Announcement

Collapse
No announcement yet.

Some notes on VHS resolution...

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Some notes on VHS resolution...

    I have the impression that people are often mistaking "lines" for
    "pixels" when it comes to determining the optimum resolution for
    video recording. I have often read in this forum that a resolution
    of something like 352x288 at 25 Hz would be "identical" to VHS-PAL
    resolution, or even that VHS recorders can't display more than 200
    pixels per line anyway. I have even read someone stating that NTSC
    is "defined" as 300x240 which is absolute rubbish. I would like to
    take this opportunity to get rid of these prejudices.


    The manual of my (PAL) videorecorder states that it has a
    resolution of 200 lines. So it should be able to show me
    roughly 200 vertical lines, that's either dark lines against
    a light background or vice versa. Come to think of it, there's
    a minimum of 400 PIXELS required to do that on a monitor
    screen, isn't it?

    Let's do a quick check to verify this. A TV has a line frequency
    of 15625 Hz. VHS has a system bandwidth of 3 MHz. Therefore, in
    one display line, there's room for 3000000/15625 = 192 full
    wavelengths. Each wave has a crest (white) and a valley (black)
    so we're actually talking about 192 lines here, which is 384
    pixels. There's some overhead involved, so about 350 pixels
    remain visible.

    The vertical resolution is 15625/25 = 625 "pixels" interlaced at
    50 Hz, of which 576 remain after subtracting overhead.
    Interlacing is a "brilliant" technique used to improve smoothness
    of motion, reduce flicker and increase vertical resolution by
    using "half-frames" that are displayed with a slight vertical
    offset.

    Every VCR is capable of displaying interlaced video, so there's
    no reduction in vertical resolution. My conclusion is, that
    my VCR displays 350 x 576 pixels interlaced !!!

    The Marvel offers you a similar resolution, 352 x 576 x 50 (i).

    Non-interlaced resolutions, such as 352 x 288 x 25Hz, that
    are used in programs like MediaStudio, reduce both the visual
    VERTICAL resolution and the smoothness of movement by half
    because instead of using two distinct half-frames, they
    simply display the same frame twice. The quality you get is
    visibly poorer then what a good VHS system delivers.


    I'm personally convinced that even the "half-pal" resolution
    352x576x50 (i) causes too much quality loss to serve as source
    material for VHS copies: If the output signal of the Marvel
    is comparable to VHS quality, and you copy this signal onto VHS
    tape, you'll get a second-generation copy. Comparable to
    copying VHS to VHS....


    So I advise everybody to stick to full-resolution when it comes
    to video editing!




    [This message has been edited by Flying dutchman (edited 05 November 1999).]
    Resistance is futile - Microborg will assimilate you.

  • #2
    You're absolutely right! The comparing of pixel resolution on a computer monitor to NTSC or PAL resolution on a television monitor is more than just confusing, it's sort of like comparing apples to oranges.

    One of the things that helps to understand what's actually going on with this is to grasp the factor of digital compression in all of this. Full NTSC video is more easily compared to 640 x 480 UNCOMPRESSED video on the computer. This requires a throughput somewhere above 20 MB per second on a hard drive. When you use the Matrox MJPEG compression codec with a Rainbow Runner or a Marvel, the hard disk only needs a throughput of around 3.1 MB per second to capture at a compression ratio of about 6.6 to 1. Consequently, right off the bat, the "resolution" of the compressed video is over six time LESS than that of full "broadcast quality" NTSC.

    As far as how much less VHS is than full "broadcast quality", it's something that varies wildly, from tape to tape. It's a sure bet, however, that the higher the quality of your source material, and the lower the compression you can manage on your captures, then the better the quality will be for your final products in NLE.

    Comment


    • #3
      Thank you for sharing this information! By the way, your explanation of
      the measurement method involved adds up to roughly the same resolution
      as I determined "by hand" : 264 x 4/3 pixels per horizontal scan line.

      Some things remain to be said though: The marvel is a DIGITAL recording
      device where every pixel has a fixed discrete position. Neither analog
      TV nor VHS have this limitation. Analog TV and VHS can position a pixel
      wherever they like, and subsequent digitizing it with a device of roughly
      the same resolution will introduce a positioning error of roughly 1/2 pixel
      (worst case). You can minimize this effect only by increasing the
      resolution.

      Digitizing a finely detailed vertical striped pattern, such as the
      test picture you mentioned, will also generate a phenomenon called
      "aliasing" on the marvel, which is an unpleasant interference causing
      moiré patterns. Aliasing can be prevented if the recording sample
      frequency is at least twice the highest signal frequency (the so-called
      Nyquist frequency) so I'd still recommend full-resolution for mastering...

