capture frame size--determined by source or target?

[chipzoller]

I have read both the capture guide and capture FAQ several times (and will continue to re-read them as necessary) but one of my primary and most important questions that I can't seem to satisfy is one of a simple nature.

Assuming that my system and capture card allow high resolution captures (me being in NTSC land, 720x480) and that the intent would be to capture losslessly to Huffyuv, I am still confused how to determine optimal frame size in the capture.

For example:
If I'm capturing through a TV tuner (from satellite, cable) with the target intent of PC backup, should I still capture at the highest frame size allowed by my card, or should I limit it to 640x480?

If I'm capturing a VHS tape through an S-VHS player with the target intent of DVD backup, should I strictly adhere to 720x480 (assuming this is the largest frame size limitation) or lower?

Also, does anyone know if this document is still around:
Capture-Cards and aspect-ratio for Dummies: Der Karl's Capture Card Aspect Ratio for Dummies (translated by Arachnotron).
The link is broken in the 'References' section here (http://www.doom9.org/capture/sizes_newbies.html)

[AVIL]

Hi,

IMHO, capturing analogic video is a matter of analogic/digital conversion. If you obtain more samples, you obtain more accurate conversion. Then (in my opinion, of course) video must be captured with maximum hardware resolution. Well settled that the card do true sampling, don't interpolate.

Put as an attachement the document by Arachnotron. I haven`t see any paragrah with copyright notice.

[jggimi]

PM sent to Arachnotron.

[chipzoller]

AVIL:
Well settled that the card do true sampling, don't interpolate.

I don't quite understand what you mean here. Could you clarify?

[AVIL]

Hi,

Could be that a card take (as an example) 640 samples per line and, in hardware, interpolate the samples to offer 768 samples to the software. I.e. the cards output 128 samples created by mathematical manipulation of the 640 real ones. If this is the case, is better capture to 640 and process after the true samples in software.

[Arachnotron]

PM sent to Arachnotron.

Of course jggimi has my aproval. My site has been off-line for some time now, but I'll restore it tomorrow, since the test DVD images mentioned in the guide are on it too.

My provider moved the domain/webhosting business to another company, but did not transfer the contents. And I never got around to doing it myself. Now where was the new password.....:p

If I'm capturing a VHS tape through an S-VHS player with the target intent of DVD backup, should I strictly adhere to 720x480 (assuming this is the largest frame size limitation) or lower?

There is no penalty in capping high apart from storage space. Since nowadays harddisks are cheap, it is not as much of an issue as it used to be.

So I would say, cap directly at DVD resolution regardless of the source format. But do correct for capture window limitations of your particular capture card.

For example, I use the Terratec Cinergy 400 mentioned in the guide. My PAL capture window is 704 DVD pixels wide. So I capture at 704x576 for all sources, crop of the dirty bits :), filter as needed, pad (not resize) untill I have 720x576 and convert to DVD.

704x576 is a legal MPEG2 resolution for DVD too, but some encoders will resize to 720x576 and I am too lazy to test for that. By padding to 720 I know the AR will be correct in any case.

[jggimi]

Attachment now approved and available; thank you Arachnotron.

[chipzoller]

Ah thanks, Arachnotron, for that translation. I learned a lot but have more questions than when I initially started.

So by reading and talking to people I've found that perhaps the best for me (in NTSC land) capturing TV (satellite) and VHS is to capture at the max res. for the card (720x480 ntsc) then crop and resize, filter as needed.

One thing: Assuming the target medium for all my captures will be DVD, should I turn on either 1.) the 3d comb filter supported by my hardware (which I understand is very good) or 2.) convert to progressive frames? It should be said that I'm not familiar at all with interlaced filtering but know there are certain problems when doing so.

[Boulder]

I don't think the comb filter has much to do with interlaced/progressive in the way we talk about those things here. The comb filter should probably be enabled - you could test it by capturing with and without it and comparing the results and/or posting screenshots here.

If you encode for DVD, keep the video interlaced. There are many threads that discuss processing interlaced material while keeping it interlaced.

[Arachnotron]

So by reading and talking to people I've found that perhaps the best for me (in NTSC land) capturing TV (satellite) and VHS is to capture at the max res. for the card (720x480 ntsc) then crop and resize, filter as needed.

Depends on your capture window. If your capture window is 720 DVD pixels wide, that would be correct. Try the procedure from chapter 10 in the capture guide to find out. (and yes, I restored my website so you should be able to download the test DVD iso images again :D )

As to the combfilter: this has, as Boulder stated, nothing to do with interlacing. It's a filter that separates Y and C when digitizing a composite signal. So the source *cable / connector* determines whether you need a combfilter, not the target format. Most cards will automagically switch it on or of depending on the source you selected during capping.

