PIX 240 Sync Drift Nightmare

[John Hopkins]

Check this:

PIX240 Sync drift on Vimeo

OH boy! We've recorded hours of footage with this issue prior to Update 1.7. How does one go through every interview and correct the 1 frame lag every 3 minutes? Is there some solution to process every 23.98 psf ProResHQ file we recorded/converted at 23.98 1080p on the PIX 240 to correct this for every interview we did? We are using an Sony F3/w S-Log and set the PIX to convert 23.98 PsF to 23.98p 1920x1280.

I've broken all the tracks down in FCP and exported into 3 minute segments and used the formula below; Is this what you did and how you prepared each in QtSync?:

Based on 60 minutes of footage (59.57 sec). ApplePro Res HQ 1920x 1280 23.98 as recorded on our PIX240 from our Sony F3, scaling from 23.98 PsF.

QtSync offers a free tool to attempt to sync this post nightmare: QT Sync

Qtsync settings: offset each segment by adding 1 frame for every 3 min. segment:
from 0 to 19 segments.

0 = (0 frames added)
1 = 1 (1 frames shifted forward on this segment to the audio tracks)
2 = +1 (etc.)
3 = +2
4 = +3
5 = + 4
6 = +5
7 = +6
8 = +7
9 = +8
10 = +9
11 - +10
12 = +11
13 = +12
14 = +13
15 = +14
16 = +15
17 = +16
18 = +17 (17 frames frames shifted forward by QtSync)

This seemed to work, not sure how precisely as I was only guessing at lip sync being on with no clappers or reference sync points seen in any of the frames being a head and shoulders interview.

However joining this back together messes up the audio in Qtsync where for a few frames it seems where it is joined with the tracks of each segment being offset. Salvages only what lies within these three minute segments.

Ideally, I'd like to conform and correct sync so as not to lose any audio at the joints between the segments.

Perhaps you figured something else out. If there is way of doing this better, I would appreciate any help.

Dual Eyes by Singular Software might be able to take the audio off the SXS cards and conform that to the PIX 240 video. No success so far but still trying. So far Qtsync is the only very awkward solution but it feels far from an accurate fix. Dual Eyes has a 30 day free trial for those who want to take a crack at this solution. If you get it to work please let the rest of us know. Singular Software - DualEyes

Needless to say this whole nerve chewing issue has been an expensive and time and resource consuming post disaster for us. The PIX 240 has lost it's initial love we bought it with. Most of the bugs are likely out, but I'd still be running an extensive series of clapper tests for any lengthy takes involving scaling to be sure. Would I buy one now. Yes. Would our producer's buy one after this experience. No. Once burnt, twice shy as they say. SD is generally solid and it's been well earned, but this is a first effort into the video recorder end of things, and the rush to market may have been too soon. The genius that designed the PIX deserves better software support, as do those who thought enough of the brand to purchase one. Sound Designs has promised to look into a precise fix for those impacted. Our production needs to find one asap as we are in post with lots of people involved and lots on the line. While no release ever seems or might never be perfect, the Sony F3 being a perfect example as well, Sound Designs is a company that will go that extra mile on support. So hopefully they will get back to us all on this soon who may have fallen into the same trap with the PIX. If you have not updated the camera yet, do not shoot another frame would be my advice. I will say that companies such as AJA and Sound Devices are generally brands which are built on robust hardware design. It just seems that with so many devices cramming onto the market and short development periods being forced upon companies, too many of these serious glitches are happening for the kind of money we are paying for these "professional" recording devices. It seems like the real testing process has been offloaded to those buying and trusting these brands. Regardless, AJA, Sony, and Sound Devices are still far above the rest of the crowd out there. Let's see what Sound Devices comes up with to help solve those impacted by sync drift problems on long takes with the PIX 240. Having identified the problem, I am sure they can help us find a precise fix.

If anyone has a solution that works, please post.

[Paul Ream]

If you've got an interview of 60 minutes and you want to use as one continuous long shot, why can't you just lose 1 single frame of audio every 3 minutes? I realise that will be 20 extra edits, but if you perform the cut during a pause nothing will be noticed?

