No one here familiar with the inner workings of the 5151 or the different Hercules modes?

MMaximus wrote:

I've also tried with a dual 15-pin 9-pin 8-bit VGA card, supposedly Hercules compatible - the display won't sync even in text mode, although this card is working in Hercules mode when connected to an LCD through an MCE2VGA converter.

Never connect the 5151 to a signal source different than 18.4kHz horizontal. The 5151 has no own oscillator and an incorrect frequency can kill it easily.

MMaximus wrote:

Compared to my other 5151, the power-on behavior is different. Whereas the first 5151 seems to give an accurately sized picture right from the start, this monitor starts with a very horizontally compressed picture, which actually takes a few seconds to "expand" to its normal size. Is this the sign of an underlying problem? Apart from this, the display quality is quite good.

Maybe some short in the horizontal circuit that opens when stuff expands while getting warm? Capacitor or transformer.

Are sure the various cards are actually outputting an 18.4KHz Hercules signal? I've had VGA cards that were Hercules compatible in that they would display Hercules if you selected it in the software, but they would still be outputting a 31.5KHz VGA signal.

dr.ido wrote:

Are sure the various cards are actually outputting an 18.4KHz Hercules signal? I've had VGA cards that were Hercules compatible in that they would display Hercules if you selected it in the software, but they would still be outputting a 31.5KHz VGA signal.

Winbond W86855AF is a very common Hercules Graphics Card clone, so yes, it surely outputs typical MDA/HGC 18.4 kHz horizontal, 50 Hz vertical - at least in normal conditions, ie. in BIOS video mode 07h, and normal Hercules 720x348 graphics mode.
Some CGA emulation software may program the CRTC to other frequencies, which indeed can be harmful for IBM 5151.

Regarding the slow horizontal picture expansion, measure B+ voltage on cold power up, if voltage drifts up, check all filter caps in B+ supply area.

Grzyb wrote:

Some CGA emulation software may program the CRTC to other frequencies, which indeed can be harmful for IBM 5151.

Back in the 80s, burning up MDA monitors for fun by using software was a common and accepted kind of sport.

Thanks for all the input so far - I plan on trying the winbond card with my other 5151 monitor when I get access to it sometime next week and then report back.

Update - just tested the Winbond card in another machine with my other 5151 monitor. Prince of Persia works correctly as I suspected. So I guess there's a problem with my newly acquired 5151, but I still don't understand why Test Drive 2 would work and Prince of Persia doesn't. Do they use different resolutions? If anyone has an answer to this it would be appreciated!

Note: picture is blurry because my phone didn't focus - it actually looks quite crisp in person.

Timings for graphics modes can be slightly different, depending what the respective application writes into the 46505/6845 registers.
Back in the 1990s I modded a number of HGC clones to feed the Eizo 4051 monochrome monitor to get 90+Hz refresh rate.
There the only thing to do was to accordingly adjust the horizontal scan pot (or was it a coil? don't remember clearly anymore).

If there is no defect in the PSU (see what @SSTV2 wrote - you have an intact monitor to compare with), then you'll have to adjust the LC constant in the monitor. Don't remember now whether the 5151 has adjustment there, if no, then probably adding a small capacitance to the flyback capacitor will fix that.