Image Map Image Map
Results 1 to 9 of 9

Thread: CGA to S-Video; has anyone done it?

  1. #1
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,901

    Default CGA to S-Video; has anyone done it?

    Lately I've been thinking about various strategies to adapt the CGA output from my Tandy 1000s to more readily available monitors on the cheap with better quality than the composite output, which generally looks pretty crummy. Looking at the schematics for the system I kind of expected the composite signal would emerge as a single pin from the "Big Blue" video ASIC, but instead it's constructed in the same way as the composite output on an IBM CGA card. (I looked up the schematic to check.) The composite output intensity is just generated by a resistor ladder with the following values:

    R: 2.2k
    G: 1.1k
    I: 750
    B: 3.3k

    Also mixed in via resistors is a composite sync signal (which comes from Big Blue on the Tandy but is generated from vsync/hsync via an XOR and an inverter on the CGA card) and the colorburst signal. Then everything is fed through an output transistor.

    In short, I was wondering if it might be be a worthwhile improvement in signal quality to remove the colorburst signal from the resistor ladder and use it as a separate chroma signal to create an S-Video-style output. I was thinking I could either do this internally by cutting the resistor to reroute the colorburst signal before it gets to the existing output, or I could leave the internal circuitry alone and reconstruct the luma resistor ladder on an external board that plugs into the RGBI output and run a parasite clip-on lead to the buffered colorburst output coming out of the 74LS244.

    Beyond using a transistor to create an impendance matching amplifier for the chroma output is there anything else I need to do to convert the CGA colorburst into according-to-Hoyle s-video?

    I realize of course this isn't going to look as good as a full-blown RGB output but I do remember with game consoles that the jump in quality between composite and s-video tended to be a lot bigger than the improvement going from s-video to component. So... yeah, I am wondering if anyone's tried it on a PC before?
    My Retro-computing YouTube Channel (updates... eventually?): Paleozoic PCs

  2. #2

    Default

    There's no harm in trying, and in some objective sense you will almost certainly get output that is higher quality than composite (in particular, 80 column text should be much more readable) but lower quality than RGBI. However, pretty much any piece of graphical software that you try this with will look "wrong". The direct colours will be the composite ones rather than the RGBI ones, but you won't get artifact colour in games designed for composite output.

  3. #3
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,901

    Default

    Yeah, I figured I probably wouldn't get artifact colors anymore, so I would probably want this mod to be reversible via a switch. (Or if I did an external replication of the resistor ladder and just tapped the colorburst signal it might be possible to run both at once, which would allow a monitor with both composite and s-video inputs to be switched between them.) I know the composite direct colors are "wrong", in terms of not matching the RGBI set, but I was playing with having my Commodore 1084 and 1702 running side-by-side connected via RGB and composite respectively and if you fiddle with the tint knob you can at least make them close-enough-ish. (And since Tandy 1000s have the 16 direct color modes that are fairly widely supported greater clarity in them would somewhat make up for the loss of artifacts.)

    (Speaking of artifacts, I have noticed what's been reported plenty of times before, IE, the artifact colors a Tandy 1000 displays are quite wrong compared to an original CGA card. It'd be nice if there were some way to fix that but I suspect there's no user-serviceable knob for doing so, at least on these later generation Tandy 1000s where the colorburst generation circuitry lives inside a great big ASIC.)

    Greater clarity in the text modes is really the win I was hoping for with this. I was testing the clarity of 80 column text with color disabled (mode bw80) with the composite output connected to the Luma input on the 1702 and I was genuinely surprised how clear it is. If an "s-video" mod would be able to colorize that level of clarity then it might make the machine genuinely useable on a more modern TV. A side-problem this might also solve is one I've I've run into with having a monochrome monitor hooked up via composite. (I sometimes use an old Apple IIc green screen as an auxiliary when I need two machines running at the same time.) It seems like there's quite a lot of text software doesn't honor "mode bw80" and turns the color back on when it starts, so the text gets trashed by the colorburst instead of coming out as a clean grayscale. (I've seen some vague references to TSRs that can work to force it off, but I haven't actually been able to track one down yet.)

    An alternative idea to trying to use the existing colorburst signal is I did find that there are a lot of similar RGB-to-NTSC encoder chips out there, a common one is the Analog Devices AD724. In theory at least it it should be possible to combine of of those with an RGBI-to-RGB circuit that includes the "brown fix" to get the RGB color set over S-Video, and the resulting widget would just plug into the RGB port of any CGA card. Maybe it'd be a smart idea to do the "hack" first to get an idea of clarity.
    My Retro-computing YouTube Channel (updates... eventually?): Paleozoic PCs

  4. #4

    Default

    I disconnected the sub-carrier from the composite output that feeds the internal Amber monitor in my 5155 (its derived from a custom feature adapter on a Video-7 card).

    IBM did not attempt to make the composite color outputs as seen on a composite color monitor display exactly like the RGBI does on a CGA monitor like the 5153. Also they got the grey scale off a little due to an error in the resistor matrix.

    Regardless of the small changes made to the resistors on the feature adapter (which are just like those on IBM's EGA or CGA cards), so as to get a good linear grey scale for a monochrome monitor, there was really no comparison between what a CGA color monitor would produce compared to what would be seen on a composite monitor. Although for the color composite monitor, the color or hue would match fairly well with the CGA monitor, but the luminance levels associated with the colors would not match the CGA monitor's image.

