I actually know the answer to that, but I want to prove that is what is happen first before giving technical explanation of timestamps. You should not rule this out as a cause of your issues.

I, as well as others reading these posts, would like you to share this info. You can give me a layman's explanation, no need to make it complicated or long.

The only reason I want touse Neoscene is that Vegas struggles with long and/or many 5D native mov files. The actual files themselves, the original 5d mov files are just fine IMO - they sync perfectly with my Z5 files, even over a 12 minute clip.

The only time any issues arise are when the 5D files are converted to AVI by Neoscene. Even then the video part of the Avi is in sync with the Z5 files, ONLY the audio has become screwed up.

I would be most grateful if you would explain to me and others, as you have stated you like to help others, how you reach the conclusin that it is my camera/card which is dropping frames. I don't want to upset or offend you, but that explanation sounds really ridiculous, however I am prepared to have my mind changed with a logical and reasonable explanation.

Simple version, the missing frames are flagged as duplicates, some cameras use these 'D' frames to pad 30p out to 60p (for a 30p, same for 25 in 50p) source, so the default behavior is we don't transcode 'D' or Duplicate frames, as those increase the data rate, for no user gain. Similar to that way 3-2 pulldown converts 60i to 24, you want the actual frames, not frames with padding. The resulting transcode is frame or frame all the decode-able frames form the source file. This is the correct thing to do for most sources, except then the source is corrupted and those dropped frames are needed to reconstruct the timing of the damaged source file. This is typically very rare. In the original file the index table points to empty untranscodable frames, in this case the previous frame is displayed (i.e. the duplicate,) so you may not notice is these stalls in playback if your source is mostly static, although motion would stutter a little for every framed dropped (note: not encoding the 'D' actually looks smoother, eye sensitivity thing.) As the audio stream is stored as uncompressed and there are not equivalent 'D' frames, the audio and video can go out of sync if the D frames are not transcoded. While we could add D frame encoding, in most situations it is not wanted, and potential a new failure point for user selecting the wrong option.

If I'm correct, all this is pointing to you are not capturing all the frames, and you will want to get new flash media so you clients have the best look video.

I'm out for the rest of the day, so please investigate what you can on your end.

So the original 5d clips are fine, they sync perfectly with the z5, even over the full 12 minute clip - this to me suggests no issue with the cards i use (which, incidently, are according to Canon well within the spec required), also the video converted by Neoscene is fine, it's just the auido conversion causing issues and somehow the card is an issue? This doesn;t sound plausible. The video, as you know is far more complex a file and makes the majority of the file, the audio being far more simple in structure and much smaller size too. So why would that (the audio) part have any issues when the video doesn't?

Wjy does the original file, both audio and video, sync perfectly? I say the card is just fine - i never get any buffer warnings on the camera, the original files are sweet, all this points to a conversion error with Neoscene.

Additonally, Neoscene costs around a hundred quid. My 32GB cards (i use 3 of them, all work just fine) cost 50% more than that. If, as you suggest, I should use a higher spec card, then the cheaper option for me is to not use Neoscen, obtain a refund from you and not have to buy higher spec cards costing £200 or more!

I appreciate you coming back to me with your repsonse, but the more i hear the more in think 'no way'! Ca you offer any explanation as to why the master 5D files sync fine, but your converted files don't. Why does the original file sync perfectly? If that file had a similar issue as the Cineform conversion, or if the buffer warnign flashed, or if there were any other indicstors that the card is too slow then trust me, i would happily accept that resolution. I would prefer not to have to come on here and spend my time and effort trying to resolve this matter. I am not convinced by what i am hearing from you - yet.

Jonathan,

David is suggesting that the CF encoder will not encode the corrupt frames so you will have dropped frames after the transcode to CF. Thus the sync issue. Your original in the MOV wrapper will happily display them thus no length difference with the Z5. The audio as he says is uncompressed so it will probably be longer due to the dropped frames. That is my understanding from his explanation.

I think your best bet at this time is to get the file to them somehow to confirm his hypothesis. It could be helpful to all of us. Perhaps if it is a big enough issue an option could be provided in the preferences to allow duplicating of the corrupt frames.

Buffering messages in these cameras only indicate the camera is not able clear its buffer fast enough to the card(ie slow writes to the card) and may not indicate random corrupt writes.

Yesterday I had a chance to transcode my 7D footage into Cineform Filmscan 1.

1080p23.97 original, same settings for output. I chose the longest file I had, which was 4.5min.

I used both HDLink and Adobe Media Encoder.

HDlink took 1m45s, Adobe ME took 7m12s for the same encode. Same machine, of course.

Dropped the original 7D footage and 2 derivatives on timeline of Adobe Premiere CS5.

Visual quality is exactly same among all 3 (at the first look.)

Audio kept in sync throughout in all versions.

However, looking at the end of the clips, you can see that HDLink clip is 1 frame shorter, and Adobe one is 1 frame longer than the original.

Lipsync was still fine in all versions though.

I don't have a longer clip at the moment, so it's not clear to me whether the frame difference would accumulate on longer footage, or will it stay as a 1 frame delta at the end.

Anyway, so far with my limited test, I did not detect any catastrophic problems with Cineform transcoding. If anything, it was 6 times faster than Adobe Media Encoder, in my setup.

Using the latest version of Neo 4K here.

Hope this helps.

Alex.


Does Premiere still count one extra frame if you right click on the clip and then select the default frame blend unticked "off"?

In playing around in cineform and CS3 with downloaded demo 5D footage (cars on forest highway and seaside wave break into rocks) it was found that simple playback speed conversions, ie., 30P to 25P worked better if frame blend was selected "off" and exactly divisable or multipliable totals of frames were selected by trimming the end of the clip so Premiere was not doing any fractional math.


Jonathan.


Is Blackmagic codec or hardware involved in any of your system?

Also for sake of curiosity, have you converted the Z7 footage in Neoscene as well and laid both clips in a cineform project? I realise this may sound a silly question but I have had issues with mixed codecs in a single project with sync going strange.

Bob, I double-checked, and all my clips have Frame Blending Off, and Field Optiosn None. The footage was progressive.