send it all over to Vegas to conform the audio to the MPEG drop frame stream I have to republish the MPEG file which both degrades it and takes a loooooong time.

HD is proving problematic because of the crummy audio on the HD10. If it's plugged into the AC power supply it just acts like one big antenna and buzzes like mad no matter how quiet a microphone I put on it. The only thing that works is running battery power only and positioning the microphones carefully to avoid the dread RF antenna effect.

And this still doesn't address the issues of the undefeatable AGC and the fact that it's not mic/line switchable. I didn't worry about the audio when I bought the HD10 cause I was planning on running DAT in tandem. Now the audio problems are about to ruin my business. Arrrrrgh!

Any suggestions? Solutions? Audio must go in the camera since sync issues are too much hassle. Has anyone tried impedance matching so you can use a line level mixer and then step the impedance down/up to match the HD10 microphone in?

If you know the exact timing difference, can't you just re time the audio with a program like Goldwave or maybe even virtual dub?
Change the length to 99.997% or whatever the math says it should be.
-Les

Unfortunatly you can't just re-time the whole audio clip. Since the video drops frames at certain intervals, not constantly, you would end up with it being in-sync only immediatly after a frame drop and then it would drift out again until the next.

What model DAT do you have? I'm assuming it doesn't have time-code otherwise you could set the DAT to drop-frame and be done with it. This is why Fostex made a big name for themselves in the film business early on with time-code DATs.

I'm sure somebody has a plug-in for Pro Tools that would fix this issue, by re-timing the audio with the proper frame-sync off-set.

Time-code issues are why more and more audio engineers are losing their hair! ;>

Good luck!

Kevin - Drop-frame doesn't mean that frames are dropped during playback but that frame numbers are skipped in the counting out of the timecode. The timecode advances by two extra frames every minute except on multiples of 10 minutes.

11:59:29 goes to 12:00:02

All frames are put out at a constant rate (29.97 or 30). Drop-frame vs non drop-frame just determines how they are counted.

Ben - The error you are reporting (8 frames out after half-hour) doesn't match the difference between 29.97 and 30 fps playback. This would be 54 frames after a half-hour.

Most likely the problem is inherent to editing with MPEG-2 transport streams which have their own measure of time (Presentation Time Stamps) and was never meant to be used for editing.

For more advanced editing requirements such as synching DAT sound with video you might want to look at other editing solutions such as those offered by CineForm.

yes, you can retime the sound file.
Sound is *not* represented by frames. Pictures are.
Sound *is* represented by samples, usually something like 44 thousand per second.
You are not 'dropping' frames of sound, in that sense.
I hope that helps.
Also, you may be able to have a tech tune the crystal in the DAT deck to be the same timebase as the camera. A frequency counter would be the tool used for that.
-Les

Wouldn't it be 48k per second?

Fixing the drift
 

<<<---
For more advanced editing requirements such as synching DAT sound with video you might want to look at other editing solutions such as those offered by CineForm. -->>>


Hi Phil.

Can you please elaborate on how CineForm could address the ("more advanced") audio/video sync (synch, synchronization, next time someone does a search all 3 synchronization abbreviations will be found) problems.

I have noticed that (and there is no way around it) when HD footage is exported to DVD (format) in Premiere Pro using MainConcept plug-in. The audio would never be synchronized (although it was synchronized in the original HD footage).
I think I get the part about counting "drop frames" business, 2 every minute but not on the 10th. It almost looks as if the issue I'm observing is due to the audio not following the congruent rule when being encoded, hence I observe the ~0.8 second drift after about 12 minutes.

Now that I think about it, I wonder what would happen if i shrank the audio 2 frames worth every minute (but not on the then) on the time line, before exporting the HD to DVD.


-Chris.

Ok, this is making me nervous!

I have the HD10u and Final Cut Pro HD - can someone tell me if I record audio with the HD10u and encode with LumiereHD into DVCPro HD codec if I'll be dealing with bad sync audio?

I've always worked DV, so audio was sync wasn't really a problem because you can click the "adjust" buttons in most NLE to allot for the 29.97 factor.

Is the HD10u screwed up no matter what you do? Anyone out there actually edit on a Mac with long video/audio footage??

Help! I'm going to be recording long-form very soon!

Murph

OK HERE'S ANOTHER GUESS!

In the camera, a crystal determines the frame timing. While accurate, they can have small errors.

In the computer, a crystal determines timing. A difference between the two crystals will result in a difference in timing.

With a digital transfer of audio, the same thing happens.

With an analog transfer of audio, the same crystal always determines recording AND playback timing, in both the DAT and the computer. So no chance for timing error.

Crystals usually have tighter tolerances than this. Sometimes, crystals can be trimmed to a more accurate frequency. This requires a frequency counter. Some software allows for timing corrections.

My guess, a crystal is out of specs somewhere.

Did you use the same camera and DAT for recording and playback?

ADDENDUM:

Crystals can easily maintain frequency to within 5 parts per million. This translates to slightly more than a half frame per hour. Expecting much more accuracy than this would be unrealistic.

Sounds to me like a DV audio v Locked DVCAM audio issue, where the problem with DV cameras is that they don't lock the audio directly to the video stream, and hence don't record a precisely 48khz rate. Perhaps this is made worse by the 6 frame GOP?

Even though the HD10 is "30p" that really means 29.97fps progressive as it's NTSC, so that's not the issue, nor would it be DF v NDF issue, as that's just how the frames are named.

Graeme