Capture Guide Suggestions Part 2

[trevlac]

You guys speak french too !?!

I feel so inadequate

[Arachnotron]

You guys speak french too !?!

I don't, only German and English. We Dutch have to learn other peoples languages, because the rest of the world can't be bothered to learn ours.

Belgium also is a small country. And they already use two languages to begin with....

In the end we will probably all speak Chinese

[erratic]

I just tested my DVD player with some NTSC video in color. In ATI MMC, NTSC and NTSC-J are in color, PAL-60 and PAL-M are not, so the output from my DVD player is true NTSC, not PAL-60.

[North2Polaris]

Originally posted by erratic
It would be great if someone in the US could test the old rage Theater chip (ATI 8500 or older card).

I have an old ATI AIW Radeon, the first in the series with the Rage Theater chip. I put this card into my system and ran the test at 704x480 and 720x480.

The 2 bitmaps can be extracted from the following:

http://home.earthlink.net/~mcartter/...les/upload.rar

4 black + (7.5 - 715.5) = 712 or 52.741 microseconds

I am starting to understand how to read the test pattern. My count may be off a few pixels. The total may be 715. I am not sure where the cut off is on the right edge.

[Arachnotron]

The problem is that the right edges of your capture window coincide with the right crop point of your DVD player.

From the 720 cap it looks as if your DVD player only outputs pixels 7-716. I think the cap area is about 714 pixels wide, but it is very difficult to see because of this.

Could you try 712x480 too? (maybe using virtualvcr if the ATI tool won't let you do this) If the area really is 712, this should look much sharper.

I am working on a guide for this. The working version can be found here. It is by no means finished, but look at the second picture for an explanation of some of the things you might see at the edges

[North2Polaris]

@Arachnotron

Thanks. I did use VirtualVCR at 4 different capture resolutions, including 712x480. All 4 bitmaps are now in the archive below.

http://home.earthlink.net/~mcartter/...les/upload.rar

[Arachnotron]

There is a second method to read this testpic.

1. search for a verticle line from the array of lines in the middle of the pic on the left side. I chose line 80; this corresponds with pixel 79 in the 720x480 cap

2. search for another horizontal line on the right side with a matching gray block pattern. I chose line 490 which has the center in pixel 492 in the cap

zo, 490-80=411 DVD pixels ended up in 492-79=414 cap pixels

the cap area = (411/414)*720 = 714.8 = 715 DVD pixels

edit
With this card it is perhaps better to cap high and resize to 714.

This second method is soo much easier than the first, and much more accurate. Also, DVD player cropping is not a problem any more.

But now I have to re-write my guide. And maybe create other test DVD's . Oh well, back to the old drawing board

[erratic]

So the ATI results are:

Rage Theater capture chip:
NTSC - 715x480
PAL - 705x576

Theater 200 capture chip:
NTSC - 705x480
PAL - not tested (but probably 705x576)

AFAIK you can't use odd numbers for the capture resolution, so with a Theater chip the obvious choice for PAL is 704x576.

[Arachnotron]

@ erratic

Mostly the horizontal resolution must be devidible by 8 for capturing.

I'll look over the other caps using the other method tonight.

[erratic]

With my ATI card every resolution divisable by 2 seems to work, up to 720. The Rage Theater chip doesn't support 768x576, but 718x576 works. I've also tried 468x576 successfully (with padding afterwards for SVCD).

[trevlac]

@wilbert

I made a copy of the guide preface with my proposed changes.
http://trevlac.us/preface.htm

I tried to follow the original outline. The things I wanted to change were 1) Fit in new resize methods 2) Give info on card/driver specifics 3) Address PC,DVD,SVCD & VCD in a common way

Also, I find it inaccurate to refer to "frame size" as resolution. I tried to fit both in because most people seem to mean frame size when they say resolution.

Here is a link on the relationship between frame size, resolution , and bandwidth. http://www.maxim-ic.com/appnotes.cfm/appnote_number/750


@all

Feel free to give feeback. I would be happy to make any changes. You could also just grab the page and edit to your hearts content. Of course, this is all dependant upon Wilbert. After all, it is his guide.

BTW: I tried to add the info on the ATI chips. I may not have the final results correct. This is great stuff you are providing.

