Best capture for laserdisc preservation

[darkbluesky]

Hi

That's it. I want to make some capture of a few laserdisc for preservation purposes. In a parallel project I'll try to do a DVD version of them, but for now I want to think on preservation only.

I know that, for NTSC, for DVD authoring I should capture at 720x480, or 720x576 if PAL. And not only because these are the DVD resolution but specially because NTSC has 525 lines , ~483 (482?, 486?) of them for picture (normally 480 for picture, which is what the capture cards allow to capture, afaik). Or similar for PAL (625, which ends in 525 for image, iirc).

The original source is progressive, so I can decomb the captures to obtain the original progressive frame (I know, I tried).

I have read that LDs horizontal resolution is something around 430 lines, that would lead to think that 1280 or 1920 pixels is overkill, but I was wondering if for Nyquist theorem 611 (http://www.afterdawn.com/glossary/term.cfm/ntsc) is the mínimum to reconstruct the analog signal -and we capture to 720 x 480- would 1280 (720p) or 1920 (1080p) (maybe 960p) improve in some way the potential information in the capture, regarding D/A conversión, for example?

I know that 720p or 1080p would not increase the quality of LDs capture, and if not managed correctly could even produce some losse of quality, but for a preservation goal:

- Could I capture NTSC and PAL LDs to 720p or 1080p?
- If so, would I loss the possibility of decomb it? I understand, please correct me if I am wrong, that when capturing -let's say- NTSC LDs to 720x480, I am capturing all the lines and nothing more, so I can decomb it (one line = one pixel row), but would it be a mess (imposible to decomb?), if I would capture 480 NTSC LD lines to, for example, 720p or 1080p?
- What about 960p (1280x960, 4:3)? It is twice 480, does it exists? Is it worth?
- For preservation, which could be the best capture?

Thanks!

[Cornucopia]

LD is pretty straightforward.

It is stored as a composite analog signal (fm modulating a carrier on ld), is always SD, and unless the source transfer was known to be a telecined film (e.g. Hollywood movies) it is actually not progressive but true interlaced. And because of LD's solid timebase output, you don't need a TBC for sync or color correction (though you might want to use one for Macrovision bypass).

Therefore you should:
Get a composite analog to digital adapter (maybe even ez cap type device as long as it is know to be high quality & reliable).
Capture to 720x480 (SD for NTSC) interlaced and save as uncompressed/losslessly-compressed.
Edit if/as needed and resave to same compression.
Then run through encoder to convert a copy that is viewable with modern digital devices (MPEG2 for DVD, MPEG2 or h264 BD, h264 for general media devices). Archive the originals.
No de-int (except film itvc)-until final view copy, nor resizing. No further oversampling either.

Scott

[darkbluesky]

Hi thanks for answer.

So, it is useless (even bad) to try to capture to other (higher) resolution, it isn't?

Your explanation, is more or less what I had thought, only that as the LD are from Hollywood movies (so not actual interlaced, but combed), I can just capture and IVTC it, so keep an uncompressed (Lagarith YUV2 4:2:2) as 'master' capture. I think this is what you say that in your last phrase.

If the source was actual interlaced, then I agree, de-interlace just at the last steps, as it is -afaik- a lossy process. Did I understand correctly?

[Cornucopia]

While "digital is discreet" and "analog is continuous", there are 2 limitations to analog Video (unlike film) which temper the concept of sampling as high as possible/oversampling.
1. Analog lines (rows) are discreet. So a 575 line signal (which has only 480-486 lines of active picture) cannot give you any more than ALL its available picture lines. So 480 (or 486 for pro equipment) is what you should stick to, vertically.
2. Even analog has diminishing returns based on bandwidth. SD NTSC bandwidth is at MOST 6MHz which gives about 400-500 pixels per line. So 720 IS actually oversampling, which should give the smoothest + sharpest digital equivalent.

Otherwise, I think you understood correctly.

ITVC not during capture, but after, during editing but before end encoding.

Also, for interlaced sources, many of the advanced deint methods available via AVISynth are preferable.
And yes, since this is a partly lossy process, it is best to minimize the loss - any loss - by doing it only once and putting it off to the last moment.

Scott

[darkbluesky]

Thanks, everything is clear now.

I am going to open another topic to discuss the source (LD version) to choose and keep the forum neat.

[jagabo]

Note that laserdisc is an inherently interlaced format (even film based material on a PAL disc is stored as interlaced video). That means any hardware upscaler will deinterlace, upscale, and maybe re-interlace to produce HD output. You will have no control of the algorithms used to do that. They usually don't do anything smart like IVTC then upscale. The best software upscalers are much better (nnedi3_rpow2, for example).

[johnmeyer]

Note that laserdisc is an inherently interlaced format (even film based material on a PAL disc is stored as interlaced video). That means any hardware upscaler will deinterlace, upscale, and maybe re-interlace to produce HD output. You will have no control of the algorithms used to do that. They usually don't do anything smart like IVTC then upscale. The best software upscalers are much better (nnedi3_rpow2, for example).

I agree: capture and then do IVTC with something like TIVTC:

TFM()
TDecimate()

[darkbluesky]

Note that laserdisc is an inherently interlaced format (even film based material on a PAL disc is stored as interlaced video). That means any hardware upscaler will deinterlace, upscale, and maybe re-interlace to produce HD output. You will have no control of the algorithms used to do that. They usually don't do anything smart like IVTC then upscale. The best software upscalers are much better (nnedi3_rpow2, for example).

I think I get what you says but just to be sure I understand it correctly; when you refer to hardware upscaler you mean in case I would capture it to a bigger resolution than 'native' (so bigger than 480 NTSC or 576 PAL), isn't? If so, I understand it, and that's why I think I'll stay at 480 (576), and in case I'd want to upscale it, I could try to upscale after IVTC.