If you use an s-VHS deck connected with a s-vid cable, you can leave it off. If you use a composite (RCA-RCA) cable, leave it on.

[chipzoller]

the source *cable / connector* determines whether you need a combfilter

Well, just for testing purposes I'm capturing from my satellite signal (Dish Networks here in the US) using the S-Video out. The only thing is the comb filter can be set to either 2D or 3D functionality, but there doesn't seem to be a way to deactivate it entirely through the hardware config. setup (maybe some registry switch).

I'll check out the capture window scenario more in-depth.
But just to confirm, the capture window refers to the frame size of capture and not literally to the window size of the program used in actually capturing the feed, correct?

[Arachnotron]

leave it on 3d. It shouldn't matter when capping s-vid.

But just to confirm, the capture window refers to the frame size of capture and not literally to the window size of the program used in actually capturing the feed, correct?

Yep. The capture window refers to which section of the total analog frame the video card captures and converts to a digital image with the resolution set in the capture program.

So in my case, if I were to hook up a DVD player to my capture card, of the original 720 DVD pixels, only 704 would make it into the captured frame.

If I were to cap the DVD player output at 720x576, the result would be the same as ripping the DVD, cropping it to 704x576 and resizing the result to 720x576. Which would change the aspect ratio by stretching the frame horizontally.

[chipzoller]

Arachnotron, so in my case it's my understanding that I should capture at full frame (720x480) from (as a test) 4:3 satellite feed, crop left and right, then resize to a 4:3 res.?
So a basic script would consist of something in this order:
source
deinterlace
crop
resize

With this ATI 550 theater pro card I have, it captures a telecined stream (29.97fps) and so I think it's better in this case to just deinterlace than ivtc.

This is what I've understood from posts and capture guides. Is it mostly correct?

[Arachnotron]

Hunting in an old thread (http://forum.doom9.org/showthread.php?p=625897#post625897) I found the theater 550. Looks like an ITU compliant card, which means that when your target is DVD, you simply cap at 720x480 and convert to DVD without any cropping for all sources to get the AR correct. No resizing needed. :)

If you want to remove ugly borders, do it by letterboxing. (or cropping and padding back to 720x480 with black pixels)

Double-check yourself wit the test DVD just to be sure, but it looks like it is that simple.

[Arachnotron]

With this ATI 550 theater pro card I have, it captures a telecined stream (29.97fps) and so I think it's better in this case to just deinterlace than ivtc.Whether or not to deinterlace or ivtc depends on the source. It could be different for every capture. Since your target is DVD anyway, why not simply leave it interlaced and encode it as interlaced mpeg2.

[chipzoller]

Yes, if the eventual target is DVD I *would* leave it interlaced (or telecined). Just for the time being I'm also testing PC capture as target and testing IVTC vs. deinterlacing.

[Boulder]

If you have a telecined source, you must IVTC it. Encode it then as 23.976fps and apply pulldown on the encoded video.

[Arachnotron]

Yes, if the eventual target is DVD I *would* leave it interlaced (or telecined). Just for the time being I'm also testing PC capture as target and testing IVTC vs. deinterlacing.In that case, cap at 720x480, crop back to 710x480 and resize to square pixel.
Or, if your card will let you, cap at 648x480 and crop back to 640x480. Either gives correct 4:3 AR.

[edit] For a solid introduction to IVTC and deinterlacing, read the documentation included in Neuron2's Decomb filter package for avisynth. Well worth the read.

[chipzoller]

For a solid introduction to IVTC and deinterlacing, read the documentation included in Neuron2's Decomb filter package for avisynth.

Yes, I'm quite familiar with telecining and IVTC theory and methods. And as I've noticed from my tests (so far only capturing satellite signal), this card captures *this* stream at 29.97 and when looking at the frame pattern it's a straight M-in-N telecining. However, when I did some PC backup tests by IVTCing the stream, the resulting file didn't play as smooth as I would have thought, so for this I was thinking of just deinterlacing (or messing with the detection parameters of tdeint).

EDIT: Well, after re-examining my test capture it appears I have a hybrid source: some sections straightforward telecining others interlaced.

In that case, cap at 720x480, crop back to 710x480 and resize to square pixel.

I'm pretty sure this is what I've been doing. I'm not sure if it was exactly 710, but I cropped off the noticeable black areas and resized to a 4:3 resolution, which in any case should yield a correct AR.


After doing some resolution and window checking on my test capture, it looks like there are about 25 pixels horizontally that had to be cropped reducing the captured res. to 695 and 10 black pixels on the bottom reducing the vertical res. to 470. From there I (for PC backup testing) resized to 640x480. According to all the capture guides I've read, TV is supposed to output 480 vertical lines, but since this is satellite I wonder if it should be the same or different? Is this normal in satellite capture cases? I'm afraid I don't have antenna or cable to provide a more varied test group.