Raven. A homemade Atari-like computer

[redpixel]

I don't have anything particularly clever to add other than I have the same card showing up in ISA_BIOS and working.
That's not saying there couldn't be some kind of incompatibility from Raven or ISA_BIOS but at least a data point showing it working in one machine.
We could probably make an ISA_BIOS with extra debug logging, and maybe a new Nessi firmware with a few more ISA waitstates if that's the issue.

Thanks for the good suggestions! I bought the card just because I saw that you also had it so I was sure that you would do magical things with it

I tried to add 2 extra wait states in Nessi, just slapping them onto the "end", i.e. counting up to ISADTACK22 instead of ISADTACK20. That did not make any difference. I also tried to replace the 68150 with another one (the one I normally use is FN40 and I tried to replace it with another FN33 that I have). Also no difference.

Then I tried to use different bus/CPU oscillators, and boom, it showed up! So I guess it is somehow timing related, but it doesn't seem to help adding wait states at the end. My testing looked like this:

Osc	Nessi 251029	Nessi +2 waitstates
25 MHz	Detected	Detected
32 MHz	No boot	No boot
40 MHz	Detected	Detected
48 MHz	Not detected	Not detected

CPU was at X2 in all cases. I have normally used the 48 MHz oscillator. In the No boot case it hangs on InitVga (I'm using an ATI Mach32 2 MB, I believe exact same model as you have also... That was by chance though!)

Not sure if you have any ideas what could be wrong, otherwise I can see if I can hook up my logic analyzer and try to measure a bit and see if I can make any sense of it. I think I might have to order one of those ISA extender PCB's for that though, to be able to have some accessible probe points.

[mikro]

Probably what needs to happen now for a more polished experience is:

- detach mouse cursor redraw from main loop.
some gfxcards has support hardware cursor but for best compatibility I think a software solution like in ScummST would be first priority.
(just redraw the cursor at every VBL if changed, essentially)

- detach music from main loop.
avoid midi/adlib timing jitter as well as stalls if mainloop is busy for a length of time (loading etc)

Perhaps not accidentally, I have looked into same things lately (although only on a theoretical level). VBL based cursors gets *very* tricky once you want to keep triple-buffering in place. Of course, not impossible to overcome and especially if you have true hardware cursors available, that way you don't have to solve dirty rects.

MIDI timing is a bit easier because audio code is assumed to work in a separate thread but then there's the issue I had been fighting for a while (and put it on hold to look into less frustrating things) -- how to actually keep the multithreading part while not stepping into the complex locking/unlocking territory.

But maybe both issues need just a fresh pair of eyes and you are definitely the best candidate for that.

Btw, while waiting for my family at Brisbane airport:

PXL_20260604_210635204.jpg (173.83 KiB) Viewed 1215 times

[kodak80]

Btw, while waiting for my family at Brisbane airport:
PXL_20260604_210635204.jpg

Wow, that is a big coincidence. It's a sign...of good things to come.

Also, guess I won't be getting that rego.

[agranlund]

Btw, while waiting for my family at Brisbane airport:
PXL_20260604_210635204.jpg

Haha, that is amazing!

[agranlund]

Yay! Raven rev.A2 arrives
Probably wont have time to solder one up until next weekend or so but was still always exciting to open the package and seeing the pcb as a physical thing rather than a project inside Kicad.

raven_a2_pcb.jpg (268.16 KiB) Viewed 1049 times

Fingers crossed everything then actually works as intended too

[dml]

Nice!!!

[Cyprian]

looks great,
I'm very curious about the DSP itself

[PhilC]

Look at all that black and gold

[LarryL]

wow - looks extremely beautiful!
Oh my God — how can I stop myself from giving in?

[agranlund]

Slowly, most of the top-layer SMD components are in place.
I had forgotten how much work it was to assemble it...

a2_building_02.jpg (244.78 KiB) Viewed 758 times