Canon DSLR Audio - "In Camera" Audio to Picture Sync Issue - Any thoughts?

[Dave]

Hi Everyone,

Noticing an issue with with my Canon 7D internal "in-camera" audio.

I would like to use Pluraleyes to do the syncing and have found it works well matching scratch camera audio to external audio recorder tracks.

However, the problem lies in the fact that the Canon 7d's "internal" audio from the camera's own mic is not truly in sync with it's own in-camera video. The audio comes in 1-2 frames before the picture. I shot 1920 x 1080 footage at 24p (internal camera settings) with the internal camera audio at it's 48kHz default. I shot a handclap to camera (no external audio at all) and then view the clip it on the FCP timeline : the audio definitely comes in early by at least one frame. My subjects hands are still spread apart yet I hear the handclap well in advance. Other people are seeing 2 frames early. I may get away with or not notice this discrepancy in general dialogue shots, but when I was shooting a singer lip-syncing to a CD track, every frame seems to count.

I guess another way to look at it is that the "video" is 1-2 frames "late" rather than the audio being early.

I'm using high speed 10-rating compact flash cards and shooting within 5 ft. of the clap.

I thought perhaps the sync was getting bunged up during the MPEG Streamclip h.264 to ProRes LT conversion process. I then viewed the raw h.264 camera file directly in QuickTime - I see that the issue exists straight out of the camera and assume that it's a 7D anomaly.

Therefore, I'm assuming that all subsequent audio in the clip is out of sync with the picture by the same 1 or 2 frames (not accounting for any possible subsequent drift). Now, if you sync any external audio with Pluraleyes or waveform matching, it too will be out sync (early) by the 1 or 2 frames.

Call me crazy, but shouldn't the in-camera audio be in perfect sync with the in-camera picture? Yes, I gather you could ignore the internal camera audio altogether and sync external audio to picture visually, the tried and true way.

I guess it bothers me that a $2,000 7D has this sync problem when my $65 Flip video camera can at least have it's audio in sync with it's own picture. I wonder if any other Canon owners might care to check their internal audio to video sync and report back? I'm seeing reports on some sites that other Canon DSLR models exhibit this same issue. I'm curious how much of a known issue this is. Hopefully a Canon firmware tweak might rectify this quirk... hope so.

Hope I've explained my issue clearly. Thanks again for your interest and thoughts everyone.

Here's a video clip where the issue is demonstrated :

All the best,

Dave

[MatthewFreedAudio]

It could be any number of things. The camera could be experiencing an anomoly, your computers video card may be outputting video and audio out of sync, your video monitor's refresh rate could be slow in which case the audio must be delayed for playback in order to be in sync for working on it, or it could be a camera setting.

Unfortunately, it will involve changing settings and checking things off the list.

[Philip Perkins]

YES! It should be in sync and it isn't! 5D too--we discovered this a couple of years ago, and offsetting everything from the incam audio a frame or two is now part of the workflow.

phil p

[api]

I suspect that it is an FCP X issue. Have you had a look at the clip on older versions of FCP or Quicktime Player 7?

I have used Compressor 4 for transcoding Canon 5D MkII clips to Prores and I haven't had any sync issues with in-camera sound.

[api]

YES! It should be in sync and it isn't! 5D too--we discovered this a couple of years ago, and offsetting everything from the incam audio a frame or two is now part of the workflow.

phil p

YES! It should be in sync and it isn't! 5D too--we discovered this a couple of years ago, and offsetting everything from the incam audio a frame or two is now part of the workflow.

phil p

Hmm, I'll do some tests, but perhaps there indeed is a one frame offset. I have shot at 25 fps with 1/50 shutter speed mostly using an external mic. Firmware 2.0.9, I think.

[Marc Wielage]

I guess another way to look at it is that the "video" is 1-2 frames "late" rather than the audio being early.

This is not the first time that a camera recorded audio a frame or two early. Actually, from a post point of view, the video is probably delayed by a frame or two due to processing issues. I seem to recall an early Panasonic Varicam having this problem, but we caught it within a couple of months of the camera coming out. Panasonic eventually solved the problem with a software update. Many camera companies these days put audio and timecode very low on their list of priorities.

As long as the offset is consistent, it's not an issue. Just fix it. This is why camera slates (and claps) are still necessary, even in a digital age.

[Dave]

Hi again Guys,

I am using Final Cut Pro 7, where I can clearly see the sync issue.... the video above shows the issue in FCP X. It would be interesting to check it in Avid and Premier too. I wonder if there's a way to scroll frame by frame outside of a computer to check this?

