Capturing VHS to DVD - Procedure check

[darkbluesky]

Hello,

I'm trying to convert a VHS recording from TV made years ago to DVD. It is a rare version of a classical (40s) movie (B&W) that is no more broadcasted neither edited in DVD (and I'm pretty sure it won't in the future neither looking at the market trends...) so before it will be lost definately, I want to made a "safe" copy in DVD.

I have been reading many times the guides on doom9.net and also 'Der Karl's Capture Cards aspect ratio for Dummies', and it seems that I have understood something about capture and specially capturing size , but before go ahead I would like to detail the process I want to do and get your advice (if I'm doing something wrong, etc):

1.- The source is VHS (NOT S-VHS), PAL B, black & white.

2.- The target is DVD PAL 720x576 (PAR 1.094, IAR 1.3675)

3.- The capture is done trough a SCART --> video stabilizer --> composite --> Gfx card composite input

4.- I own two capture cards: Nvidia MSI 5900XT ViVo (SAA71xx) & Conceptronic CTVFMi BT878 Chipset

5.- I have tested, according the explanations and Test DVD in doom9 -- http://www.doom9.org/index.html?/capture/start.html chapter 10 --, and I have checked/obtained:

• NVIDIA with their WDM 2.26 drivers capture 53.333 us
  BT878 with BTWincap 5.3.8 drivers capture 52 us

6.- If I have understood correctly, the target format (DVD PAL) will send to TV the 720 pixels wide at 13.5 MHz, i.e., it will send a width of 53.333 us, (720/13.5) which is just fine to capture with the NVIDIA (SAA71xx chip) and 720x576 according to chapter 16.4 of the forementioned link, and so sending a 1.3675 IAR

7.- If I capture the VHS with the NVIDIA ViVo (SAA71xx), at 720x576 I get a frame in which there are 9 pixels-wide black vertical bars at each side of the image, and the image itself is centered and 702 pixels wide. IF the movie was originally transferred using Pan&Scan from IAR 1.37 (35 mm) to IAR 1.33 (i.e. cropped off two thin vertical portion of the image at each side to make it less wide and so adjusting it to 1.33 IAR, but not changing the "PAR", i.e. not deforming it), I can ASSUME then that, according to http://www.doom9.org/index.html?/capture/start.html, chapter 16.3 - example 4, the PAR is already correct (DVD PAL PAR = 1.094).

8.- I want to make a version to avoid the loss of image due to overscan phenomena of CRT TV and later another version full-screen (i.e., not considering the overscan). For the moment I want to work only in the first version (to avoid loss of image).
To achieve this I should add black borders around the image with the same thickness that the areas that are hidden by the overscan. So, first of all, I must define which is the area visible in a CRT TV of a DVD.
To define this area I used the Test DVD mentioned above, and also looking into http://en.wikipedia.org/wiki/Overscan_amounts.
As a result I consider that the visible area of the PAL DVD when saw in a CRT TV is 658x540 pixels, which BTW have the same IAR than the motion area of the captured frame (658/540=1.218, and the motion area, 702/576=1.218), and so, making the next step easy:

9.- As I must resize my image (702x576) to the 'visible' 658x540, and they share the same width/height relation, it is so easy as multiply the captured frame dimensions (720x576) using a 540/576=0.9375 ratio (new_height/old_height). So I must resize to 720*0.9375=675, 576*0.9375=540. New resolution is 675x540 --> 676x540 (rounded to even).

10.- I must add black borders to achieve the DVD PAL frame size: add (576-540)/2=18 pixel in the top and also in the bottom.

11.- Add black borders to sides (720-676)/2=22 --> 22 pixel by side

12.- Now I have a 720x576 compliant PAL DVD frame with a border enough to avoid the hiding of part of the image due to oversize scanning, and with a correct PAL DVD PAR.

So before going to capture I wanted to post it here. Do you think I'm doing something bad? Maybe I have not understood the 'thing' very well?

Any comment is apprecited. Thank you (and sorry for the long and maybe boring post )!

[Capmaster]

5.- I have tested, according the explanations and Test DVD in doom9 -- http://www.doom9.org/index.html?/capture/start.html chapter 10 --, and I have checked/obtained:

NVIDIA with their WDM 2.26 drivers capture 53.333 us
BT878 with BTWincap 5.3.8 drivers capture 52 us

What are 52us and 53.333us? Is that what you paid for it?