      I also accept your explanation, that the half-pal (352x576) resolution
      is sufficient for recording from VHS tape without loss of detail, but
      that still doesn't work in the opposite direction (saving video from
      Marvel to VHS)! As far as the VCR is concerned, it is getting video
      from a good VHS master, and not from a full-bandwidth original, so the
      copy is second generation....


      Resistance is futile - Microborg will assimilate you.

      Comment


      • #4
        Hi,

        Thank you for an excellent remark.
        However, the history or resolution is:

        1. In TV, the resolution is measured in numbers of individual picture elements, but not in a number of lines that are visible as different lines on some background. So, if you take a picture with black lines and white lines going one after one, BOTH lines are counted as picture elements.
        2. The resolution is measured not for entire line, but for its part, equal in length to the screen HEIGHT (in centimeters of screen display). For typical 4:3 aspect ratio, you have to take 3/4 of pixels in line. So, 352x3/4= 264 pixels, responsible for hor resolution.

        All above is an explanation how the resolution is measured. Don't ask me WHY.

        For full screen, theoretical limit of MJPEG digital video is 532 lines, and 720x3/4=540 lines for DV format.

        In VSH, the resolution is measured as follows: If you record sine wave in BRIGHTNESS (or luma) with 100% contrast, it will have less than 100% contrast for high frequencies after retrieveing from tape. 70% contrast level (-3db) is taken as threshold. Now, count the number of dark lines and add the number of bright lines, but only on a line width in centimeters, equal to screen height. This value should be 240 for VHS, 400 for S video.

        If you create an MJPEG movie with still frame containing 100% green lines over R=G=B=150 grey background, the luma channel will have NO modulation. Only chroma difference channels will be modulated.
        Looking at his movie, you can determine the chroma channels resolution on screen.
        For composite signal, the lines of 5 pixels in width with 5 pixels distance look as slightly modulated green bar.
        On the VHS VCR output, the test pattern has not more than 50 (!) green lines that you can detect as lines on screen, or ~70 lines or "resolution", as it is determined in TV.
        Half width Mjpeg format has 264 lines for luma channel and 132 lines for chroma components resolution.

        Finally, VHS has ~120 periods of test sine wave per a portion of line that is used in resolution measurement. To gitize this analog signal, you need 240 digital samples, which NEVER decrease the resolution. Take CD audio as an example - 44.1 kHz is used for 20 kHz bandwidth. For MJPEG, the digits are 264 samples per 120 periods. Good fit.
        The ability to keep the resolution is proved by theory, however the actual result depends on HOW the conversion is done. For CD audio, the result is - no loss of bandwidth, or resolution in our terms.
        Yes, some professionals say that 48 kHz or even 96 kHz are better for CD audio, but most of us simply cannot hear the difference at all. Anyway, everybody confirms that the differentce is NOT in resolution (bandwidth).

        Note, MJPEG at half width also as a huge overhead for chroma components in resolution.

        Conclusion:
        We can use half width MJPEG successfully, without ANY loss of luma resolution, for VHS video.

        Grigory

        [This message has been edited by Grigory (edited 05 November 1999).]

        [This message has been edited by Grigory (edited 05 November 1999).]

        Comment


        • #5
          Flying Duchmen,

          A very interesting discussion on VHS resolution.

          As a novice in the area of video capture your comment at the Marvel offers a similar resolution 352 x 576 x 50 was of interest. I looked at the technical info on the Matrox website for the Marvel G400-TV but could not see this resolution option in the list! Could you clarify?


          Ernie.
          Ernie

          Comment


          • #6
            When you make audio CD from 20-20,000 Hz bandwidth source, you have only 44,100 samples per second (line for video).

            Then, if you write this CD audio output to tape, you get same 20-20,000 bandwidth with a minimal drop.
            What I try to say, is that the video signal on the output of Marvel can have exactly the same bandwidth as on its input, if the input is VHS, the digitizer has 352 samples per line, and everything is working as Nyquist said.

            As for VHS, because you copy VHS to VHS, you obviously get second copy, even if your cable has 100 MHz bandwidth. My idea is that 352 pixel per line Marvel is about the same as such "cable" in terms of resolution loss.

            Grigory




            [This message has been edited by Grigory (edited 09 November 1999).]

            Comment


            • #7
              Hi Ernie,

              this 352 x 576 x 50 resolution I mentioned is "interlaced". It consists of two 352 x 288 "fields" that are displayed with a small vertical offset.
              Resistance is futile - Microborg will assimilate you.

              Comment


              • #8
                To: Grigory.