Trev


Edit:
I wanted to thank midelic for letting me try to explain resizing in this thread: http://forum.doom9.org/showthread.php?s=&threadid=65809
It was a great conversation for me. It showed me the limits of my simple method with a simple explination.

[trevlac]

Originally posted by erratic
With my ATI card every resolution divisable by 2 seems to work, up to 720. The Rage Theater chip doesn't support 768x576, but 718x576 works. I've also tried 468x576 successfully (with padding afterwards for SVCD).

My experience is also that factors of 2 are what is needed. I assume this is because 2 pixels share color information for 4:2:2, in 24 bytes. So samples are converted to pixels in groups of 2. Macropixels.

By this logic, if you cap in a 4:1:1 (4:2:0) color, you could only use factors of 4.

For MPEG and/or DV, there may be a requirement of (8s ?)16s. Due to Macro blocks. I'm not sure.

I'm not sure about all of this. Just what I'm guessing.

[Arachnotron]

@ Trevlac

My experience is also that factors of 2 are what is needed. I assume this is because 2 pixels share color information for 4:2:2, in 24 bytes. So samples are converted to pixels in groups of 2.

It depends on the brand. Philips drivers will accept anything devidable by 2, but I think Hauppauge drivers accept only divisions by 8. But I am at work right now, so I cannot check it.

But I think it would be more of an restriction set in the driver than a hardware issue, and that devidibility by 2 is the only 'hard' limit for the reasons you mentioned.

By the way, nice work Trev. I have some reading to do tonight....

[erratic]

Well, with an ATI Rage Theater chip it's possible to capture NTSC@714x480 and then crop/pad to 704/720x480 for DVD authoring (no resizing necessary). That's how I would do it anyway.

[Arachnotron]

@Erratic,

You are perfectly right of course. Much easier.

[trevlac]

I just wanted to make some corrections to my macro pixel comments.

For Packed 4:2:2 fromats like YUY2, each macropixel is 4 bytes (16 bits per pixel) and contains 2 pixels. YV12 is not packed so no macropixels. Maybe you can cap in this at 1 pixel intervals (if the driver and card support it).

http://msdn.microsoft.com/library/de...eosubtypes.asp

@erratic

I use BTwincap and cap at 712 and crop to 704 (while I cap). I use VirtualVCR's built in crop filter for this. I noticed you mentioned a problem with VVCR. Just another thing for me to look into.... The learning never stops.

[Arachnotron]

I just opened a new thread for the discussion about determining the capture area. That should leave this thread free for discussions about the new Capture Guide.

I recalculated the data accumulated so far, but no surprises :-)

Everybody interested in this, please move over to:

http://forum.doom9.org/showthread.php?s=&threadid=66191

[Arachnotron]

@trevlac
@erratic

About the issue about dividibility by 2, 4 or 8 : the driver may not give you a problem, but the codec might on later processing. Look at this thread a few days back:

http://forum.doom9.org/showthread.php?s=&threadid=65159

[erratic]

Originally posted by trevlac
I use BTwincap and cap at 712 and crop to 704 (while I cap). I use VirtualVCR's built in crop filter for this. I noticed you mentioned a problem with VVCR. Just another thing for me to look into.... The learning never stops.

AVI offset calculator: This is intended for people who use WDM capture device/programs. These often create AVI captures with internal audio and video offsets, which make their playback sync correctly but tend to ruin things when you try to edit them. This tool calculates the total offset, sets the internal offsets to zero, and renames the file so that you can manually adjust the audio skew in VirtualDub or other application prior to editing. This should result in proper sync for playback and editing.

A few threads you should read:
http://forum.doom9.org/showthread.php?&threadid=50978
http://virtualvcr.sourceforge.net/fo...t=ST&f=4&t=541
The most annoying problem is that the offset (reported by the AVI offset caclulator) sometimes needs to be fixed, but sometimes it doesn't, as you can read in the thread from the VVCR forum. That was driving me mad, so I decided to capture with VirtualDubMod instead.

[erratic]

Originally posted by Arachnotron
About the issue about dividibility by 2, 4 or 8 : the driver may not give you a problem, but the codec might on later processing.

Yes, I suppose people should test this first by capturing 714x480 and then try their usual post-processing on the AVI file until they're sure everything works.