[FulciLives]

In the past I had used a BT chipset capture card so I do have some experience with this although I have to admit it has been a while now since I have used it.

I know that using the supplied manufacturer's driver that NTSC was 688x480 (then pad to 720x480) and for PAL it was 696x576 (then pad to 720x576). I captured many NTSC and PAL videos at those settings and it worked well. My understanding is that the manufacturer's driver for BT chipset capture cards almost always uses these settings.

However I did eventually move on to the BTwincap driver and you have to capture at a different resolution with that driver for a proper aspect ratio. I know that with NTSC that you have to capture at 712x480 then pad to 720x480 for a proper aspect ratio. In fact the image gets a slight shit to the right as I recall so normally I would capture at 712x480 then crop 8 off of the left side to make it 704x480 then pad to 720x480.

Unfortunately I never did PAL using the BTwincap driver so I forget now what the proper capture resolution is for PAL when using the BTwincap driver. I want to say it is 702x576 then pad to 720x576 but I'm not 100% on that.

I think most of the Philips chipset capture cards give a correct aspect ratio at 704x480 for NTSC and 704x576 for PAL (pad both out to a width of 720) but again I am not 100% on that.

It has been a long time since I've read the doom9 guide on capture resolutions etc.

These days I use a Pioneer stand alone DVD recorder (model DVR-531H-s) and I don't have to worry about such things anymore LOL

Good Luck !!!

- John "FulciLives" Coleman

P.S.
These numbers that I have given are for capturing to and encoding at Full D1 resolution for the creation of a DVD Video.

[darkbluesky]

Thanks both for your comments.

us stands for micro seconds, the time of the video signal/wave that defines a line of PAL or NTSC image. More info on doom9 capture guide and 'Karl's Capture Cards aspect ratio for Dummies.pdf'.


About BTWincap, I agree I think it is 702x576 and pad to 720, is the only way I understand it can be done, but I must think more on it. Nevertheless, for this time I am going to use the Philips chipset (Nvidia card).

I musr think more on it though.

[FulciLives]

I did just go ahead and take a look at the doom9 capture guide. It's been a while but if I understand correctly what I just read along with my past memories of "fooling around" with this stuff then I am as positive as one can be that my numbers above (in my first post) are correct.

BTwincap is 712x480 for NTSC and I would crop to 704x480 then pad to 720x480 with 8 pixels of black on each side.

BTwincap is 702x576 for PAL and I would crop to 696x576 then pad to 720x576 with 12 pixels of black on each side.

Philips is 704x480 for NTSC and I would pad that to 720x480 with 8 pixels of black on each side.

Philips is 704x576 for PAL and I would pad that to 720x576 with 8 pixels of black on each side.

Doing all that will give you a proper aspect ratio for a DVD Video at Full D1 resolution.

Now if you want to resize to a lower resolution in order to try to minimize TV OVERSCAN then things get more complicated ... at least for a BTwincap capture.

For a Philips capture though you can easily use FitCD to figure out the resize to account for TV OVERSCAN.

For instance assuming you captured PAL at 704x576 using a Philips chipset capture card then here is what FitCD would look like:



The final script would look something like this:

AVISource("C:\TEST\test.avi")
SeparateFields()
LanczosResize(656,264,4,4,696,280).Weave()
AddBorders(32,24,32,24)

I hope this helps

- John "FulciLives" Coleman

[thecoalman]

I want to made a "safe" copy in DVD.

If it's that important don't store it on DVD, that's a lossy format. Capture using huffy.

[darkbluesky]

thecoalman: I am capturing it with huffy 2.2 but it takes 74 Gb, I can't store this in any disc, I think...I would like though. Is there a way?

Coleman: tx for taking the time. I agree with BTWincap part but I don't understand the Philips PAL. According to anex 16 (pixel aspect ratio) of the doom9 capture guide, as a DVD player send 53.333 us window and the Philips/nvidia capture (from VHS) also a 53.33333 us window, if I capture at 720x576, when the DVD player will play it (with a DVD PAL PAR 1.094), it will keep also the original PAR. Or I am wrong?

My nvidia/Philips captures a 720 wide frame, but the motion area is 702 wide though. I don't know if I make myself clear...

