VideoHelp Forum




+ Reply to Thread
Results 1 to 13 of 13
  1. I am interested in the frame size of NTSC captures. It seems to me using iuVCR's built in BT8x driver, one should cap at 712x480 or 754x480 for DVD. This should then be cropped/padded or resized to 712x480 and then cropped/padded to DVD-video size.

    Reading Houspig's guide on LordSmurf.com, it is recommended that one cap at 720x480. However, the example shots on iuVCR site show caps at 768x576. 768 corresponds to the 754 number above. 768 is the max width at the BT8x chips sample rate for PAL. To get back to the original picture aspect, a PAL user would size down to 702 and pad.

    Any thoughts?

    Am I Just crazy? :P

    Nobody care?


    By the way: As far as caps at 720 vs 352. I'd hazzard to say 90% of the capture cards out there capture the full picture and then resize to the width you ask. SO the only "quality" difference is if you like the cards resize or not. If you want to get the full info back from the card without a resize you need to ask for 754 or 768. If you want the source picture at it's original width, you need to ask for 712 or 702.
    Quote Quote  
  2. Member
    Join Date
    Dec 2002
    Location
    United States
    Search Comp PM
    Don't mix PAL and NTSC standards.

    Anythingx576 is PAL. Anythingx480 is NTSC. (by anything I mean the usual resolutions, not some oddball like 690 pixels)

    You should be looking at 720x480 or 704x480 for NTSC captures. How much overscan you want to deal with is up to you. I usually capture in 720x480 and go from there.
    To Be, Or, Not To Be, That, Is The Gazorgan Plan
    Quote Quote  
  3. Thanks for the comments.

    However, I mentioned PAL and NTSC in my post. I was not mixing. I was covering both.


    Speaking only NTSC:

    Your capture card captures the active NTSC picture as 754 pixels width. That is because it samples at a higher rate than the 13.5 in the standard. For the standard, the active picture would be 712. The card then shrinks the 754 to the width you asked for. If you ask for anything other than 712, the width is too small or too big. The picture is distorted (not very much). 720 is too wide. 704 is too narrow.


    Speaking only PAL:

    The capture width is 768. The picture width in a correct 4x3 aspect is 702 TV pixels.


    Here is a link to the BT8x cards specs: http://www.conexant.com/products/entry.jsp?id=404

    Here is a link to a BBC writeup on picture sizes: http://www.bbc.co.uk/commissioning/branding/picturesize.shtml

    Here is a good definition of Aspect Ratio: http://www.quantel.com/domisphere/infopool.nsf/HTML/565C7D89F2D64F3680256C8000444A4D


    Ever wonder why 720x576 is not 4:3? Well, 768/576 is. Exactly 1.33333333333. And would look good on a square pixel PC. 702 would look good on a fat pixel PAL TV. That is the number on the iuVCR website example page.


    NTSC is a bit more complex. OR maybe I am crazy :P

    Edit: Sorry, the BT8x spec is a doc on that page. 100199a.pdf. Here is a direct link: http://www.conexant.com/servlets/DownloadServlet/100119a.pdf?FileId=542
    Quote Quote  
  4. Member housepig's Avatar
    Join Date
    Jan 2003
    Location
    the Plains of Leng
    Search Comp PM
    Originally Posted by trevlac
    Reading Houspig's guide on LordSmurf.com, it is recommended that one cap at 720x480. However, the example shots on iuVCR site show caps at 768x576.
    iuVCR's developer is in PAL-video land... I am in NTSC-ville, and all my sources are NTSC, which is why I made the suggestions in my guide.
    - housepig
    ----------------
    Housepig Records
    out now:
    Various Artists "Six Doors"
    Unicorn "Playing With Light"
    Quote Quote  
  5. Member
    Join Date
    Dec 2002
    Location
    United States
    Search Comp PM
    Given the source of my captures, 712 or 720 doesn't matter. I usually put TV episodes in 1/2 D1 resolution anyway, so 704 is better since I'm going to decimate by 2 for speed. Usually I mask the outside to black then re-size. My TV is 27" and it doesn't show. In fact, I see the artifacts from the cable TV a lot more than I ever see encoding artifacts.