    There are some color & monochrome images I took in this article:

    http://worldphaco.com/uploads/FITTIN...N_IBM_5155.pdf

    In any case, there is fairly severe interference of the color sub-carrier with the luminance signal causing patterning effects in the composite signal (which is aggravated by the wider than usual bandwidth of the lumininace component, because this is much wider in a computer system's composite video out, than a television system) so if you separate the sub-carrier off the luminance signal or better, don't have it mixed on there in the first place, it has to be a good idea.

  5. #5
    Join Date
    May 2011
    Location
    Outer Mongolia
    Posts
    1,901

    Default

    Quote Originally Posted by Hugo Holden View Post
    IBM did not attempt to make the composite color outputs as seen on a composite color monitor display exactly like the RGBI does on a CGA monitor like the 5153. Also they got the grey scale off a little due to an error in the resistor matrix.
    Thank you for posting the link to your PDF. I happened to have the machine fired up here so I've been doing some quick comparisons between the composite direct colors and RGB (have both feeding into the 1084 and flipping the input switch) and I think your observations about the chosen values for the composite resistor ladder being off are definitely on-point. Tandy used slightly different values than IBM did but I would say that it shares the same problems of blue being too "hot" and insufficient distance between normal and intense. (Although I will say that at a given brightness setting on this particular monitor the composite output itself seems like it's running on the hot side. If I tune the brightness so CGA's dark gray color 8 is quite dim in RGB mode and flip to composite it becomes about as bright as RGB's color 7, and color 7 is about as bright as RGB 15.)

    If the resistor ladder was fixed I think you could probably tweak it to be *fairly* close to the RGB color set in brightness at least, although I assume that the "Brown" color would still be dark yellow. (And red would probably also remain its more magenta/pink-ish tint. Presumably fixing that would require tweaking the colorburst generation circuitry itself.)

    Maybe it'd be worth ordering a few of those AD724 chips just for laughs. They seem to be pretty cheap on eBay. (Much less cheap from Digikey.) I think I already bought the parts to make a brown-fixing RGBI-RGB adapter to try out with one of those lousy GBS scaler boards anyway, tacking the AD724 onto it would be an interesting side quest.
    My Retro-computing YouTube Channel (updates... eventually?): Paleozoic PCs

  6. #6

    Default

    Quote Originally Posted by Eudimorphodon View Post

    Snip...

    Greater clarity in the text modes is really the win I was hoping for with this. I was testing the clarity of 80 column text with color disabled (mode bw80) with the composite output connected to the Luma input on the 1702 and I was genuinely surprised how clear it is. If an "s-video" mod would be able to colorize that level of clarity then it might make the machine genuinely useable on a more modern TV. A side-problem this might also solve is one I've I've run into with having a monochrome monitor hooked up via composite. (I sometimes use an old Apple IIc green screen as an auxiliary when I need two machines running at the same time.) It seems like there's quite a lot of text software doesn't honor "mode bw80" and turns the color back on when it starts, so the text gets trashed by the colorburst instead of coming out as a clean grayscale. (I've seen some vague references to TSRs that can work to force it off, but I haven't actually been able to track one down yet.)

    ..snip
    I had this exact same problem hooking up a monochrome Amdek to my CGA Deskpro. I ended up writing my own TSR to fix this issue:
    http://www.vcfed.org/forum/showthrea...ite-monochrome

  7. #7

    Default

    Quote Originally Posted by Eudimorphodon View Post
    If the resistor ladder was fixed I think you could probably tweak it to be *fairly* close to the RGB color set in brightness at least, although I assume that the "Brown" color would still be dark yellow. (And red would probably also remain its more magenta/pink-ish tint. Presumably fixing that would require tweaking the colorburst generation circuitry itself.)
    I have some information in another article about the pink issue and the yellow vs brown. With a conversion I made to get a 5153 monitor to display EGA signals it required a color 6 fix circuit, its on pg 16 of the pdf, but it was more about getting an EGA signal to display correctly for at least 16 colors:

    http://worldphaco.com/uploads/INCREA...OR_TO_AUTO.pdf

  8. #8

    Default

    Quote Originally Posted by Hugo Holden View Post
    IBM did not attempt to make the composite color outputs as seen on a composite color monitor display exactly like the RGBI does on a CGA monitor like the 5153. Also they got the grey scale off a little due to an error in the resistor matrix.
    I believe the second revision of the IBM CGA card had the composite colors purposely changed in order to show up better on the 5155's internal monochrome display.

  9. #9

    Default

    Quote Originally Posted by vwestlife View Post
    I believe the second revision of the IBM CGA card had the composite colors purposely changed in order to show up better on the 5155's internal monochrome display.
    Yes what IBM really needed was two composite video outputs, one for an external composite monitor and a duplicate one (with so color subcarrier) for the monochrome monitor in the 5155. It would have taken only an extra transistor and a few resistors.

    The monitor in the 5155 has an excellent luminance frequency response and the video amp in it is good to well over 7MHz, which is why it has good sharp graphics, but they are ruined when the subcarrier is imposed too.The difference, with and without the subcarrier being present on the feed to the 5155's monitor is shown on page 18 of the pdf:

    http://worldphaco.com/uploads/FITTIN...N_IBM_5155.pdf

    So on their early CGA cards you can just disconnect the resistor that mixes in the subcarrier (that is if you don't want to use an external colour composite monitor). But in my 5155, because I got the video for the amber monitor from a custom feature adapter on a video 7 card, I simply just omitted the sub-carrier in the resistor matrix.

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •