Actually, the EGA card supports three timings: The CGA-type timing (15.6kHz / 60 Hz), the enhanced EGA timing (21.8kHz / 60 Hz) or the MDA timing (18.4kHz / 50Hz). The EGA card is supposed to choose one of the three timings depending on the DIP switches on the EGA card. It does so by reading the switches and storing a copy of the switches into the BIOS data area (on POST), and then using the BIOS data area to choose the supported modes. An EGA card configured to "color monitor" is not supposed to ever output MDA/Hercules timings, no matter what the color/mono switch on the mainboard is set to.

In CGA-type timing, the EGA card takes the pixel clock from the OSC pin at the ISA bus, which is supposed to be 14.318MHz. For enhanced EGA and MDA timing, the EGA card uses its local 16MHz oscillator. So I have two ideas why your EGA card might behave strange in the XT286: Possibly bad RAM causes the DIP switch copy in the BIOS data area to get corrupted (but in that case, the card should output proper mono video), or you try to operate it in a mode at CGA timing, and the 14.318MHz signal on your XT286 is bad. If the 14.318 signal is bad across the whole main board, you should also be observing that the DOS time doesn't run at the correct speed.

Well, maybe I could observe the DOS time running at the wrong speed if I could see what’s on the screen! Alright, I’ll check the 14.318 crystal.

The output at OSC seems kind of dirty, but it seems to be about the right frequency (the frequency counter on this oscilloscope seems to have large steps to it and is not very precise)

That signal looks OK. I don't think the signal is "dirty", but it seems you didn't set a proper trigger (at least in the first screenshot it clearly looks this way). The trigger setting indicator at the top right corner says "rising edge, channel 1, level 0V", which means the oscilloscope synchronizes with the point in time the signal on channel 1 crosses 0V on a rising edge, i.e. goes positive from being negative. The signal never is negative, though, so the oscilloscope doesn't trigger on the signal at all. As it is in the "AUTO" mode, it still "auto"matically captures some waveforms to help you find out why it doesn't trigger. If it were in "NORM"al mode, it would not display anything at all. You should put the trigger level to ~2.5V for this signal to be displayed cleanly - well, unless you use "AC coupling" for the trigger source, but I get the impression you didn't do that.

Anyway, as the OSC frequency seems to be correct (quantized to a period of 70.0ns instead of the expected value of 69.8ns), the OSC signal is likely not the problem. Furthermore, as your EGA card works correctly in a different system, the fault is likely not on the card. Can you show a photo of the "wrong geometry"? I understood this as "the aspect ratio is wrong", but that wouldn't prevent you from reading the time in DOS. At the moment, I ponder the idea that possibly some data bit for I/O port writes is not properly sent over the ISA bus, mis-initializing the card. OTOH, if the ISA bus is that broken, the 5152 shouldn't be able to boot as well.

mkarcher wrote on 2025-04-25, 17:36:

That signal looks OK. I don't think the signal is "dirty", but it seems you didn't set a proper trigger (at least in the first screenshot it clearly looks this way). The trigger setting indicator at the top right corner says "rising edge, channel 1, level 0V", which means the oscilloscope synchronizes with the point in time the signal on channel 1 crosses 0V on a rising edge, i.e. goes positive from being negative. The signal never is negative, though, so the oscilloscope doesn't trigger on the signal at all. As it is in the "AUTO" mode, it still "auto"matically captures some waveforms to help you find out why it doesn't trigger. If it were in "NORM"al mode, it would not display anything at all. You should put the trigger level to ~2.5V for this signal to be displayed cleanly - well, unless you use "AC coupling" for the trigger source, but I get the impression you didn't do that.

Anyway, as the OSC frequency seems to be correct (quantized to a period of 70.0ns instead of the expected value of 69.8ns), the OSC signal is likely not the problem. Furthermore, as your EGA card works correctly in a different system, the fault is likely not on the card. Can you show a photo of the "wrong geometry"? I understood this as "the aspect ratio is wrong", but that wouldn't prevent you from reading the time in DOS. At the moment, I ponder the idea that possibly some data bit for I/O port writes is not properly sent over the ISA bus, mis-initializing the card. OTOH, if the ISA bus is that broken, the 5152 shouldn't be able to boot as well.

Thanks for this info! I need to learn how to use this oscilloscope more effectively.

And now, I present to you: the real problem all along. I even tried a different cable early on — didn’t suspect the gender changer. I began to smell something fishy when it didn’t even work with my 5150+CGA haha.

Thanks for your help anyway, mkarcher. Now at least I know more about how the EGA card gets its pixel clock.