User:Mjb/video capture hardware

From Offset
< User:Mjb
Revision as of 19:05, 27 December 2020 by Mjb (talk | contribs) (+horizontal resolution)
Jump to navigationJump to search

Digitizing videocassettes involves a playback device (pick one):

  • VCR with composite output
  • S-VHS VCR with S-video output
  • 8mm camcorder with composite and/or FireWire output
  • VHS-C camcorder if you have those types of tapes

It also requires a capture device (pick one):

  • Digital8 camcorder (will output DV encoded video & PCM audio via FireWire)
  • internal PCI or PCIe capture card
  • external USB-based capture "card"

VCRs notoriously output unstable signals with lots of horizontal and vertical jitter and "flagging" which CRTs were very forgiving of. But capture cards require a relatively stable signal, otherwise they may drop frames, lose audio sync, or just not work at all. The stability problems can be partially or fully resolved with the help of a time base corrector (TBC). One may be built into some VCRs and some camcorders. TBC or TBC-like functionality can also be found in some DVD recorders.

Time Base Corrector (TBC)

  • Getting hard to find as a separate item. Very expensive. They were $100+ to build and now run ~$1000 on eBay.
  • A TBC doesn't clean the image. The main purpose is to ensure picture is stable (no up/down image bounce, i.e. "jitter"), distortion-free (good horizontal alignment of scan lines), and to output a constant video signal for capture; without it your pic may be jumpy and the capture card may drop frames. maybe not necessary. (DigitalFAQ forum admin "lordsmurf" insists it is necessary but he is a perfectionist)
  • Some DVD recorders have line TBC or something close to it, and you can use them as filters without invoking the DVD recording functionality at all. Best ones are the Panasonic DMR-ES10 and DMR-ES15. I bought a DMR-ES15. One forum user says a Philips DVDR3576H is even better.
    • The DMR-ES10, ES15, and ES25 are in the same league as one another. Tests show some very slight degradation of color, but even that seems to be well within the tolerances of analog gear and not noticeable outside of test signals. I mean, two different tapes, VCRs, capture devices, or sources could easily have the same amount of color difference.
  • Some S-VHS VCRs have actual TBC + noise reduction, but quality varies greatly and they don't do all the things a real standalone TBC does. Good overviews here: http://www.digitalfaq.com/forum/video-restore/2251-tbc-time-base.html http://www.digitalfaq.com/forum/video-restore/1853-alternative-avt-8710-a.html#post9889
  • It is recommended to use a TBC in conjunction with an S-VHS VCR that also has a full-field TBC, ex: Panasonic AG-1980P, although JVC is better. JVC models with TBC include HR-S7500U, 7600, 9500, 9900. However JVC is also reportedly "soft" in image quality. (I agree! at least with Picture Quality mode set to Auto. Other settings for this mode are Soft, Sharp, and Edit. Edit is no/minimal filtering, but also means more noise. Ref: http://www.digitalfaq.com/guides/video/capture-playback-hardware.htm and http://www.digitalfaq.com/forum/video-capture/4296-jvc-s7500ek-pal.html )
  • Some Digital8 camcorders have TBC but it may just be a line TBC (horiz. cleanup only). Still, better than nothing!
  • My JVC HR-S3600U S-VHS VCR has a "Video Stabilizer" toggle in the settings. According to https://forum.videohelp.com/threads/286055-VCR-buying-guide-(S-VHS-D-VHS-Professional)#post1746696 this is not a TBC, but does do some vertical stabilization. In my experience, it does not help at all. Leave it off unless you test and make sure it works better for a particular tape.
  • A real TBC: AV Toolbox AVT-8710 (a.k.a. the Cypress CTB-100 outside USA) is cheaply made, overheats after 1.5 to 5 hrs, freezes last good field when encountering a bad one, may weaken color, adds dot crawl and ghosts, etc.
  • A real TBC: DataVideo TBC-1000 is the gold standard, but expensive

S-VHS VCR

  • Some of them include a TBC and various bells & whistles
  • Professional or prosumer preferred, from mid-1990s to mid-2000s. Panasonic post-1995 not so great.
  • Panasonic AG-1980P has a full-field TBC, best ever included in a VCR, possibly with integrated color noise filter, but only available used, probably needs its remote, very expensive. Works well with EP/LP and hard-to-track tapes.
  • Panasonic AG-5710 supposedly is the same but w/o tuner, not easy to find.
  • S-VHS is a format that encodes more scan lines on VHS tapes, resulting in more detail, but with more color noise. Nothing to do with S-Video, a type of output that sends luma and chroma separately. However, S-Video outputs are mostly only found on S-VHS VCRs.
  • JVC decks (maybe just the cheaper ones?) have poor S-Video implementation, especially on upscaled 240p content like the on-screen menus, but also on tape playback. The luma and chroma aren't adequately separated. The crosstalk manifests visually as checkerboard/diamond/grid/crosshatch pixellated fuzziness, also called dot crawl. See discussion at http://forum.videohelp.com/threads/331681-s-video-artifacts?p=2209636&viewfull=1#post2209636 (and note differing analysis in subsequent posts; could even be a connecting cable issue). It may be possible to get better results by capturing this jankety S-Video and then processing it with a comb filter. I would like to see proof of that.
  • Some recommendations here (says JVC 2000-5000 series are to be avoided): http://www.digitalfaq.com/guides/video/capture-playback-hardware.htm and http://www.digitalfaq.com/forum/video-restore/1567-vcr-buying-guide-2.html#post17383
    • JVC SR-MV50US, SR-MV55US
    • JVC S-VHS & W-VHS (broadcast) USA models with TBC have TBC+DNR in one function, normally a button on the front panel.
    • JVC D-VHS USA models supposedly do not have TBC, just DNR. They also add "DigiPure" functions. TBC may exist and just be always-on. No agreement on VideoHelp forum.
    • JVC 9911 = great TBC/DNR, unreliable mechanism
    • Panasonic AG7650 = not recommended
    • Mitsubishi HS-HD2000U = good

Horizontal resolution

You may have noticed that SD capture hardware normally outputs a frame that is 720 pixels wide. This frame is supposed to be 4:3, but it's anamorphic—slightly stretched for NTSC or squished for PAL—because the number of scan lines is fixed and produces non-4:3 ratios. This happens because of the Rec. 601 video capture standard, which dictates a particular horizontal sampling frequency which produces slightly rectangular pixels.

This may lead to the question of how many pixels are needed to fully capture the horizontal content. It turns out that the number is quite a bit lower than 720.

There are no pixels in analog video, just scan lines which compose a visible 4:3 frame surrounded by various synchronization signals. The horizontal resolution of the visible part of such video is rated in a very poorly defined standard of discernible "lines", which you'd think would just equate to pixels, but it doesn't. The measurement is subjective, with a huge margin of error, and the math behind it doesn't quite match up with real-world performance.

Testing resolution properly is hard because you can't know for sure that a videotape, even one you made yourself, has the same resolution for recording as for playback. A simple way to estimate, though, is to capture the most detailed VHS video you can find, and see how much you can horizontally shrink it before fine details start to smear. You should see that VHS captures will require luma resolution of no more than 320 to 360 pixels wide. VHS chroma resolution is no more than half that number!

S-VHS offers slightly under twice the VHS horizontal resolution for luma, so will probably need around 600 pixels.