Oh and make sure you're not using 1.05 any more! (They're up to v2 now)

[John Hopkins]

To find sync you need to shift and increasing amount of frames beginning approximately every three minutes, and not cut out frames which would cause glitches in the sound track. So the only solution I could think of was to cut the interview up into 20 segments. You could re-sample and modify possibly but that would be extremely finicky and expensive in post with an audio engineer on the pay roll.

[John Hopkins]

Had some help from Dror Lahat whose tests confirmed this very worrying bug. He suggested not to offset the audio but rather increase it over time like you suggested. *Went back and re-tried QtSync which is a free program found here:*QT Sync This seemed to do the trick and was able to get it close from the beginning to the end of the interviews affected.*

Using QtSync, set the DURATION values and "arming" the two audio channels for "17 frames," for an interview which went 59:57 minutes/frames. In a shorter interview of *27:05:12 minutes/seconds/frames it was 10.5 frames. t would be nice to get an exact formula as one has to guess and experiment until it looks and feels right on sync. Perhaps SD's engineers might help there. I was told they are looking into this, and may have an answers for those impacted soon. But for anyone struck by the same problem I suggest you try this QtSync method.

Initially, I entered the values in QtSync using the offset window at the top and could only get sync that way by breaking everything down into 3 minute segments. Not sure what the bug is, Sound Designs seemed to indicate it was something to do with the way the PIX interpreted PsF while converting to 23.98 using the PIX scaler, and recording that result to the SSD. Here's what Dror said he believes the bug is or was not that SD has corrected the bug in anything from 1.07 update up:*

"Remember that audio has no frames. Also, I eventually realized that the PIX wasn't adding frames to the video; the PIX has no capability for buffering like that, nor can it predict signal it hasn't received yet. What was the likely cause for the bug was inaccurate framerate (speed) metadata in the QT, at the Quicktime Atom level or something. I think the audio played correctly at 48,000 samples per second (or whatever audio settings were), but the video ran at 23.98fps instead of 23.976fps."

Everything we shot with the Mini KiPro was fine, but our long interviews the PIX stuff was alarming when discovered the sync problems. We chose the PIX because, using large capacity SSD's, we did not have to interrupt our interviews to change (very expensive low capacity) CF cards. Certainly it's fantastic QtSync could do this, when the paying programs I tried failed. The pic attached shows where I entered the values. You have to play with it a bit to figure out how many frames to enter when setting the "Duration" for both audio tracks. It depends on the the length of the interviews. The longer the interview the larger amount of frames you enter but you have to experiment with the number and just watch closely to see if the interviews are syncing with the lip movements.*

[David C Wright]

Hi John,

Have you found any more solutions to this problem? Have Sound Devices been a help, they have always been good about customer service

[John Hopkins]

Yes sure, I wrote the Michael Gerbes who wrote the QtSync software looking for some feedback to ensure we can try to find as precise of fix possible using his terrific little and free program. He was very kind and wrote back to me with this, and here's his answer to share with others similarly impacted with this issue:

Hi John,

"removing a frame of video" means that the video has to run quicker.
This can be achieved by using the "Duration" selectors. (If the duration is shorter, the track has to run quicker.)

In the QT Sync "Track Selector" Window (the window that opens above the video) select "Show as Frames" and "Total". The duration of all tracks will now be displayed ( in frames ).

Let''s assume you have 25 fps and a 10 minute video, the duration will be shown as 15000. ( Total number of frames ) in case of 23,976 fps and a 10 minute video, the duration will be 14386.
Lets call this value "A".

you will have to do a little bit of math here: (don't worry, its easy ;-) )
How many frames do you want to "remove" from the video in total ? (In other words: How long is you video ?) If its only 3 minutes long you want to remove 1 frame.
If its 30 minutes you want to remove 10 frames (1 frame every 3 minutes)
Let's call this value "B".

Now lets do A - B. (Total number of frames in your movie MINUS Number of Frames you want to remove from the movie). Enter this new number into the Duration box of the video track AND PRESS ENTER.

Hope this helps,
Michael