BUG: Match alternates out of Sync to AAF

[Frank Kruse]

Hi,

I've finally found the time to trace an issue we've encountered for years now but that we always manually "fixed". Conformed original BWAVs are out of sync by up to a frame compared to the AAF.

Here's the normal workflow:

-Production sound recorded on Cantar, Deva, SD788 etc. Direct import into AVID MC converts to MXF audio.
-Copy all AVID audio media to PT and transfer cut via referenced AAF
-Expand to tracks using original BWAVs from production recorded.

So far so good. When listening to the conformed tracks against the AAF there are random sync inaccuracies from bang-on to light phasing to terrible echo.

I always wondered how this can happen when the files are identical incl. meta-data so there's no way a machine could mistake an in-point in samples after midnight for a different one.

So did some tests and it turns out that ALL the MXF media is not identical in length compared to the original BWAVs by random amounts between one frame minus 1 sample to one sample.

And the reason is that the AVID MC seems to fill any file that doesn't end on the full 24fps grid (or 25 if it's a PAL show) up to the next full frame.

On my current film the sound was recorded on a DEVA Fusion and this machine doesn't close it's files on full frame boundaries so there are all sorts of sync issues in the conform.

Bottom line is: The issue is a combination of Zaxcom not recording full frame length files and AVID manipulating the audio without notice.

I think this should be addressed as it's causing labour intense hand-syncing on a clip-by-clip basis like in the days of tape -based conforms.

Here's the "evidence":

Screenshot of AAF and conformed BWAV (in all screenshots the BWAV conform is on the top, the MXF AAF on the bottom):



As you can see the conform is out of sync.

So I wound out both clip to their start (100% handle) and snap the out of sync conform so it starts exactly where the MXF starts:



Then when you look at where both clips end ypu can see that the MXF file has digital zero audio added to it's actual end to fill the gap till the next full frame while the original BWAVs ends where there recorder dropped out of record:



Needless to say that the amount of "zero" at the tail is exactly the amout of shift between the two clips.

I have no idea why the AVID MC adds audio to each file on import. This shouldn't be happening. Could also be related to the Deva not starting clips on full 00-Frames or not ending them on full frames? No idea.

Since the digital zero is added to the end it wouldn't explain the shift since sample #10000 is the same in both files.

Hope this wasn't confusing and I didn't miss anything.

Frank,.

UPDATE: This happens in PT10 and 11, btw.

[conleec]

Hmm, interesting. I too have noticed slight shifts in audio when doing field recorder matching. This makes sense, as Avid always works against video frames. It happens the other way too. If you do an AAF export back to Media Composer from PT, I think you even get a warning: something to the effect of, all clips will be truncated to frame boundaries, or something like that.

I'm not sure what the solution is, honestly. But at least I can understand why the random differences exist.

Chris

[Frank Kruse]

Originally Posted by conleec

Hmm, interesting. I too have noticed slight shifts in audio when doing field recorder matching. This makes sense, as Avid always works against video frames. It happens the other way too. If you do an AAF export back to Media Composer from PT, I think you even get a warning: something to the effect of, all clips will be truncated to frame boundaries, or something like that.

I'm not sure what the solution is, honestly. But at least I can understand why the random differences exist.

Chris

Even if the AVID has to add those samples there's still no logical reason why the conform is out of sync since the in-points (samples after midnight) are still the same.
In your example nothing goes out of sync. It has to be done because of the nature of MC vs. PT.
The solution is to fix this bug I'd say
We've had similar bugs like this before where a head-fade on a swapped out clip would push the underlying audio later out of sync by the length of the head-fade.
This time PT seems to push the conformed audio forward out of sync by the amount of added zero at the end of the MXF file

frank.

[Frank Kruse]

Update:

Just checked BWAVs from an SD788 and they always start and end on 00-frames so it would be interesting to find out if a project based on SD788 files becomes out of sync as well.

If not that would pretty much confirm this as a PT bug or something you want to look out for on Deva-projects.

Frank.

[Craig F]

Frank,
You should re-post this on JWSoundGroup http://jwsoundgroup.net/
Jeff is a Zaxcom user as are several other members.
That might get Zaxcom's attention on the frames issue.

[Frank Kruse]

Originally Posted by Craig F

Frank,
You should re-post this on JWSoundGroup http://jwsoundgroup.net/
Jeff is a Zaxcom user as are several other members.
That might get Zaxcom's attention on the frames issue.

A friend already forwarded the link to the zaxcom forum. Thing is: Even if zaxcom changed the machine to reflect Aaton and SD behavior I think this is still a bug. And then there's still lots of recordists that set their machines to 25fps TC on 24fps HD shoots over here so the issue could still occur. So ultimately PT should be fixed.

But let's see who's faster.

Frank.

[mfranken]

Hi Frank,

I saw this problem myself a good year ago on a TV drama. Main unit was deva and 2nd unit was SD. Main unit had sync issues and 2nd unit didn't.