    NTSC is a pain in the butt compared to PAL.

    And Yes, IUVCR has a definate PAL flavor to it (the BTWIN drivers force me into PAL only mode).
    To Be, Or, Not To Be, That, Is The Gazorgan Plan
    Quote Quote  
  6. housepig:

    I don't mean to argue with you. I was just trying to better understand a strange "fact". Tell me to suck eggs if you don't care.

    Translating from 768 PAL gives 754 NTSC. This is also for PC AVI watching, not DVD TV watching. If you want to translate from PC to TV watching, you would get 702 for PAL and 712 for NTSC. See the BBC link above for some PAL info.

    Actually you can cap at 768x480 because the drivers support that big of a width. I see the initialized buffer on the right when I do this. The card does not provide any info for the driver to put there. It's the famous green color.


    Now the differences are not much. I will gladly keep my thoughts to myself if no one cares. For practical purposes, one might use any number that makes making a DVD easy.

    Edit: To clarify/correct, 754 is wrong above for NTSC PC watching. 648 would be the correct number for this (648/486 = 1.33333). 754 is the max width the card can give for NTSC. 922 is the PAL number. The drivers may not support 922 as a PAL number. And if you want to be really anal, 711 is the NTSC TV pixel number not 712. 711/486 = 1.33333333 711 and 486 are not good numbers for digital. You get 486 because 6 are chopped off of the TOP of the picture. At least that is what I have read. I did not confirm.
    Quote Quote  
  7. Originally Posted by Gazorgan
    Given the source of my captures, 712 or 720 doesn't matter. I usually put TV episodes in 1/2 D1 resolution anyway, so 704 is better since I'm going to decimate by 2 for speed. Usually I mask the outside to black then re-size. My TV is 27" and it doesn't show. In fact, I see the artifacts from the cable TV a lot more than I ever see encoding artifacts.
    You are VERY right. It doesn't really matter much. I just wanted to see what people were doing. I find the easy 1/2 D1 method (using virtualvcr) is to cap at 368x480 and then just crop to 352. This throws the aspect off by 12 pixels vs the 8 you mention. Fitting NTSC into 1/2 D1 is 712 / 2 = 356. You must then crop 4 more to get DVD spec. My method is ment to let the card do the resize (which is what it does. it grabs 754 and shrinks it to 368). All of this can be done in virtualVCR at the cap step, using the built in crop filter.

    Why 368 ? Well for some reason, doing anything less is mucked up by the card's resize. Maybe it's the driver (BTwincap) that I use.

    Does 352 give you poor results, but 368 look ok for you?

    I have a sample of the bad resize. It actually keeps the detail. It just seems to mess up the fields. I'll post a link if I can find it.

    Trev
    Quote Quote  
  8. Member
    Join Date
    Mar 2003
    Location
    Uranus
    Search Comp PM
    Have you guys completely forgotten that the player
    resizes the horizontal to the DAR so discussions of
    Sample Aspect Ratio don't mean much ?
    Quote Quote  
  9. Originally Posted by FOO
    Have you guys completely forgotten that the player
    resizes the horizontal to the DAR so discussions of
    Sample Aspect Ratio don't mean much ?
    What are you saying? For NTSC analog out? The player goes from 720x480 to 711/486 using a resize? I doubt it. I'd say the player just outputs the 720 centered on the scan line. It then overlays the edges with blanking/sync signal info required to be in "analog" spec. Finally, it adds a few lines at the top or bottom to make 711/486. Picture putting a 711/486 frame over a 720 (or 704) by 480 picture. The player has to convert from the digital spec to the analog one.

    Have you tested what you are saying? In a test I'd bet various DVD players do the frame on pictue thing I described above in slightly different ways. They may throw off the center line. They may pad on top or bottom.

    Don't get me wrong. I am just a hobbiest who likes to understand the details. For practical purposed, maybe noone cares.

    PS:

    A more interesting thing about picture aspect ratio may be that for 30 years movies were shot in 1.37 720x480 is real close to this. To put this on a dvd one would just pad a 720x479 image by just 1 line and then send it to DAC. So (for old movies) is 1.33 really 1.37 seen thru a 1.33 frame?

