The problem is that Canon, in it's infinite wisdom, "called it" 30 fps in the freakin' header of the mov file, which then proceeds to screw up any decent app or NLE that pays attention to the header information. Sometimes, with respect to video in the 5D2, I wonder if they would have been better off using some random users from the Internet than their own engineers, doh!

Personally, in my opinion, this NEEDS TO BE FIXED asap by Canon.

Funny thing was in Check Westfall's last tech tips, even he was ignorant of the issue and played off the drift issue as a 30p thing. We made comments in that post of his, and although he responded to everyone else, he didn't respond to our comments. Canon needs to get a clue on this.

(BTW, I may be a little biased considering I had to send my BRAND NEW 5D2 into warranty because it became useless with Err 20 lockups not to mention a hot pixel in video mode that can't be mapped out by the user -- yup, that's not one, but TWO DEFECTS, straight out of the box.)

I have to disagree somewhat. While Canon may have a header labeling issue, the company preparing transcoding should be looking at actual file structure in designing its transcoding software. I would expect that Cineform would have done so. Neo Scene had 5D support specifically added to it. I suspect that there is more to it than just a header issue, but I hope I am wrong about that.

Yeah, that's a semantic blooper. Basically, it means whatever you set the input to, that's what the output will be.

Don't take for granted that the 30fps header is incorrect. I have heard both 29.97 and 30 as the true frame rate at different times directly from Canon, and all of our tests indicate that it truly is exactly 30fps. On a 10 minute clip, separately recorded audio syncs perfectly with the footage playing back at 29.97, AFTER you slow down the audio to 99.9% of the original speed. This would indicate that the video was really 30fps, but who knows how headers and such are effecting the software when I attempt to manually reinterpret at 29.97.

Hopefully we will hear something from one of the David's on this issue shortly. And Cineform, as a developer, may have some kind of information on that oft "rumored" firmware update, and may be some reason to wait on that.

Dave Posted New Beta
 

For those who have been following, David Newman just posted a new beta of NeoScene, which appears to me to resolve the synch issues.... I did one conversion of a long file, and seems to match well now.

Chris,
Thanks for the positive report.

Anyone have the Nero issue? That should also be fixed, just looking for add end user confirmation.

Apparently, Canon USA support doesn't even know, since I contacted them via email recently and they still claim that the frame rate is 29.97 (NTSC). The text of their response is posted elsewhere in the dvinfo forums:

http://www.dvinfo.net/conf/1029179-post6.html

But that said, I have no idea what the frame rate in the 5D2 MOV files is and as you stated, the frame rate value stored in the MOV file metadata could be "goofing up" any attempt to figure that out. The audio and/or the video tracks of the original MOV file could be being decoded incorrectly, if the header information is used and the true frame rate really isn't 30.00 fps.

I do know that up until the most recent update to NEO Scene, the converted CFHD files also had the frame rate in the metadata as 30.00 fps. At least, that issue has been addressed. Keep in mind also, that the problem of the A/V synchrony isn't that obvious, because the maximum duration of a 5D2 clip is only about 11 minutes, depending on content. That's less than 20 frames of "slop" in the longest possible converted file. It is still an unacceptable result, but for shorter clips, the lack of synchrony gets less and less noticeable.

It wouldn't surprise me at all that if there is another firmware upgrade for the 5D2 and the frame rate is officially changed to 29.97 for 30p files.

That would make a lot of sense - in conjunction with 25p support (and hopefully 24p and/or 23.97p.)

Frankly, 30p doesn't satisfy anybody. If they were to change it to 29.97p without adding 25p, that would just dis the European/PAL market even more.

I have difficulty getting my pea sized brain around all of this sometimes. What Cineform has done here is changed the playback rate, right ? David indicates here, and elsewhere that there are not frame blendings or anything like that, only that audio is pitch filtered and stretched to meet the new frame rate as employed by the Cineform intermediate.

So, on an esoteric level, because it doesn't mean much to me in the max 12 minute clip one way or the other, which frame rate runs true to a real time clock ? I know when I drop the camera file and the new Cinframe file on parallel time lines in Vegas, the camera's file ends about 8 frames faster in a five minute clip than the Cineform transcode file on a time line set up using the Vegas intermediate preset. All this has me thinking of time warps and theory of relativity....

We believe real-world clock rate for the Canon 5D MkII is 30.0. As that is mostly useless, we have now conformed the output to 29.97. I'm guessing the camera will be updated to 29.97 (and other rates) at some point.

Wow, that's one hell of a prediction there. Here's to praying that ends up being true. I mean the 29.97 would be nice, but "other framerates," now that gets my attention! Do you know something we don't, heh...

I know nothing more than other Canon outsiders. There have been rumors about firmware video fixes and 24p coming, I'm guess a video fix might include 29.97. Of course with Neo Scene you don't need to wait for that particular fix.

This is why it's hard to a) Understand why Canon used 30.00 fps and b) test whether it's really 30.00 fps. But I digress. Every independent test shows that on a 29.97 timeline, the original 5D2 MOV files will be out of sync with separately recorded audio and/or video of the same footage. That leads everyone who has tested this (except Canon) to conclude that the MOV file is truly 30.00 fps. I've moved on from that a while ago, just because, until Canon does something about this, the only option is to work around this issue. If the only thing that you are editing is one or more 5D2 MOV files, than this issue is somewhat lessened, because you are going to use a different final format out of your NLE anyway. Prior to the upgrade to NEO Scene, your best option was to use a 29.97 timeline and correct the audio yourself in your NLE.

But if you are converting to Cineform, you might as well have the metadata header in the converted file set to 29.97 fps and then correct the audio. As I understand what David Newman has stated, there is no interpolation between video frames to do the conversion. The 5D2 MOV file frames are converted to Cineform frames, of course, by decoding the h.264 and encoding into Cineform.

Why do the neoScene transcoded files being read as non-progressive by vegas? In Vegas, the files are indicated as Upper Field first.