Blog Stats
  • Posts - 32
  • Articles - 0
  • Comments - 3
  • Trackbacks - 9

 

It is impossible to grab PAL DV of Mini-DV tapes with NTSC camcoder after all.

It is impossible to grab PAL DV of Mini-DV tapes with NTSC camcoder after all. Took me learning raw DV format to give up on ingesting my PAL tapes with NTSC camcoder and it looks like NTSC camcoder just grab 10 DIF sequences per frame from tape that has 12.

Here is what I tried so far:

  1. Forced NTSC camcoder to dump the stream over FireWire despite that camcoder complains about incompatible tape format. This could be done using GraphEdit and Microsofts DV DirectShow filter. The result would be Type-1 DV avi with avi headers stating that the content is NTSC (120000 bytes per frame) DV.
  2. The next step was to extract DV stream from avi hoping that once repackaged back as PAL (144000 bytes per frame) it would work. This was done with AVI_Dump(in, 0) from transcode. Quicktime could play output .dv file and correctly identifies it as PAL (720x576 25fps), but the picture doesn't look right - the video is there and is partitioned into macroblocks, but those macroblocks are not placed properly.
  3. This got me to look into details of DV stream that consists of 10 (for NTSC) or 12 (for PAL) DIF sequences of 80 bytes DIFs. Each sequence consits of one header DIF, 2 subcode DIFs encoding timecode, 3 VAUX DIFs followed by 135+9 DIFs of interleaved video and audio. Using libdv as reference I wrote a small program to look at DIFs and within the sequence they appear to be fine. However the first thing that become clear was that the DIF sequences within a frame were numbered from 0 to 9 as they should be in NTSC. This was in *every* DIF (2nd byte). I was still hoping that NTSC camcoder simply renumbered DIFs still moving raw data as it was on the tape which was PAL, so I hoped that renumbering DIFs back could recreate the original stream as on tape. But renumbering didn't help and the picture didn't change at all.
  4. Next, I noticed that timecode encoded in 2nd and 3rd DIFs within the sequence looked weird. Timecode is actually fine, it is just that judging by the timecode from 2nd and 3rd DIFs 25 frames per second are recorded, but there are only 10 sequences with the same timecode, which is not enough to recreate PAL frame. So if we assume that camcoder dumps every DIF from tape without modifying them apart from renumbering, we should see more then 10 sequences with the same timecode. But we are seing 10 sequence with the same timecode, properly numbered, alas only 10 instead of 12. So it looks like NTSC camcoder at best just dumps 10 sequences per frame and moves on to the next frame.
So it is impossible after all to grab PAL DV of Mini-DV tapes with NTSC camcoder, because 2 DIF sequences per frame are missing.

Feedback

# re: It is impossible to grab PAL DV of Mini-DV tapes with NTSC camcoder after all.

Gravatar Wow, very impressive sleuthing! Why not just run DVIO on a laptop to grab the stream? 12/1/2005 8:14 AM | Lorin Thwaits

# re: It is impossible to grab PAL DV of Mini-DV tapes with NTSC camcoder after all.

Gravatar What is DVIO and how is it different from Microsoft DV Capture filter which is as low-level into FireWire as Microsoft would let one get?

Anyhow, I believe it is camcoder that is at fault and thus no software could fix it. 12/1/2005 3:37 PM | Aleksey

# It can be debated endlessly

Gravatar Endless debate impossible 9/2/2009 5:14 PM | SandersSays

# No teme

Gravatar Hello from Russia) 10/12/2009 5:42 PM | Polprav

Post a comment





 

 

 

Copyright © Aleksey Sudakov