    The math for the above is:
    720 * 72/79 = 656 x 479 = 1.36994.
    711 * 72/79 = 648 x 486 = 1.33333.

    72/79 is a conversion from PC square pixels to NTSC skinny pixels. BTW: PAL pixels are short and fat. (is that why all of those actors on the BBC are short and fat? )

    See the bbc link or this one for reference: http://www.quantel.com/domisphere/infopool.nsf/HTML/565C7D89F2D64F3680256C8000444A4D
    Quote Quote  
  10. Member
    Join Date
    Mar 2003
    Location
    Uranus
    Search Comp PM
    I'm simply saying that the aspect ratio of the capture has nothing to
    do with the display aspect ratio.
    neither of 352 x 480 and 720 x 480 is 4:3, yet they both display
    as 4:3 on a TV.
    If this discussion is about overscan , that's a slightly different matter.
    I don't worry about it. maybe I should.
    Does a captured line have more pixels than will display on a TV
    when put on a DVD ? possibly
    In that case the capture is getting more than is visible, and should be
    adjusted.
    Quote Quote  
  11. Hey FOO,

    Let me 1st say I am relatively new to this stuff. However, I have been recently reading the official ITU specs, running some tests, and piecing this together. The specs are free to download and shorter than you would think. I've also been pondering specs from the BT8x implementation of this stuff. These can also be had for free in a pdf.

    Originally Posted by FOO
    I'm simply saying that the aspect ratio of the capture has nothing to
    do with the display aspect ratio.
    neither of 352 x 480 and 720 x 480 is 4:3, yet they both display
    as 4:3 on a TV.
    Let's be clear. The aspect ratio of a picture/image is width/height, measured by a ruler. The pixel frame size (ie 720x480) is a measurement of a picture/image in pixels. If pixels are not square, measurement in pixels must be adjusted to make sense. I think the "squareness" of pixels is referred to as SAR. 1/1 is a square pixel. NTSC TV pixels are 72/79 or about .911. PAL TV pixels have a SAR of 128/117 or about 1.094. See my links above for a confirmation.

    Getting this out of the way, we can then calculate the picture aspect ratio of an NTSC picture in square pixels. 711 TV pixels wide by 486 lines is:
    711 * .911 by 486 == 648 x 486 == 4/3 == 1.33...

    So to bring it on home ... If you have a card that captures the active scan line and then scales it to what you ask (like the BT8x does), then not adjusting it back to the original apect ratio distorts the original picture/image.

    Said another way: 711x486 is 4:3. Capturing a 711 image at 720 streatches it.

    I don't think 720, 704, and 352 are the aspect ratio of an picture/image. That is dependant upon the source. The frist 2 numbers are just the picture size (in non square pixels) that the dvd player shows thru the 711x486 frame of a TV. Padding and cropping to get to them does not effect the aspect ratio of the picture. 352 is probably streatched to 704 and then just shown with black fill by the dvd player.

    If this discussion is about overscan , that's a slightly different matter.
    I don't worry about it. maybe I should.
    Does a captured line have more pixels than will display on a TV
    when put on a DVD ? possibly
    In that case the capture is getting more than is visible, and should be
    adjusted.
    My understanding of overscan is with regard to what you can not see on the edges of a retail TV. To be most efficient with regard to seeable picture, one would cap at 712, crop to 704. If you think black borders compress better than picture, you could then crop up to 5% more off of each side and add back black to get back to 704. With VirtualVCR one could do the cap at 712 and crop all you want in 1 step. This would get you the smallest file size. You'd then have to add back using something else.

    Regards,
    Trev
    Quote Quote  
  12. Member
    Join Date
    Mar 2003
    Location
    Uranus
    Search Comp PM
    I don't know what you are trying to say.
    What I'm trying to say is the capture rates horizontally and vertically
    are not uniform, or "square" but that is not related to what comes
    out of the DVD player .
    No matter what the horizontal pixel count is, the player
    will spit out a full scan line that fits the TV width.
    Quote Quote  
  13. Thank you for bearing with me. I am not being clear.
    Originally Posted by FOO
    What I'm trying to say is the capture rates horizontally and vertically
    are not uniform, or "square" but that is not related to what comes
    out of the DVD player .
    Here I disagree. It is related, and it is a rational process. (see below) The point is D1 is not what you may think. 1/2D1, 2/3D1 are a little different but also very rational.

