I can try this later today on a Small Wonder (does it make any sense? You've already measured this one) and a Schneider EuroPC.
I also own the IBM's "Monochrome Display and Printer Adapter", but it doesn't fit in any of my XT cases. The EuroPC is obviously one of them 😀

I don't think my program will run on an MDA card. Aside from the Hercules graphics mode not working, I don't think the MDA card has a vsync bit in the status register. So my program will get into an endless loop waiting for the bit to change.

Scali wrote:

Graphics: ~23440 ticks Text: ~23923 ticks […]
Show full quote

Graphics: ~23440 ticks
Text: ~23923 ticks

Total pixels:
Graphics: 864*370 = 319680
Text: 882*370 = 326340

I don't own a Hercules card, but I find the graphics value quite odd.
The last time I did research about the register values, the horizontal total is 98*9=882 in text mode, and it should be 110*8=55*16=880 in graphics mode.

But when typing this, I realized the HGC graphics mode is set by user not the BIOS, so where did you get the values to set up the graphics mode?

Edit: I also realized the Fractint Hercules driver uses 864, but some other sources use 880.
Oh well, so one needs to write the register settings, and then measure with PIT what's the vsync rate to get the pixel clock. Then use other settings for the other pixel clock. I don't know what the monitor likes about slightly incorrect timing, but it appears you need a certain vsync rate for something. Do I smell a Hercules demo at some easter-time demo party ?-)

Be careful with timing on a real MDA monitor.. It's very sensitive.
I heard stories about joke programs which destroyed the fly-back transformer within a few seconds.
Better try it on a VGA card in HGC emulation mode first and use a LCD monitor.

Jepael wrote:

But when typing this, I realized the HGC graphics mode is set by user not the BIOS, so where did you get the values to set up the graphics mode?

I got them from the INIT.ASM file included here: http://www.minuszerodegrees.net/software/hercplus.zip
These should be the original disks included with the HCG Plus card, so they should come straight from the horse's mouth.
Indeed, the first thing I thought when I analysed the CRTC values, is that they could have increased the horizontal total by one to get 880 instead of 864.

Jepael wrote:

Oh well, so one needs to write the register settings, and then measure with PIT what's the vsync rate to get the pixel clock. Then use other settings for the other pixel clock.

Well, the pixel clock is the same, the amount of pixels per frame is different, resulting in slightly different refresh rate.
Another complication is that apparently there are at least two variations of crystal frequencies used on MDA/Hercules-compatible cards. So you have to be careful when using tightly timed routines, that you time it against the actual refresh rate of the card, and not just 'some value that happened to work on my machine'.
CGA doesn't have this problem obviously, since it takes its clock signal from the ISA bus, and therefore is always in sync with the other components running on the same clock (such as DMA, PIT and the CPU, at least on 8088 machines. Later machines may run the CPU asynchronously from the chipset/ISA bus. CGA cards probably don't even work on many later machines, if the ISA bus doesn't deliver a proper NTSC-compatible clockspeed).

Over at VCF, user Xacalite ran my timing program on an original Hercules card (type GB102):

1Hercules timing - v1.0c
2Graphics:
3Time 0: 23842
4Time 1: 23839
5Time 2: 23843
6Time 3: 23842
7Time 4: 23843
8Time 5: 23840
9Time 6: 23843
10Time 7: 22809
11Time 8: 23842
12Time 9: 23842
13Text:
14Time 0: 24338
15Time 1: 24338
16Time 2: 24337
17Time 3: 24338
18Time 4: 24338
19Time 5: 24340
20Time 6: 24337
21Time 7: 24337
22Time 8: 24339
23Time 9: 24342

This is the result of a 16.000 MHz crystal, as opposed to the 16.257 Mhz one on my clone.
The numbers seem to check out, about 1.5% difference in frame time between the two.
So if one wants to be compatible with a wide range of hardware, one should be flexible enough to handle both timings robustly.

I don't know if this is relevant, but MDA (or Hercules text mode) uses 350 active scan lines (25 text lines, 14 pixels character height), while Hercules graphics mode uses a resolution of 720x348 pixels. I have even seen a routine in a hardcopy program which uses the difference in frame time (or maybe counts the h-blanks) to differentiate between graphics and text mode. (Hercules registers are write-only so they can't be used for detection, and the BIOS doesn't know about graphics mode, it only knows the card is running mode 7.

It is correct that there are more active scanlines in MDA mode than in graphics mode.
However, the difference in frame time is not because of this, because graphics mode is compensated for this in the inactive scanlines, both have a total of 370 scanlines per frame.
The difference in timing is because the scanlines are 864 pixels wide in graphics mode, and 882 pixels wide in textmode.

So you can indeed detect which mode you're in by carefully timing a frame. I did not know that there were programs that actually tried to do this though (I wonder what use that would have, aside perhaps from some kind of debugging/overlay tool, where you want to present some additional information on screen, without changing modes).
It is not a very failsafe method however, because nonstandard graphics modes as seen in various games could throw off the timing.
It also doesn't work if you don't 'calibrate' your timer beforehand, by timing both the text and graphics modes, and recording the values. Because as we've seen, there can be different pixel clocks. If you don't know beforehand which of the two you have, you may mistake textmode for graphics mode, because the pixelclock is higher than you assumed.

Hi scali 😀
Great that you do thing's for hercules. I'll too. My machine is EuroPC(I).
Plattform TASM1.02 (Speedware).

Goals:
1. Getting Hercules-Demo Codes working
2. Converting existing CGA-StarfieldCode to run directly on native Hercules instead of running it with CGA-Emulation on Hercules.
3. ?

I have no experiences in programming, but I am interested in it.
My current problem is, finding out, why the existing Hercules-Code does not work on my machine.

Greetings
Doc