[guns1inger]

Try Lagarith. It can use between 30 and 50% less space than huffyuv, and is also lossless. (savings vary from video to video - YMMV)

[FulciLives]

My understanding is that the Philips chipsets give a proper aspect ratio in NTSC when capturing 704x480 and I thought it was the same with PAL (704x576) but I suppose it could be 720x576 for PAL.

My suggestion is to capture the same clip twice ... once at 704x576 and again at 720x576. Then take a screen shot of each (try to get the exact same frame) and using Photoshop (or some such similar image editing program) add 8 pixels to either side of the 704x576 capture to make it 720x576.

Then overlay the two images and check to see if there is a difference. The 720x576 will either have more info on the sides and thus "match" the image in the 704x576 (except for the missing side information) OR the 720x576 will have the same image as that of the 704x576 except the image will be stretched to fit the 720 frame (in other words 704 stretched to 720).

This will be easy to see if you do what I say with Photoshop (or whatever image editing program you use).

Let me explain to make sure that this is clear to you.

Let us say the correct aspect ratio is 704x576. If you capture at 720x576 then only 704 width is captured but then stretched to fit a width of 720.

Let us say the correct aspect ratio is 720x576. If you capture at 704x576 then you get the whole capture window minus 8 pixels on either side. If you capture at 720x480 then you get the entire 720 window.

For a test I would pick a DVD that you know (from playing on your computer) has an image that goes from one side to the next. Some DVD's are like that whereas some have a bit of black on either side (which is of no use in this test).

I hope that all makes sense

- John "FulciLives" Coleman

[darkbluesky]

BTW, it is necessary to make the resize divisible by 16, if later I'll add black border up to reach the 720x576 (which is already multiple of 16)? I can understand that the compression will work better if the block is not shared between motion and plain black, but it is that important even to change the aspect ratio?

[FulciLives]

BTW, it is necessary to make the resize divisible by 16, if later I'll add black border up to reach the 720x576 (which is already multiple of 16)? I can understand that the compression will work better if the block is not shared between motion and plain black, but it is that important even to change the aspect ratio?

If you are making a DVD then don't worry about it. Most of that block stuff is a bunch of crap if you ask me and applies more to other formats (like MPEG-4) than to MPEG-2 DVD spec.

- John "FulciLives" Coleman

[darkbluesky]

I have made the test. Both captures have the same content including two side vertical black bars (which are not in the original frame).

Each capture is 'stretched' to their respectives widths (720 or 704), though.

Personally, I think that as the philips chip capture a width (53.33 us) greater than the PAL standard width for the motion area (52 us), it captures more than the motion area, thus we have black vertical bars at each side.

So, summarizing, IMO the chipset captures always the same content, it adjust later to the width and height specified in the driver, 704, 720 or whatever.

Don't know to obtain more conclusions though.

[FulciLives]

I have made the test. Both captures have the same content including two side vertical black bars (which are not in the original frame).

Each capture is 'stretched' to their respectives widths (720 or 704), though.

Personally, I think that as the philips chip capture a width (53.33 us) greater than the PAL standard width for the motion area (52 us), it captures more than the motion area, thus we have black vertical bars at each side.

So, summarizing, IMO the chipset captures always the same content, it adjust later to the width and height specified in the driver, 704, 720 or whatever.

Don't know to obtain more conclusions though.

Your observations and my experience seem to indicate to me that 704x576 is the correct PAL resolution. If 720x576 was correct you would obtain more information on the sides but since 720 is just 704 stretched out to 720 ... well then ... that tells me that 704 is the correct width to capture at. Then just add 8 pixels of black to either side to make it 720 width for the final DVD spec video.

- John "FulciLives" Coleman

[ronnylov]

If you have followed the doom9 capture window guide and got 53.33 us capture window (which seems to correspond by the table in the doom 9 guide, SAA7108 and nvidia wdm v 30.82 with PAL) then you should capture 720x576 resolution to get a correct PAL DVD aspect ratio.

I myself have a Terratec Cinergy 400 with the philips driver and my capture window is 704 pixels for the SAA7134. So in my case I capture to 704x576.

I think you should be careful when resizing down to avoid overscan. All TV does not have exactly the same amount of overscan. Resizing will decrease the quality and if the source is interlaced it complicates the rezising also. Sometimes I crop and add black bars in the overscan area to mask out video edge garbage but I don't resize anymore since I discovered a loss in image quality.