    No matter what the horizontal pixel count is, the player
    will spit out a full scan line that fits the TV width.
    This may be true, but the picture will be distorted if you don't follow the correct relationship/process.



    The following represents the process of DV filming a round ball, broadcasting the image, capping it, and putting it on a DVD at D1. In order for the ball to continue to be round, the steps in the process must be correct. I'm saying not capping at or resizing to 712 with iuLabs (or BTwincap or maybe any BT8x) drivers, breaks the process. The ball will not be round.

    In order to follow it is easiest to visualize the initial picture/image passing thru a series of frames as it travels thru the process. I give the frame measurements in square and {non-square ITU-601 NTSC} pixels. The point is that if you stretch the picture after it passes thru a frame, you will be distorting the picture. Your opportunity to do this is between when you capture and put the image on a DVD. If you assume the frame size is not what I give, you are distorting the picture. The frame aspect does not really matter. The frame just crops/pads the picture as it passes thru. I give the frame aspect just as a reference.

    Code:
    Image passes thru       Frame Size           Aspect of Frame
    
    1)ViewFinder            648x486 {711x486}    1.333
    2)DV Spec               656x480 {720x480}    1.367
    3)TV Broadcast          648x486 {711x486}    1.333
    4)Capture Card          649x480 {712x480}    1.352
    5)DVD Spec              656x480 {720x480}    1.367
    6)DVD out to TV         648x486 {711x486}    1.333
    Notes:
    1. This is artificial. I picked this as a clear starting point. There really is no pixel measurement until you go to digital. An LCD on a cam probably is 4:3.
    2. DV spec captures more picture width than would show thru a TV (with no overscan), and probably more than you see thru the viewfinder. I imagine TV cameras line up with broadcast specs, ITU-470.
    3. The 601 spec says the with of a 720 image is 53.333 microseconds. Analog TV broadcast spec ITU-470 says the active picture is 52.666 microseconds. To go from one to the other you effectively crop the picture. Analog images are taller than digital images. You have to add 6 lines. (Top or bottom or both, I don't know)
    4. BT8x cards pull the active picture as 52.740 microseconds. This is probably because 712 is a much nicer number than 711, and the analog spec gives a +/- value. After all, it is analog. The chip also samples at a higher rate than the spec (14.296 vs 13.5). That is why you can ask for up to 754. The card also chops off 6 lines. (Top or Bottom, I don't know). If it's the same black lines as the DV to Broadcast conversion added, there would be no picture lost. It seems it is. VHS tapes may vary. If you cap at 720 you are effectively asking the card to cap a 712 line and then stretch it to 720. Some drivers may be smarter than you an limit what you do to keep the aspect. I don't think the iuLab or BTwincap drivers do this. To test your drivers you could cap at both 720 and 704, pull a similar frame out in virtualDub, and see if the "active picture"s are the same. If they are, the driver is adding a bigger boarder to a 720 cap and cropping the 704 cap. If they are not, the 720 is stretched, and the 704 is squished. The existence of a boarder is not enough to know. The source may have a boarder.
    5. Going to DVD spec, one needs to pad or crop to 720 or 704, if your drivers did not do it for you. You would also set the DAR to 4:3. Your only other real option here is 16:9. If you did that, the DVD player would try to stretch the picture.
    6. Your DVD player (thru some analog out, not DVI) has to convert from ITU-601 spec to ITU-470. It sees the DAR as 4:3 and knows the picture is not anamorphic. It then centers the picture on the 53.333 scan line and starts overlaying the edges with analog sync stuff. This will effectively crop the picture to a 52.666 width (711 pixels). It also has to make up for the missing 6 lines. I don't know if it adds them at the top or the bottom.



    Since I typed all of this up, could you respond with some type of comment?

    Thanks

    Trev
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!