Announcement

Collapse
No announcement yet.

ATI TV Wonder or Rainbow Runner

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

  • #16
    Hi Paul,

    Most people will output to mpeg if they plan on showing their clips on systems that don't have the hardware to decode the clip for playback. If you want to show your clip on a vcr, then don't worry about it. Capture with the hardware codec and dump out your clip to a tape.

    Hard drive space may be cheap but a lossless video system isn't

    Ideally, you want to capture with as minimal compression as you can without sacrificing quality. The only way to do this is:

    - Capture in RGB (uncompressed) but this means you need a system that can handle ~ 20Mbytes/sec sustained and gigs a gigs of hard drive space, depending on how long you want to capture for.

    - Capture with some kind of hardware codec that will not sacrifice your quality while at the sametime, using minimum compression to save on HD space and bring down the sustained thru put problem.

    You can use any hardware mjpeg or mpeg2 card for this.

    - Get a firewire card and dump your video via firewire to firewire onto your system. Firewire keeps your data in the digital domain and the datarate is a joke even with DV50 rated capture cards.

    However you are correct about the effects. For each effect, overlay, video filter, etc, the editor has to create a new frame which means something is getting recompressed. Only way around this is to get a real time card that actually works.

    Depending on people's budgets, they will get a mjpeg card and then convert it to mpeg format. I'm personally not crazy about this because this means that a software is recompressing and re-scaling my original capture which will result in some kind of quality loss if I capture at full res and then convert it to mpeg 1.

    However, you can bypass the re-scaling if you convert to mpeg2 but this brings in another problem.

    If you capture in 704x480 or 352x480, you are capturing both fields. By standard, mpeg2 displays both fields. Since most software converters are cheap, they will not convert both fields. So in reality, all you have is an mpeg1 clip with a faster data rate.

    To solve this, you can either get an expensive converter package that actually converts both fields or use a dvd authoring tool like dvd it.

    I use dvd it with my RT and the final quality is really is extremely close to a lossless clip.

    Haig

    Comment


    • #17
      Hey Haig,
      why doesn't Matrox close a deal with Ben Rudiak-Gould over his HuffYUV codec (to include with a possible new product for instance). It works very effectively and is "lossless" (you can use 24bit or YUV, YUV ofcourse being lossy but not in the extreme since most >MJPEG/MPEG steps do this conversion anyway). I've been using it for some months now, and I 've never been this happy with any MJPEG or MPEG codec before!!! What's more: it's for free and the source is too!!! (maybe you can arrange a really nice deal with him to develop a special Matrox version in exchange for some hardware & codec-support) Datarates can easily be handled by a modern IDE drive for half res capture, full res will probably work on the faster models just as well... The author uses a celeron at 416mhz himself, so it should work fine on most PII/III systems...
      If you'd combine this in a package with LSX Encoder (be sure YUV input is possible) you'd have a real winner
      http://www.math.berkeley.edu/~benrg/huffyuv.html


      [This message has been edited by Stormlord (edited 13 September 2000).]
      --

      David Van Dromme
      A.K.A. Stormlord/WOW
      Former C64 Scener and Advanced Gravis P'n'p betatester.

      Comment


      • #18
        Hi Haig,

        Thanks for the explanation.

        With the both fields bit, you are talking about capturing from an interlaced source? I'm using a progressive scan DV camera, which I think means the image is already non-interlaced.

        I have been trying to figure out what the trade off is here. If I have an interlaced source, then I have lower resolution for each field because each field only contains half the lines in the image, but greater accuracy in the time domain because each field is truly spaced at 1/50 of a second (for PAL).

        However using progressive scan mode on my camera, I get a non-interlaced image at 25fps (well, 50fps, but every second frame is identical) - so the image is sharper but movement is not as smooth because I'm missing that extra information about what went on between each 25fps frame.

        So I was thinking that maybe I should be switching my camera out of progressive scan mode, into standard interlaced PAL, as the smoother motion may be worth the trade off.

        But, reading through my editing software( Adobe Premier) with an interlaced source, my understanding is that it discards one field anyway. When I output to mpeg2 (in this using my capture card) the image is re-interlaced, but since the interlaced field was previously discarded it is simply the other lines of the 25fps imaged that was de-interlaced by the software.

        If this is the case, then I might as well be recording using progressive scan?

        Thanks,

        Paul

        Comment


        • #19
          Hi Paul,

          AFAIK, very few DV cams can do progressive scan. Most of them export by dropframe by default even if they can do progressive scan. You can check your manual to verify this. The Canon Elura can do progressive scan as can the higher end Canons and the Sony's.

          Importing the ps to the hd shouldn't be a problem if you use generic firewire importation but the problem will be blending them into 29.970 fps dropframe.

          I think that VirtualDUB might be able to handle this. If not, then Premiere &
          Ulead MSPro 6.x might to if the clip were rendered to fieldB at 29.970 in your final step.

          With the both fields bit, you are talking about capturing from an interlaced source? I'm using a progressive scan DV camera, which I think means the image is already non-interlaced.

          Correct the image has no fields just frames.

          Haig

          Comment


          • #20
            Why would you guys want to use mjpeg anyway?
            The mpeg2 chips are still too expensive to use on cards aimed at the mass market especially for an all in one type of card.
            Using a software codec isn't that bad though. Even if you don't have a powerfull system, you can still capture at half res with a PIII 600.
            ... and then came the Marvel G400 eTV. Why would you want MJPEG when you can have MPEG2, right Haig? Of course, you land up buying a new video card. Will their be a RR-like add-on to the G450? Isn't clear if this card is using a HW or SW codec. I assume its SW due to the P3-600 minimum requirements. If so, how does this card assist MPEG2 encoding/decoding? What is IBP format and what affect does it have on video quality? I assume it uses a lossy compression algorithm. How much does it compress the data stream (having already undergone MPEG2 compression)?
            <TABLE BGCOLOR=Red><TR><TD><Font-weight="+1"><font COLOR=Black>The world just changed, Sep. 11, 2001</font></Font-weight></TR></TD></TABLE>

            Comment


            • #21
              Nope,

              no RRG type of add ons slated nor any type of video add ons slated. Mpeg2 is totally software. The PIII 600 spec is for half capture. If you want full res capture, you need at least a PIII 800. If we had motion comp on the board, then this would lower the specs but we don't.

              I can answer the rest of your q's once I get a chance to play with it myself

              Haig

              Comment


              • #22
                Anyone see any reviews of this new beast yet?
                "Whoa..."
                Keanu Reeves.

                Comment

                Working...
                X