derkom's attempt

[derkom]

Confirmed my messy whites were just a cabling issue. It looks great now.

I fixed the serial port, installed the ACSI port, swapped out the 32 MHz crystal for the 32.084988, and attempted to jumper out the DMA buffer. I say "attempted" because the system started crashing after that. But I was using solid core 24 AWG pulled from some cat6, and that stuff wasn't reliable for logic probe connections earlier, so I imagine it's just not making good contact in the socket. Will probably wire up something better tomorrow.

[derkom]

H4-016.jpg (45.1 KiB) Viewed 1158 times

Okay, with IC6 jumpered out properly, DMA is working fine. Will replace with the bodge board later.

[exxos]

[derkom]

Yesterday I tried to implement Icky's DMA fix, without success. But I'm wondering if the problem is my build of the fix or if there's a different problem with DMA that I didn't notice before. (Foolishly, I verified that the US detected when I jumpered out IC6, but I didn't actually use it.)

My build of Icky's fix:

H4-017.jpg (113.65 KiB) Viewed 1127 times

H4-018.jpg (119.55 KiB) Viewed 1127 times

(Two different connections to FDCS are due to my debugging attempts.)

Is there anything noticeably wrong with this build of the fix? Yes, that is a 74LS04 on top, and I have verified with the logic probe that it's reversing FDCS and feeding that into pin 19 on IC6. When I was pulling FDCS in from the 1.44 GAL port, I seemed to be getting inconsistent behaviour. Occasionally bombs, occasionally scrambled device names on ACSI, occasionally good device names, but never fully working. When I moved over to pulling FDCS directly off IC5, I am consistently getting device names in HDDRIVER, but always three bombs when I run HDDRUTIL.

If this build looks good, I probably have another problem with DMA somewhere, although I have been very careful to get good solders this time. I figure I'll pull the fix out and jumper over IC6 again and to a better job of testing DMA without IC6, but wanted to see if there's anything glaringly wrong with my build of the fix first.

[derkom]

So it turns out my problem seems to have nothing to do with DMA or with Icky's fix (which does fix it), but with time. Before playing with DMA, I hadn't actually run the system for more than 10 minutes or so at a time, and it seems my H4 is stable for only about 30 minutes before things start to go haywire. If I start the system cold and run GEMBench in a loop, it'll crash after a while. Sometimes just a freeze, sometimes two bombs, sometimes three.

I've swapped all the ICs with another ST to see if it was just a wonky chip. I've moved from Falcon RAM to SIMM. I've used three different PSUs, both inside and outside the machine. Nothing is running hot except the DAC, where it's expected.

Occasionally, as the system is in the process of becoming unstable, it seems to run very very slowly before crashing (screen redraws glacial), and sometimes the mouse pointer isn't black anymore.

Very odd stuff.

[PhilC]

Just a guess, but try pushing the board in a few odd places, it sounds like something is failing due to heat. You might be able to emulate this by flexing the board in places.

Have you also tried a different psu?

[derkom]

Just a guess, but try pushing the board in a few odd places, it sounds like something is failing due to heat. You might be able to emulate this by flexing the board in places.

Have you also tried a different psu?

I have tried poking around at the board, with no luck causing a crash.

I've tried three different PSUs, two of which I know are good. The third is a recent acquisition and I haven't really tested it out, so no guarantees there.

[exxos]

Occasionally, as the system is in the process of becoming unstable, it seems to run very very slowly before crashing (screen redraws glacial), and sometimes the mouse pointer isn't black anymore.

I would suggest re-soldering everything which is connected to BG,BGACK,BR lines on the board, SIL arrays, CPU etc etc Also resoldering the MFP.

[derkom]

I would suggest re-soldering everything which is connected to BG,BGACK,BR lines on the board, SIL arrays, CPU etc etc Also resoldering the MFP.

I knew you were going to say that. I've reflowed everything so many times at this point...

It did occur to me while I was thinking about this that when I reflowed all the PLCCs, I didn't reflow the DTACK pins because I didn't want to deal with redoing that bodge at the moment. I'm going ahead and doing that right now, with little hope, but you never know.

[exxos]

I only suggested a few pins though

Assuming you are not using the blitter, make sure BR,BG,BGACK are high all the time.