                Ah! So that's the misunderstanding. The point is,I'm NOT copying VHS source material. I won't even try, because even a direct VHS-VHS copy looks poor in my eyes. VHS is just the target format (can't avoid that).
                My raw material is mostly S-VHS (the 400 line stuff) and partly recorded direct from TV (broadcast quality). I find the conversion losses highly visible in 352x576 format. By the way, do you know if it's possible to let the Marvel act as a TBC?

                Resistance is futile - Microborg will assimilate you.

                Comment


                • #9
                  Hi Flying Dutchman.

                  After your last post I'd like to interfer! You are wrong, if you think that you record "broadcast quality" from your TV. I know that the TV-stations use pro-gear, such as Betacam, that has resolutions around the 650 lines. It is however so, that the transmitter-gear, the transmitters that sends the signal to your arial and/or cabelprovider is only capable of sending with approx. 380 lines! So if you record something with a Super-VHS camcorder and show it directly on your TV, it will have a higher resolution than a direct broadcast!

                  Now, getting the signal via satelitte to your own sat-reciever, that's a whole other story.

                  Cheers!

                  ------------------
                  ASUS P2B-S, PII-350 (o/c to 412MHz), 128MB RAM, Cheetah 9.1 GB, Matrox Mill. G200SG, SB 64AWE, Plextor 32x CD-Rom, Sony CDU-924S CD-R, Canon BJC-7000 InkJet and Canon CanoScan 300 Scanner.
                  ASUS P2B-S, PIII-550 (o/c to 565MHz), 512MB RAM, Seagate X15 & Cheetah XL, Matrox Mill. G200SG, SB LivePlayer, Plextor 32x CD-Rom, PlexWriter PX-R820T CD-R, Canon BJC-7000 InkJet, OkiPage 4W Laser and Canon CanoScan 300 Scanner.

                  Comment


                  • #10

                    Hey, I agree with Vikingman for a change!

                    In my part of North America, our cable supplier delivers approximately 330 lines of resolution. I'd be pretty happy if I was receiving 380 lines. Of course, this might be a PAL/NTSC issue also.

                    Comment


                    • #11
                      Hell Patrick, I'm sure we can agree on more....hmmmm.....let's say women and cars!

                      Cheers!

                      ------------------
                      ASUS P2B-S, PII-350 (o/c to 412MHz), 128MB RAM, Cheetah 9.1 GB, Matrox Mill. G200SG, SB 64AWE, Plextor 32x CD-Rom, Sony CDU-924S CD-R, Canon BJC-7000 InkJet and Canon CanoScan 300 Scanner.
                      ASUS P2B-S, PIII-550 (o/c to 565MHz), 512MB RAM, Seagate X15 & Cheetah XL, Matrox Mill. G200SG, SB LivePlayer, Plextor 32x CD-Rom, PlexWriter PX-R820T CD-R, Canon BJC-7000 InkJet, OkiPage 4W Laser and Canon CanoScan 300 Scanner.

                      Comment


                      • #12

                        Yeah, I love those fair-haired Scandinavian women!!!

                        Comment


                        • #13
                          Yeah, and Scandinavian cars - RIGHT!?

                          ------------------
                          ASUS P2B-S, PII-350 (o/c to 412MHz), 128MB RAM, Cheetah 9.1 GB, Matrox Mill. G200SG, SB 64AWE, Plextor 32x CD-Rom, Sony CDU-924S CD-R, Canon BJC-7000 InkJet and Canon CanoScan 300 Scanner.
                          ASUS P2B-S, PIII-550 (o/c to 565MHz), 512MB RAM, Seagate X15 & Cheetah XL, Matrox Mill. G200SG, SB LivePlayer, Plextor 32x CD-Rom, PlexWriter PX-R820T CD-R, Canon BJC-7000 InkJet, OkiPage 4W Laser and Canon CanoScan 300 Scanner.

                          Comment


                          • #14

                            Sorry buddy, but the Italians really know how to make a car. Isn't that right, Elie.

                            Of course, I'm stuck cruising around in a Pontiac Fiero these days...

                            Comment


                            • #15
                              Yep! There's only ONE - Ferrari, but I'm talking affordable cars....

                              ------------------
                              ASUS P2B-S, PII-350 (o/c to 412MHz), 128MB RAM, Cheetah 9.1 GB, Matrox Mill. G200SG, SB 64AWE, Plextor 32x CD-Rom, Sony CDU-924S CD-R, Canon BJC-7000 InkJet and Canon CanoScan 300 Scanner.
                              ASUS P2B-S, PIII-550 (o/c to 565MHz), 512MB RAM, Seagate X15 & Cheetah XL, Matrox Mill. G200SG, SB LivePlayer, Plextor 32x CD-Rom, PlexWriter PX-R820T CD-R, Canon BJC-7000 InkJet, OkiPage 4W Laser and Canon CanoScan 300 Scanner.

                              Comment

                              Working...
                              X