Also taken the keyboard controller socket off so that it's easier to give the area a good clean.

I'll give it a good clean with IPA again, and then rub down all of the light corrosion and tarnishing with some fibreglass pencils, before fitting the new caps, BIOS and keyboard controller sockets and replacement keyboard port. Then re-test again.

Cleaned the remaining corrosion from around the battery area:

There's only one broken trace that I can find - the obvious one from the keyboard socket, which should be an easy repair. The rest has cleaned up nicely. I'll get the DIP sockets replaced and then fit those new caps and see if we get a fully set of lights on the POST card.

Good job 😁

Nice!

I have a fully working version of this board with MR-BIOS installed and everything. If there's anything you need me to check something out, let me know. 😀

I love this board.

DIP sockets fitted, new caps in place, and a new non-corroded keyboard port soldered on:

Quickly retested:

I have +12v again (hurrah!) and +5v, but no -12v. RESET is being held and CLK is, well, the best description is intermittent.

What I'll probably do next is go through and replace all of the other capacitors of the type that were next to the AT PSU header.

Keep at it, you are giving me inspiration to get a 286 I have working. They are surprisingly simple boards these old PCs

You turned this into a shiny beauty (again). I love hopeless cases with happy endings.

My regular AT-PSUs don´t start when there´s a short or overload at any of the rails.
But when I´m in a hurry I connect one of my more powerful ATK-PSUs to the board (that doesn´t refuse to start with a shorted tantalum), turn it on and watch the culprit go up in coloured smoke. This saves time, but there´s a risk of burnt traces and the stench lasts for days. I wouldn´t recommend this method for any valuable board or one that you have spent a lot of time on.
If there´s an interruption in the -12V line a look under the AT connector might be useful.

Recently I was working on an unknown Socket 7 board with unknown jumper settings. There was a 2-pin connector labeled "JP 7", all headers for Reset, power on etc. were "J XY", so I put a jumper on "JP7" for a test, turned it on and the 12V traces all the way through all ISA slots and PCI slots went up into smoke. "JP 7" was in fact the fan connector for a CPU fan - labelled by mistake as a jumper. I had to rewire all slots from the backside.

Thanks for the encouragement!

I've replaced all the electrolytics I can - but I'm short a few of them, I thought I had a bag of 20 or so, which would be more than enough, but turns out I only had 10, so I've got another 3 on this board to do.

Looking at the AT PSU header I think that is more likely to be the (power) culprit; if I plug and unplug the PSU a couple of times, I can get +12v, -12v and +5v all to appear, so it's either a dodgy connection or some corrosion on the pins that I cannot see. I'll see if I can find another header from somewhere.

In the meantime I need to figure out what is causing the stuck RESET signal.

I've found several traces on the back of the board that are cut, near the cache sockets:

I don't know if that will have anything to do with the RESET signal, but they will definitely need repairing - at least two of them have no continuity.

At least PC-chips cannot be blamed for that.

Traces at the edge of the board can be for anything. I had to learn that due to the automatic routing of the pcb design software some traces that are lower frequency / low current are routed in a way that they make lots of detours just to allow more importand traces to be short and have low inductivity.

Repaired the traces - not my finest work, in fact 5 years ago I probably would have scoffed at the quality of the repair, but I find that what once I could do fairly easily is now pushing at the very limits of what I can achieve... it's not as if I'm close to retirement, I'm only 42!

Anyway, the traces are repaired and there's continuity again between them (and no bridges). No difference to the operation of the board: +12v, +5v, -12v steady, RESET held on, CLK pulsing irregularly.

Oh, and also completed the change of those remaining 10uF caps - so they've all been replaced now.

Another thing is clear, the 14MHz clock signal is anything but clean - it's bouncing all over the place and I don't get any measureable signal from it; certainly not a nice 14MHz clock.

The 32KHz clock is there, but the 14MHz signal is nowhere to be found. I'll try removing that can and replacing it with one from another board.

Confirmed. The 14MHz can was dead - I've quickly desoldered another 14MHz part from another ISA motherboard, and:

+12v, +5v, -12v all steady as before, but CLK is now steady, FRAME now steady. RESET lights for a second and then goes out. Perfect!

My cheap Hantek scope now sees a 14MHz signal from the oscillator:

It's still not getting to the point of getting any POST diagnostic codes, but it's certainly progress from where I was previously.

-

Thought I'd try to read the BIOS ROM to see if it is corrupt, blank, whatever, but it looks like either it is a device that isn't supported by the TL866, or there's something wrong with it.

Tried both 27256 and 27512 style devices, usually the TL866 will say, "Ah, you asked me to read from a Motorola 27c256, but it's really an Intel 27128"... this time it just doesn't know what these devices are:

$ minipro -r file.bin -p "M27512@DIP28"
Found TL866CS 03.2.86 (0x256)
Invalid Chip ID: expected 0x200D, got 0xFFFF (unknown)
(use '-y' to continue anyway at your own risk)

$ minipro -r file.bin -p "M27256@DIP28"
Found TL866CS 03.2.86 (0x256)
Invalid Chip ID: expected 0x8904, got 0xFFFF (unknown)
(use '-y' to continue anyway at your own risk)

So, I peel back the label on the top of the ROM and:

Classy, PC-Chips, real classy. FFS. This is some knock-off ROM instead of an industry standard 32kbit or 64kbit EPROM. Will the money saving never cease?

Anyone had luck reading one of these wierdo ROM chips?

I had one of those, it's a cheap OTP EPROM I think. Just ignore the lack of ID bytes, as long as you are not writing the Vcc is 5V for all of these. Obviously you won't know the size but generally I just pick the biggest device for the pinout and then look inside the resulting file to see if there are any mirrors of the data. Or if half is blank/noise. Then I just adjust the file size. Note the valid data doesn't have to start at the zero offset in that case.

Hmm, I get nothing back if I force read it as either 32k or 64k device. Just a file of empty bytes:

$ minipro -r file.bin -p "27C512@DIP28" -y
Found TL866CS 03.2.86 (0x256)
WARNING: Chip ID mismatch: expected 0x89FD, got 0xFFFF (unknown)
Reading Code... 1.34Sec OK

$ hexdump file.bin
0000000 ffff ffff ffff ffff ffff ffff ffff ffff
*
0010000

I've got a few (probably fake) Winbond W27C512 eeproms coming. I'll give it a try writing one of the available M321 BIOS images to one and see if that starts things up. The fact I can't seem to read anything from this chip is concerning.