Generational loss in 16:9

farss wrote on 5/10/2007, 3:16 AM

This came up in a discussion about HDCAM Vs HDCAM SR today:

Because HDCAM is 1440x1080 every time you add a super the pixels are stretched to 1920 for the composite and then squished back to 1440. Do this several times and things fall apart. SR avoids this.
This sounds like BS to me for several reasons:

1) It'd make more sense to squich the 1920x1080 super to 1440x1080 and then do the composite.

2) If this were true, wouldn't it also apply to any non square PAR format, even 16:9 DV or DigiBetacam. Even regular 4:3 DV would suffer although I guess the argument is the more ARCing the worse the effect.

3) If this were true DVCPro HD is in even more trouble than HDCAM.

4) Modern hardware and software compositors are pretty damn good at resampling anyway, even if this is what's going on.

The rest of tirade against HDCAM made sense, this part didn't after I thought it through.

Bob.

Comments

apit34356 wrote on 5/10/2007, 3:51 AM

Farss, stretching 1440 to 1920 debate is not new and there is a lot of good software for it, but 1920 to 1440, details will be loss, even the best compressors can't take 1920-4x4x4 to 1440-4x4x4 without some loss, tho it may not be "visual" at 30p.

I'm not sure, Bob, what your direction is, off-line storage or delivery format or just pointing out work flow issues.

farss wrote on 5/10/2007, 4:46 AM

Just pointing out work flow issues, well more to the point, are there any?

I've always assumed that the PAR only matters when the image is recorded and displayed. I'd imagined that when a non square PAR frame is processed in anyway the pixels are not stretched to square and squashed back again.

For example assuming I'm compositing text over 16:9. I'd assumed the text is matched to the video's PAR and the pixels composited, no stretch - squish of the underlying image happens, the text gets squished and the video is left alone.

Hope I'm making sense.

Bob.

GlennChan wrote on 5/10/2007, 11:46 AM

I think when you do image processing with HDCAM material, you typically have to move the material over HD-SDI as a 1920x1080 image. Then you apply image processing, then send it back to HDCAM where the deck downsamples to 1440x1080.

The only way you can get HDCAM material as 1440x1080 is if you can transfer the native HDCAM material over in its compressed form. I don't believe there are any systems that currently do this (Xpri used to; I think there was another system which could also do this, but I forget).

2- Perhaps part of the problem may be linked to monitoring. HD CRTs don't have that great resolution compared to 1920x1080 LCDs. The ~22" Sony BVM CRT for example is noticeably blurrier compared to a 1920x1080 LCD (the 32" is slightly better). The 22" is also small, so you have to be close enough to the monitor to see everything. If you monitor on a HD CRT, it can be difficult to see the advantage of 1920 versus 1440.

3- The most useful thing to do would be to test things out.

farss wrote on 5/10/2007, 3:02 PM

Thanks Glenn,
that's the problem! You can't get to the bits on the tape.
Whenever the broadcasters are moving the footage down HD-SDI to another deck losses mount up.

So anyone getting HDV dubbed to HDCAM for say festivals would do better getting multiple copies made directly from the HDV tape or HDD rather than dubbing from a HDCAM master.

Bob.