BitWrangler wrote on 2023-01-28, 03:59:

You've got me wondering what happens if you stick one of those 14.375ish Mhz crystals in there, whether it synchs on a TV or video monitor. Then maybe it has overscan?? IDK

My head is also going toward the simpler Amiga scandoubler plans and the stuff Grant Searle has done for composite display, using atmels, where maybe it could intercept the signal digitally then videofy it.

These are some interesting thoughts! 😁 👍
Since MDA/MGA are 1-Bit in its simplest implementation (pure mono video, omitting intensity for now), an A/D and D/A conversion for TTL<->Video might be possible without so much trouble.
I mean, a framebuffer in its simplest form (holding a 1:1 bit pattern of zeros and ones) would do. Timing/sync are a different story, perhaps, needing some dedicated glue logic.
But I'm optimistic the process of "digitizing" would very well be possible with little code. CGA would be more complex, at least in 4 colour modes, I guess.

Speaking of Arduinos/Atmels, here's something interesting:
External RAM with Atmel: https://spritesmods.com/?art=avrcpm&page=2
Compsoite (VBS, actually) with Arduino/Atmel 328p: https://nootropicdesign.com/video-experimenter/

BitWrangler wrote on 2023-01-28, 03:59:

You've got me wondering what happens if you stick one of those 14.375ish Mhz crystals in there, whether it synchs on a TV or video monitor.

An MDA is not going to synch on a TV or video monitor by just changing the crystal to 14.318 MHz, for multiple reasons:

• The horizontal timing parameters for mode 7 and mode 3 differ.
• If you manage to find the correct crystal frequency (If I calculated correctly, it is 13.783MHz) so that the horizontal timing of an MDA card is dropped from 18.4kHz to 15.6kHz, the vertical frequency will
  have dropped from 42.4Hz, whereas NTSC video timing requires 60 Hz. This is due to the vastly higher scan line count of the MDA mode (350 visible lines) compared to TV timings (around 240 visible lines).

I'm still gonna wonder what happens though. Analog ain't all that notchy, it's like bellmouth vases where if you flip a coin and it lands on the rim it will slip in, even though it's inches off the perfect center. Though yah, it ain't necessarily gonna be repeatable across a range of different CRTs

I've seen a number of old (early 90s and I think some late 80s) VGA and SVGA monitors with the specs including a minimum H-frequency of 29 or 30 kHz. A fair number will probably tolerate a bit lower than that, as with the 25 MHz crystal. (though 26 MHz would probably be a more reliable bet = 29.5 kHz and 79.7 Hz)

You should also be able to get a paper white B/W output by running video to R G and B lines. Since this is 1-bit mono, the issues of weighted RGB for grayscale output also doesn't matter. (though this is really an issue specific to RGB outputting to a grayscale monitor, like the IBM PS/2 Model 25 used, or various industrial VGA monitors: I had used one in my first PC my dad built me around '93 or '94)

I wonder if something similar will work with overclocked Atari ST video output, but for RGB modes rather than mono (since Mono is already SVGA compatible at stock speeds).
There's various synchronous overclock mods that usually halve the GLUE chip (that provides lower clock rate sources and video synch and blank signals), but it would be a matter of leaving the GLUE at synchronous speed and only isolating a few of the lower clock rate things (which I the STe already does). Using the 50 Hz PAL mode with 64 MHz oscillator would result in 31.25 kHz 100.16 Hz 312 line video (with 200 lines used, or up to 284 with overscan mods ... and some other limit via software overscan). 58 or 56 MHz would give 28.32 or 27.34 kHz.

Then again, 312 lines might be stressing compatibility more than 370 lines of Hercules.

But then, as with the Herclues mod, any software using the v-sync/blank (V-scan rate) timing would run too fast. Presumably a lot of games and demos would fall under that.

Hmm, for Hercules compatible games, as I recall there's a few graphic adventures that support it, and probably some wireframe or polygonal 3D ones.

Hmmm... I am wondering if it will be easier to convert Hercules video to composite (combine TTL video, intensity, hsync and vsync signals and get it into the 1 Vpp range), then feed the resulting signal into a cheap AV2VGA converter. The ASIC in these things (MS1851) can process both PAL and NTSC, but can it handle 18kHz? That's the question.

itsemast wrote on 2025-05-26, 06:48:

Hmmm... I am wondering if it will be easier to convert Hercules video to composite (combine TTL video, intensity, hsync and vsync signals and get it into the 1 Vpp range), then feed the resulting signal into a cheap AV2VGA converter. The ASIC in these things (MS1851) can process both PAL and NTSC, but can it handle 18kHz? That's the question.

Both PAL and NTSC work with 15 kHz HSYNC, so I wouldn't expect it to support any other frequencies.

Also, there's the problem of signal quality - CGA 640x200 already looks worse via the Composite output than via the RGBI one, with MDA/Hercules 720x350 the difference would be even greater.

Yes, that's right, just found the datasheet, and it advertises support for NTSC(J,M,443) and PAL(B,D,G,H,I,M,N,Nc,60), so I guess, it would fail with MDA.

Some CRTs with composite input can be tweaked to lock on 18.432Khz, so converting to composite would make more sense. I am thinking of translating TTL video and intensity signals to 0.7 Vpp, then feeding them to the inputs of AD724/AD725 along with Hsync and Vsync, and using it's Luma output as composite signal (basically, grayscale video with Csync, no chroma). The encoder datasheet also states that it needs either PAL or NTSC crystal, but it is needed for the color burst generation, which is discarded in case, so maybe it will work.

Grzyb wrote on 2025-05-26, 08:29:

Also, there's the problem of signal quality - CGA 640x200 already looks worse via the Composite output than via the RGBI one, with MDA/Hercules 720x350 the difference would be even greater.

Not without the colour carrier, I think. Plain VBS looks fine and is the RGB equivalent of the monochrome world. :)
Back in the day, VBS was the native video signal of b/w TVs and b/w video monitors.
Originally, video monitors had used SO-239 "UHF" connectors, before BNC and RCA/Cinch/Phone connectors got common.

Here in Europe, to my knowledge, many IBM PC compatibles had on-board CGA without NTSC circuit.
The monochrome video signal via Cinch was high-def, it was being used to drive green monitors and amber monitors. IMHO.

Jo22 wrote on 2025-05-26, 19:32:

Here in Europe, to my knowledge, many IBM PC compatibles had on-board CGA without NTSC circuit.
The monochrome video signal via Cinch was high-def, it was being used to drive green monitors and amber monitors. IMHO.

I've seen a clone PC with a dedicated clone CGA card (i.e. not on-board) that had two cinch jacks. One is grayscale and the other one color. A green monitor was connected to the b/w output, and displaying a nice sharp image. If you were connecting it to the color jack, it would show strange high-frequency dot patterns.

Also, I really like how your Winbond Hercules card looks with mostly SMD components on it. Unfortunately, I can't find any high-res pictures of this exact card, I have attached the ones of a similar card that I found online, but they are not good enough to make out all the traces. Can you take some bore pictures of both sides of your card with good lighting?