Capture Guide Suggestions Part 2

[North2Polaris]

Originally posted by trevlac

I think Jukka Aho's method is still too complicated. This is because he uses pixels. I say forget the pixels. Reduce the process from 7 steps to 3.

1) Look up source "sampling matrix width", and target "sampling rate" from the table.
2) Multiply this by target (source sampling matrix width) * (target sampling rate)
3) Crop/Pad

@trevlac
Your 3-step method is a more elegant solution and is probably less likely to result in calculation errors than longer methods that involve working with PARs.

The second figure in the Der Karl article, “Capture Cards and Aspect Ratios for Dummies”, shows an "analog wave" and the relationship between sampling matrix width and active pixels. Including this figure in a guide using your method could help many users "make the leap" from pixels to sampling matrix width.

The use of "target sampling rate" can be explained fairly simply. What is needed is to resample that part of the analog TV signal or wave that is included in the sampling matrix width using the target sampling rate. This calculation will give the target width in pixels that can then be used to resize the video in the new digital video format.

As someone who uses AviSynth, this method will change the script order that I use.

For example, converting DV to XviD for viewing on a computer.

sampling matrix width for NTSC DV = 53.3333
target sampling rate = 12 + 24/79 MHz
target pixels = 656.2 or 656

The script based on your method becomes:

BilinearResize(656, 480)
crop(8,0,-8,0)

The recommended script based on the DV FAQ is:

crop(8,0,704,480)
BilinearResize(640,480)

When using this script from the DV FAQ, cropping to 702 would be better than 704, but 702 is difficult to get to using AviSynth in YUY2 colorspace. Your method may give a more accurate resize than the method recommended in the DV FAQ, but I am not sure how to evaluate this.

The script based on your method also works equaly well for the the "TV capture" using the ADVC converter that I described in an earlier post -- 640x480 without black borders.

Thanks to all of you for your help on this. If you need a newbie to review drafts of the new guides, I would be happy to help.

[erratic]

Originally posted by Arachnotron
Most likely the real numbers will be 704 and 712.

The weird thing about the NTSC number (712) is that ATI MMC 8.7 (latest version) only supports 704x480 for AVI captures and 720x480 for MPEG-2 captures. Better capture 712x480 AVI with another app.

For PAL it's 704x576 AVI and 720x576 MPEG-2. PAL AVI is correct, but I prefer VirtualDubMod anyway. I used to love Virtual VCR but then I found out about the very annoying DirectShow AVI offset and the AVI offset calculator.

[Arachnotron]

Oops! To fast! I made a big mistake here!

For NTSC:

5.5 up untill 17.5 + 2 black on the right = 713 + 2 = 715 pixels!
715 equals 52.96 µs

For PAL:

12.5 up untill 717.5 is 705 = 52.22

You count the white lines as 'whole" ; the gray lines are the result of a pixel 'hitting' a peak on the flank and count roughly as half.

Very strange numbers indeed! I don't think this is an 13.5 MHz based card.

I suppose this means a lot of resizing. Sorry

[Arachnotron]

@Wilbert and erratic:

I just missed Wilberts remark about this card having a SAA7108 encoder/decoder.

That explains part of the crappy composite grab, since this chip has only a very basic combfilter. But it is a 13.5 MHz chip with an internal scaler.

ATI must really have done a number on the driver for this chip.

I will later produce a SVCD version of this DVD.

[erratic]

If PAL is 705 then I guess 704x576 is good enough without resizing.
Fortunately, NTSC (715) is not really my problem.

[erratic]

Originally posted by Arachnotron
@Wilbert and erratic:

I just missed Wilberts remark about this card having a SAA7108 encoder/decoder.

That explains part of the crappy composite grab, since this chip has only a very basic combfilter. But it is a 13.5 MHz chip with an internal scaler.

ATI must really have done a number on the driver for this chip.

I will later produce a SVCD version of this DVD.

I'm confused. I thought ATI made their own capture chip?

According to this page ATI's Rage Theater capture chip appears on many ATI cards, from the Rage Pro 128 to the Radeon 8500DV.

In this thread Metaluna writes that the AIW 9x00 boards (and the AIW VE) all have a new generation of capture chip, the Rage Theater 200, while the 7500, 8500DV, etc. use the older Rage Theater chip.

[Wilbert]

I'm confused. I thought ATI made their own capture chip?

Ohhh, I'm very sorry. I misread it I thought Arachnotron was asking for someone having a SAAA7xxx chip.

You are right, they have their own chip

[Arachnotron]

Check here

It looks like you are right about it being an ATI capture chip.

From the specsheet of the theater 200 here I gather it communicates with the rest of the board using an ITU-656 interface; that implies a 13.5 MHz chip.

So probably the original theater chip is too.

But this is really getting off-topic :-)

[trevlac]

Originally posted by Wilbert
Yes, something like that. Trevlac, could you give it a shot (i.e. make a proposal)?

Sure,

Here is 1st cut at a table. I'd like to write some text around it and check the numbers, maybe breakout PAL and NTSC, but basically the table contains all of the resize widths for all of the cards from Arachnotron's tests. I may also add ATI as 52.148 for both PAL and NTSC. I'd base this on erratic's comments. I think he has commented on ATI. I may not remember this correctly. I'm tired ...


http://trevlac.us/Destination%20Widt...e%20Matrix.htm

[erratic]

Yes, I've tested ATI's Rage Theater capture chip.
Strange numbers: 705 PAL and 715 NTSC.
I hope someone else will test ATI's newer Theater 200 chip.

[North2Polaris]

I have an ATI ALL-IN-WONDER 9000 PRO. I will try to run the test this weekend.

Additional information on the THEATER 2000 below. Should this discussion be moved to a new thread?

http://www.tomshardware.com/technews/20020104.html

"ATI says THEATER 200 is designed for use in consumer electronic equipment requiring video and audio capture, including multimedia PC to TV video graphics cards, set-top boxes, analog personal video recorders, LCD TVs, and other multimedia platforms. The chip captures Composite, S-Video, and Component (including High Definition) analog video signals and converts them to ITU-656 compliant digital video along with capturing the Sound IF (Intermediate Frequency) or composite audio outputs available from most tuners. It performs audio demodulation, stereo decoding, and outputs the decoded audio via I2S (inter-IC sound - a serial bus for digital audio). THEATER 200 features analog front-end circuitry that includes input source selection, anti-aliasing filters, advanced AGC (Automatic Gain Control) and Clamp circuitry, and dual 12-bit ADCs (Analog to Digital Converters). The video decoder includes Macrovision detection, a vertical and horizontal scaler, and an adaptive 2D comb filter. The chip supports video standards including NTSC, PAL, and SECAM, and audio standards including BTSC, DUAL FM, EIA-J, and NICAM."

[trevlac]

@erratic

I missed the discussion of your ATI tests. I missed that whole page #3, until just now. I see you have an 'official' number from the Rage Chip. Great!

@Arachnotron
@Wilbert

I don't think I mentioned it before, but in my focus on simplicity, I missed the fact that not using PAR makes it hard to match how many people think about and discuss this topic. Thanks Arachnotron for pointing this out.

I don't really like the table I made. I think it would be better to just list the Card/Driver capture line lengths and the sample rates of the standards, and let people do the math. I put some text around the explination, in another thread. I will re-read it tomorrow and try to come up with a 'proposal'.


@North2Polaris

It's great that you are participating in this. I consider myself a newbie. I've just been thinking about this topic alot lately.


As far as the ATI chips being ITU-656, I wondered about that when I read about them before. As far as I know, ITU-656 just "describes the means of interconnecting digital television equipment". If the cards have a digital out, this makes sense. I'm not sure what ITU-656 in the absence of ITU-601 means.

I suspect they are like the other chips/drivers. They capture a portion of the full line and scale it to what you ask.

Here is a posted version of 656 http://www17.cds.ne.jp/~stray/ccir656.html

[Arachnotron]

@trevlac

The 13.5 MHz was just a theory, I have no way of knowing for sure. I just thought that because of the emphasis placed on ITU-656 in the specsheet, they might have made it 13.5 MHz since that would be the simplest way to do it. Leave the system on 13.5 and do the rest through scaling. But once again, it's just a guess.

Visual inspection might give an answer by looking what kind of crystals are on the card. The frequencies are written on them. You still need to be carefull with them, since the actual sampling clock is mostly generated from them

For instance, my philips cards are equipped with 24.676 MHz crystals which are used to generate an 27 MHz reference pulse.

[North2Polaris]

@Arachnotron

I ran your test on a ATI AIW Radeon 9000 Pro at three capture resolutions using VirtualVCR:

704x480
712x480
720x480

I also ran the test on a Canopus ADVC converter.

The 4 bitmaps can be extracted from the following file:

http://home.earthlink.net/~mcartter/...test_files.rar

As for the ATI AIW 9000 Pro, I "counted" 700.5 pixels, which would give a capture window of 700.5/13.5 = 51.8889, if I have followed your method correctly.

The ATI 712x480 and the 720x480 show evidence of scaling, correct?

How would you interpret the results of the ADVC converter?

Please let me know if my calculations are incorrect or if you would like me to run these tests again at different resolutions. I hope this adds to the database.

[Arachnotron]

@North2Polaris

Your DVD player crops and only displays pixels 7 up untill 715, so ca. 709 in totall

Als a result you captured some black pixels at the sides which DO count.

I therefor would set the ATI at 14.5-715.5 + 3 black =704 0r 52.15 µs

The ADVC is clearly an ITU device, since it captures

7 b + 7.5-715.5 + 4b = 709 + 11b = 720 DVD pixels or 53.33 µs

Notice also that your DVD player only displays pixels 7-715=709 pixels with blanking on either side. Unfortunatly, these numbers are completely different for each brand of DVD player which makes it impossible to say anything about the exact location of the center.

As to the underlying sampling which is re-sized, this is difficult to tell. It might be that the underlying sampling grid is 13.5 MHz, but slightly shifted compared to the DVD output. Or the underlying grid is not 13.5 and the pixels were made by scaling.

In the ADVC case I noticed the resulting ghosting is the same across the whole width of the capture. Notice that when yoy look at the ADCV cap at the range of verticle lines in the middle, they always have the same ghost on the left. From this I suspect 13.5 MHz sampling with the offset a little shifted compared to the DVD. So, no scaling at 720x480!

With the ATI, note that the ghosting in the lines on the sides does not match the ghosting of the lines in the center in the 704x480 cap (which should be unscaled if this were an 13.5 MHz device)
So I supect the ATI has a different 'base' sample rate than 13.5 and the ITU pixels are made by scaling.

(You can see this the clearest by cutting out a line from the outher edges and pasting it under a verticle line in the center at 800% zoom)


Notice by the way I am a complete amateur and may be completely off the mark.

[erratic]

I wonder why North2Polaris gets different NTSC results. My Rage Theater chip captures 715 lines, his newer Theater 200 chip captures 704 lines. I assume we're both using the same drivers, Catalyst 3.9 (the latest).

ATI's capture software (Multimedia Center) used to support 720x480 for NTSC, but lately that has been changed to 704x480. 720x480 has been removed from the AVI capture settings. That's why I already assumed that 704x480 would be correct, but my results were different.

Is this simply due to the newer capture chip or could my DVD player's NTSC output have something to do with it?

It would be great if someone in the US could test the old rage Theater chip (ATI 8500 or older card). We also need someone in PAL land to test the Theater 200 chip.

[Arachnotron]

@ erratic

Does your DVD player have a switch to force NTSC mode? If not, it probably put out a PAL-60 signal.

Perhaps that is the difference?

edit
By the way, the timing of a PAL60 signal is the same as NTSC, only the color coding is different. Which you would not notice with a grayscale testpic. But the ATI may treat it differently. The BT878/BTWincap combination certainly does judging by my earlier results
(PAL 704, NTSC 712 but PAL60 699 )

Your use of a composite lead increases the error somewhat (perhaps a pixel off) but not that much.

[erratic]

My DVD player does not have an NTSC switch, but I can capture the test pattern as NTSC, NTSC-J, PAL-60 and PAL-M. PAL-60 and PAL-M look worse (blurrier) than NTSC though. I don't think the capture area is different, however. Still 715 lines apparently, although it's hard to tell because of the blurriness.

[Arachnotron]

My DVD player does not have an NTSC switch, but I can capture the test pattern as NTSC, NTSC-J, PAL-60 and PAL-M. PAL-60 and PAL-M look worse (blurrier) than NTSC though

I suspect the signal was PAL-60, since most PAL DVD players do this when confronted with an NTSC DVD. The additional blurry could be caused by the comb-filter kicking in. But you are correct, if the area stays the same with all those settings PAL-60 should not be the problem here.

I'm baffled by this.

If it was your DVD player you would expect the PAL value to be strange also. By the way, what kind do you have?

[erratic]

My DVD player is a Neufunk 2210S (aka Bluesky 2210S).