exxos's Atari, Amiga & retro forum

I managed to hack it to start the diag cart by driving EDTACK low when address ff8940 is hit. Then it continues on past that. Sadly stuck somewhere further on though, before anything on RS232. Still where its next stuck might give me more info.

foft wrote: 14 May 2022 14:47 Yes, one exists for Falcon? I only knew of diagrom for the Amiga.

Terriblefire did a Atari port. Its on the forum somewhere. It runs with a 030 CPU, but not sure if anyone has tried it on a falcon yet.

Some progress :)

After fretting for some time about having to order and replace the SDMA chip I decided to double check all the signals going into it.

I noticed BMODE is a bit glitchy, so tried reburning/replacing U68. It has the same glitch.

Several clock pins all seemed valid, then I noticed the main one CPUCLKA on pin 110 is missing :oops:! I'd inadvertently knocked a wire off the bus speeder mod, so there was no longer a connection to R221.

Now I fixed that its much more alive, not surprisingly! The diagnostic cart starts successfully, at least if I press down on the ram module - the SIMM socket is damaged and the 3 elastic bands aren't enough...

Now the diagnostic cart boots and I get the menu via RS232 and video on-screen. The keyboard works, the sound test works.

The blitter test on burn-in seems to just hang, though perhaps I just need to wait longer?

I optimistically tried to boot without the diagnostic cart and just get a grey screen. I don't have a battery yet in the RTC, could that cause that?

One more, thing. This means I have a BERR or IPL issue most likely right?

Checking FPU...
EF FPU not found
Initializing PSG...Ok
I6 Checking exception handling...
Falcon 030 Production Diagnostic - v1.27b
1993, Atari Microsystems
...

foft wrote: 16 May 2022 21:29 The blitter test on burn-in seems to just hang, though perhaps I just need to wait longer?

Not sure about the burn-in test, but the blitter test in the second menu ('D', 'G', from memory) is virtually instant.

I optimistically tried to boot without the diagnostic cart and just get a grey screen. I don't have a battery yet in the RTC, could that cause that?

TOS4 will get in a strop without a functional NVRAM chip, but the battery being dead is not normally a problem. Sometimes the chip contents is inconsistent and needs resetting, but I believe somebody mentioned that the diagnostic cart actually does that for you.

EmuTOS 512k variant (should you have a burnable 27C4096) will tolerate more things not working and give more useful feedback too.

BW

TOS works now! After I took off my FPGA board I forgot also to put back on W11.

Still some issues: some diag cart tests not working and the bus speeder crashes it (Bus Error starting cart: FF8936 access error at PC FA03A6)

Also I need to cut some of videl off to reconnect EXT CLK, it has 3 pins missing, I thought I'd managed to solder this one to the stub but hadn't...

I must say, I'm really impressed by your hardware debugging skills!

But for anyone else, I think this is a nice example why not to overcomplicate things. From what I could read, most of the damage could have been fixed by a far simpler diagnostics:

- new RAM
- fixing bent/missing pins on the Videl/SDMA
- replacing the CPU (if it ever was faulty and not just damaged by the reflowing attempt?)
- replacing GALs (ok, this could have been tricky to diagnose without a LA yet easy to do even as a futile attempt)
- putting the jumper back ;)

But if time is not your enemy... you've gained incredible knowledge here!

I definitely learned a lot and definitely not very efficient at all!

I think the CPU was broken from the initial symptoms, though I do now wonder if it was actually the 2nd broken GAL (U63) that was the issue. As a I reminder I replaced the Centurbo GAL to fix the clocks then replaced U63, which handles DTACK, after I'd removed the CPU.

I'm keen to pursue the FPGA board on expansion, though that'll have to wait until the chip crisis is over. My ribbon cables are too long to be reliable at this clock frequency I think. @terriblefire said he might make a 68060 in HDL;) I'm tempted too, though its a lot of work...

Still got plenty to learn on this project:
i) Work out how to expose the lead frame on VIDEL without damaging it, so I can connect EXTCLK. I tried a rotary tool on another chip (before using the real thing) and it was a nightmare!
ii) Work out why I get some bus errors when using the bus speeder. The diag cart menu does run if I switch it on while its running.
iii) Figure out why FPU doesn't work properly when I put it in in the socket (its a tested one from Exxos, so problem is here)
iv) reliable pad replacement with epoxy and uv solder mask
v) Play with some Falcon software now I have a working Falcon :D

I should also add: Many thanks for all the help on this.

exxos wrote: 14 May 2022 14:59 Terriblefire did a Atari port. Its on the forum somewhere. It runs with a 030 CPU, but not sure if anyone has tried it on a falcon yet.

I did once. It didn't work for me, but I didn't escalate it as I think my particular configuration assumed entry point at E00030 and it wasn't.

Does work in Hatari IIRC.

BW

Now looking at the bus speeder - Centurbo 3b.

The CLK clock is driven by COMBEL and the SDMA clock is from a gal pin on the speeder. Not entirely sure why, since they are the same clock. Perhaps for a phase shift? Or perhaps he just had them separate since he was planning to try a different bus and cpu clock? They are listed on different lines here: file:///home/markw/Downloads/CENTurbo-1_Evolution-3b_FR.pdf. (Also found this: http://strider.untergrund.net/toxicmag/ ... nt_ev3.htm)

There is a low pass filter between the GAL that drives the clock for the SDMA chip. Its the 33ohm resistor and 470p capacitor here:
http://powerphenix.com/rodolphe.czuba.f ... /CT1e3.jpg
At 16MHz I get pk-pk 4V (no load) and at 25MHz I get pk-pk 2V. I guess the latter is too low for SDMA to use.

fc = 1 / (2 * π * R * C) = 1/(2*pi*33*470*10^-12) = ~ 10MHz.

I can't see how that'd ever work at 25MHz, it cuts the voltage too much surely?

Someone had adjusted the board here in the past. The trace from capacitor to ground was gone and a solder bridge to the ground plane (scraped off) replaced it. I tried to remove the cap to check it was really 470pf, but managed to break that!