ljbk wrote:Remember that this machine might be the slowest of all (8.0071 MHz) but i would say that the music has also probably its role in this as it probably uses timers (for SID) and the problems are not constant.
Yes I am not at all surprised that the CPU speed difference would show up here. As you correctly point out the use of a SID tune in this menu meant a lot of extra work - I rewrote both the replayer as well as added some timer magic to the menu to make that work. Unfortunately while it's rock solid on my STE and STF it's obvious that it's not true for all different machines.
(The SID tune is actually the reason why I used our dual-Timer border removal instead of the currently recommended HBL-based one - my conclusion was that it suffered worse from jitter)
It might be that the solution is really simple, but without hardware to test with myself any changes are pretty much blind. Hopefully I'll get enough feedback from people who test on real hardware to at least conclude where the differences lie.
Btw, is it border removed/not removed or bitplane flickering? I had issues with the jitter triggering +2 lines but used wakeup detection and suitable EXG-replacement to "make sure" +2 couldn't happen. "It works here" ...
Besides the two old demo screens work fine: i think i saw one frame flickering during 10 minutes for the Sowatt unlimited sprites fullscreen.
I also tested the original screen with the Sowatt demo and that one really had stability problems but not with the top border. It had probably to do with the stabilizer position you changed.
Yes, the changes made to the old classics were both left border position changes as well as adding an updated and self-calibrating top border removal. The existing top border flickered on STE even after having fixed the left border. The self-calibration code in the classics run throughout - which is different from the menu (the newly coded screen) where I calibrate for 50 VBL and then lock down the timing.
It's interesting that you see instability on the old versions of the classic - I've never heard of that before. I do think there are a few scanlines that have one nop displaced switches compared to what they should have (and the rest are) and with the new knowledge about wakeups it might indeed be a case of that introducing instability in let's guess WS2 ... I'll have to try that myself.
Again about that screen, may i ask why do you have those 3 groups of 16 pixels at the top (red and dark red now (one blue with the original screen)) ?
Redhead wrote those screens and I'm guessing it's leftover debug output His TV also didn't display much of the borders IIRC.
Many thanks for testing on real hardware. I'm now very much hoping that these issues are isolated to MSTE (or at least NTSC STE) and "slow CPU" STF machines and that it works just as well on all other machines as it does on the two I have ...