By the way, my Canon 7D firmware is the latest Version : 1.2.5. The compact flash card I used is an ADATA 16 GB "Speedy" 533 X

http://ca.adata-grou...&piid=41&lan=en

I wonder if I played the file straight from the 7D's AV outputs into my old PD150 and then bring it into the computer to check it??

I'll see what they say over at the Pluraleyes site - they seem to be proactive about navigating DSLR sound stuff.

If it proves to be a camera thing, which I have a feeling it may be, I do hope Canon can do something about it.

If someone has the new Canon 5D Mk III, it would be interesting to see if it has the sync mis-match too.

Lastly, the sync problem I had yesterday was when I was in 24p mode - maybe I should check what happens in other frame rates?

Thanks again for your interest.... take care.

Cheers,

Dave

[Marc Wielage]

Ask around on the DVXUser Forum:

http://www.dvxuser.com/V6/

There's a big section there for Canon users, and I'm certain this has been discussed. The reality is, these are crappy cameras with many issues. I don't like them for picture because they're 8-bit, and the images get "choppy" at both extremes (over-exposure and under-exposure). They're bad for audio because the audio circuitry overloads easily and doesn't sound good.

Having said that, they can make decent pictures when well exposed, but I think you basically have to treat them as a film camera and sync stuff up the hard way. BTW, be aware that many LCD monitors also delay the picture by about 1 frame, so there's an extra frame of apparent sync. This may be different on different monitors and projectors.

[pverrando]

Perhaps the statement should be- "I've only paid $2000 or so for this camera, so a 1 or 2 frame sync error is something I can live with."

The primary objective of this camera is to capture high quality raw image data, and sound is a distant second priority, as evidenced by the mini-plug audio input. If you desire high-end audio capture, consider a professional video camera, or double system.

[Philip Perkins]

Our tests (from real jobs actually) were with a wireless feed of the master audio going to camera, which reveals the sync issue much more clearly than camera mic audio. Were using a TC slate--the editors sent me a screen grab from FCP which clearly showed the slate closed on screen and the audio waveform of the clap (in the cam audio track) out of sync. It's relatively easy to just PluralEyes the tracks and then move everything a frame.

phil p

[Marc Wielage]

Perhaps the statement should be- "I've only paid $2000 or so for this camera, so a 1 or 2 frame sync error is something I can live with."

[ChrisH]

Perhaps the statement should be- "I've only paid $2000 or so for this camera, so a 1 or 2 frame sync error is something I can live with."

I agree, at only $2000 you cant expect to get pro level camera features. The 5d and 7d are pro level still cameras and take pro level still shots. When it comes to video they are not professional at all. The red one is $20,000 without lenses and its audio and timecode are not much better than the cannons. Only when you get into sony arri and panasonic HD cameras do you really start to see proper internal timecode sync and audio circuits. we are talking $40-$75 k before the accessories packages.

I also agree that its inexcusable for cannon or red to have made these mistakes but what you gonna do?

[Philip Perkins]

I agree, at only $2000 you cant expect to get pro level camera features. The 5d and 7d are pro level still cameras and take pro level still shots. When it comes to video they are not professional at all. The red one is $20,000 without lenses and its audio and timecode are not much better than the cannons. Only when you get into sony arri and panasonic HD cameras do you really start to see proper internal timecode sync and audio circuits. we are talking $40-$75 k before the accessories packages.

I also agree that its inexcusable for cannon or red to have made these mistakes but what you gonna do?

I'm going to work with them the best I can without complaint, since a huge percentage of my clients love them.

phil p

[cmgoodin]

One additional variable in your test is the "Slate App" you are using to indicate a sync mark.

Since the slate app plays a sound and animates the screen there is no reason to believe that it is always in perfect sync. It is a processor subject to interrupts and driver latency and as a computer program must execute commands sequentially. A better choice for checking sync would be a mechanical device like real wood slate sticks or snapping a close-pin with a mike close by to pick up the sound. When you are talking about 1 frame any electronically generated sync indicator can just add to the variables in the sync chain making it more difficult to determine where the error is occurring.

[Marc Wielage]

I agree with Courtney. Simple is best: wooden sticks have worked for 90 years, they're easily visible from a distance, they hold up to abuse, they can survive a 5' foot fall onto concrete... hard to argue with that. I call this "a low-tech solution to a high-tech problem."

[lutzray]

Well, this thread is 10 years old, but for anyone wanting to test their camera for in cam audio synchronization (with subframe precision), I documented an electro-mechanical process here. I'll send you my (dirty) python code if you ask. Sorry for the big movie... I've resized it though!

 

Here you can read about the whole process (measurements and computations): In Cam Audio Lag