When I spotted one of the deva location audio files to original TC I found that the WAV file didn't start on the frame edge. Plus the distance between start of the file and frame edge was the amount the conformed version of this WAV was out-of-sync with the AAF.

From what I am seeing the problem is that the deva doesn't start recording on a frame edge. MC doesn't understand sub-frames so it starts the file on the frame edge. Hence the MXF or media in the AAF is out-of-sync by the number of samples the WAV starts before the frame edge. I don't think it's a PT problem, but I may be wrong.

So from what I can see either the deva should start recording on frame edge or MC should truncate the start of WAV files so they do start on a frame edge on import, or both!

I followed this up with a recordist here in Sydney who uses SD recorders. The word back from SD was that they start all their recording on the frame edge for all their HD recorders. This adds up with what I saw on my drama show where the SD files conformed correctly.

Hope this helps.

Regards,

Mark

[Frank Kruse]

Originally Posted by mfranken

Hi Frank,

I saw this problem myself a good year ago on a TV drama. Main unit was deva and 2nd unit was SD. Main unit had sync issues and 2nd unit didn't.

When I spotted one of the deva location audio files to original TC I found that the WAV file didn't start on the frame edge. Plus the distance between start of the file and frame edge was the amount the conformed version of this WAV was out-of-sync with the AAF.

From what I am seeing the problem is that the deva doesn't start recording on a frame edge. MC doesn't understand sub-frames so it starts the file on the frame edge. Hence the MXF or media in the AAF is out-of-sync by the number of samples the WAV starts before the frame edge. I don't think it's a PT problem, but I may be wrong.

So from what I can see either the deva should start recording on frame edge or MC should truncate the start of WAV files so they do start on a frame edge on import, or both!

I followed this up with a recordist here in Sydney who uses SD recorders. The word back from SD was that they start all their recording on the frame edge for all their HD recorders. This adds up with what I saw on my drama show where the SD files conformed correctly.

Hope this helps.

Regards,

Mark

Mark,

not sure if what you said matches with what I've found. As you can see in the screen-shots the AVID adds samples to the end of the media on import not the beginning to force the files to comply to a full frame count length.
But generally yes. The deva is the only machine among the usual suspects that causes this.

Unfortunately the consensus at Zaxcom and the location recording forums seem to suggest that this is an AVID issue although they (zaxcom) are the only ones that are causing this.

Having said that, I don't understand why PT swaps the matches out of sync. There shouldn't be a reason because sample-count up to the point where the added AVID silence starts should be identical. Unless of course the AVID changes the timestamp in "samples after midnight" at the same time as well which would clearly alter the reference used for conforming because PT reads the reference from the AAF media. If the latter was true it would be a combined issue of AVID/PT bugs surfacing by the way the Devas work.

I hope either AVID of Zaxcom can offer a fix for this. I hope Zaxcom will adjust their firmware as this can also help THEIR users in a scenario where they need to cross-jam different TC rates. I've suggested this to them several years ago (via deva owners) but there was little (actually no) interest to change the behavior although all their competitors use 00-frame starts and full frame count files.

Anyway, I'm hoping that this thread might change something.

Frank.

[mfranken]

Originally Posted by Frank Kruse

not sure if what you said matches with what I've found. As you can see in the screen-shots the AVID adds samples to the end of the media on import not the beginning to force the files to comply to a full frame count length.

I'd say that the deva starts and stops on button push regardless of frame edges. MC loads the file out-of-sync (as the WAV doesn't start on the frame edge) and as the file doesn't end on a frame MC extends it with silence.

Originally Posted by Frank Kruse

Having said that, I don't understand why PT swaps the matches out of sync.

Because the media in the AAF isn't where it should be. Well that's what I recon. PT expands the WAVs correctly as it understands sub-frames/samples. Of course this would mean that all sound loaded from a deva in MC is up to 1 frame out-of-sync. Not good as a guide reference.

If the above is true I guess someone could write an app to truncate the start of deva WAVs to start on the frame edge before loading into MC :)

Regards,

Mark

[SBP]

Hi -We also noticed sync errors matching between material supplied as part of AAF and original production sound recordings. We went someway down the road of tracing the issues and sending the information that we found to AVID. -this is some 2 years ago now. -the problem was still in existence eairly this year MC6+PT10. -I don't know if it is corrected in MC7/PT11.

Initially production sound was being recorded in such a way that production recordings did not start hard on frames -but even when we corrected this not all of the sync errors we saw when matching back to original sound files were corrected.

We found that it was also important to stop the AVID media composer from re-rendering any files when it makes the media for the AAF -a direct copy was fine -however if the AVID media composer re-rendered -to change bit rate or file type or whatever -then we would often see large and variable sync offsets when matching to the original production sound -and the production sound was clearly OUT OF SYNC. -in these cases the AAF audio was IN SYNC.

The work around for us was not too arduous. [no re-rendering of files] -once we had it sorted .

All the best

Bruno