Check if the legs of the flat pack are all connected to the board. Maybe there is one lose of the pcb connection.

You could try an Deoxit threatment over the whole board.. ( your board looks clean from corrosion, but it might help otherwise.

Or the oscillator might be bad / non functional.. (due age )

Next is to use a tool like a oscilloscope to find clock signals ect.

Caps can also been dried out.

Would also check the memory chips or they are good or faulty.

Robin4 wrote on 2022-11-30, 23:17:

Check if the legs of the flat pack are all connected to the board. Maybe there is one lose of the pcb connection. […]
Show full quote

Check if the legs of the flat pack are all connected to the board. Maybe there is one lose of the pcb connection.

You could try an Deoxit threatment over the whole board.. ( your board looks clean from corrosion, but it might help otherwise.

Or the oscillator might be bad / non functional.. (due age )

Next is to use a tool like a oscilloscope to find clock signals ect.

Caps can also been dried out.

Would also check the memory chips or they are good or faulty.

Thank you. Yes, the board had been cleaned by the guy who gave it to me. And I cleaned and checked again the battery damaged area.
-- Memory chips are now out. The POST card should show something even with no memory inserted.
--
I poked the legs of the chipset IC one by one but those all seems to be soldered firmly.
--
Yeah, the OSC may possible be faulty but I have CLK on ISA which I am able to check even with my advanced DMM and it is derived from the main clock I think. I need to install a socket to be able switching OSC - for sure.
--
Capacitors - the tantalums checked for shorts. However there are planty of electrolytes. I just took out one to test it and it measure good. ESR 1.6ohm and 11uF (it is good for a 10uF CAP)..
--
What is weird is if I take a needle probe and check the voltage on the VCC of the 80286 I see 4V (sometimes a bit lower) which I think is weird. It is a 5V CPU.
--
I also checked the CLK pin on the CPU but it is out of range of my DMM so no frequency reading. The VOLTAGE level on the CLK pin is 5V. Is this ok??
--
I can check the other small "auxiliary" OSC and they both read good.
--
The CPU had been tested by the guy before in another board.
--
Is it OK the 8042 KBC is warm after some minutes of running? Not hot. Just luke warm. The CPU i warm as well. Chipset as well and also the 74xxxx chips by the bios got a bit warm.
--
It is possible it can ´t post with no CMOS battery? But I have seen a photo from the guy when it was running with no battery

Hello guys, very often people do not provide any update and those repair threads simply fade away. However I have an update and one remaining glitch with the board I can ´t figure out.
I have managed to get the board running after a number of failures and problems. Long story short:
- main issue were some loose legs all around the chipset - after several attempts I made it reliable again
- problematic sockets for 27C256 BIOS chips, dirt, corrosion, age...chips cleaned, sockets replaced for a pair of new ones in a good quality - random POST problems disappeared
- totally worn out memory sockets for 4256 chips - age, probably molested by the previous owner, than the recent owner and me as well we finalized the damage by pulling the chips often out. Replaced for 8 new better quality ones - good
random "no memory or bad memory" issue disappeared
All the parts came through out meticulous cleaning, multiple application of deoxidizer etc.
NEXT ISSUE (emotional Roller coaster) - it works now 100% reliable when in de-turbo regime = 10Mhz. And it works reliably if you press the turbo AFTER the machine is already running.
BUT:
There were issues - basically no POST "-- --" on a POST analyzer when I tried to start the machine with my TURBO on = 20Mhz. I found another issue - a broken trace from a ceramics 270pF capacitor
which is in-between the GND and the power-good signal. (Shaping the signal somehow?) (Surprisingly no issue in 10Mhz regime). I have fixed the trace and the machine is now (in 99% I think) able to start even with turbo on in 20Mhz regime.
BUT AGAIN:
- No issues at 10Mhz
- but if you ran the machine at 20Mhz (Turbo on) and you press the RESET button sometimes it works, sometimes (often) it crashes. The image on my monitor goes black, but I see the same numbers still on my POST analyzer, or I see some random numbers and the POST process is stuck. Very often just pressing the reset 2- times, or holding it for 3-2 seconds helps. But sometimes a power cycle is needed.
(a new update now is that with one older more worn-out PSU I can ´t POST in 20Mhz until I press the RESET, but with my testing NewOldStock PSU its OK - but still the reset button issue in turbo regime is persistent)
--
I have tried to trace the reset button circuitry - seems to be intact. It simply connects the PWR-GOOD signal to GND for a second through out a reversed diode and a resistor. My understanding is the button takes the power-good away and thus force the CPU to reset. The power good is obviously present - otherwise there is no chance for the machine to boot. All the resistors there have good readings. I have even replaced the 270pF capacitor for a new one. And even tried a bit stronger one. But it seems to have no influence in a good / bad way.
Those issues strictly occurs only when running turbo (20mhz). And even with no memory chips installed the reset behavior is still there. (it obviously beeps the memory is missing ). The issue looks like not connected to any IS card or memory. It is strictly isolated on PWR_GOOD (Reset) circuitry. My assumption is there is some problem with accepting the pwr good signal, or resetting the cpu once it is running on 20Mhz. Why the hell if I press the button would ´t the CPU simply RESET?
(Any suggestion will be appreciated - too much effort and time had been invested in to that board. It is usable if your PSU is good and you know how to tackle the issue - but would be nice to have it 100% perfect)
Thank you guys!

H.W.Necromancer wrote on 2023-08-09, 12:15:

Why the hell if I press the button would ´t the CPU simply RESET?

Because the CPU RESET signal is generated by the chipset (on older designs - by the keyboard controller) and has a few sources:
1) POWER GOOD signal
2) manual reset button
3) software reset

It's because of (3) that the chipset does it, and the reason why is that 286 can enter protected mode but not exit it. It needs to be reset to return to real mode. Some software might require a temporary switch to PM and back to RM - XMS managers for example, although some HIMEM versions would just use undocumented LOADALL instruction to get around that. Windows 3 also does switch between both modes. So the mobo must be able to reset the CPU but preserve everything else to make this operation as transparent as possible.

Long story short, it migh just be that this chipset can't handle reset well at high clock speed. Perhaps whatever it does internally takes too long and with 20MHz clock ends up being incomplete, thus hanging the system. If everything else works well at 20MHz then I'd say it's the chipset.
The CPU itself also need the reset pulse to be applied with some minimal duration, it's possible the chipset is timing this for 10MHz and at 20 the timings are violated and CPU is not being reset correctly. That is still a chipset issue though.

I see. And thank you for this more complex explanation. I like 286 systems and trying to gather as much information as possible.
I can do one test - as my OSCilator is already socketed I can try to use more conveniont 16Mhz clock and try the same set of tests.
However if you dig through my long post you will see that the RESET button is basically grounding the PWR-GOOD. Like not directlly, there is a diode and (probably current limiting resistor?) and a (filter) capacitor) but the trace go like that. The endpoint is the spot where the PWR-GOOD is entering the system. It looks like pressing the button basically make the power good signal low/flat for a second and tus forcing the chipse to issue a reset. But it seems like its not enough.
But when you powercycle and the pwr good comes directly from the PSU which is happening in a fraction of a second the ystem works. Thats kinda weird.
(and the fact that with one PSU it starts always but with another PSU it behaves even worse brings me to a suspition there is something wrong with the propagation of the PWR good in the system).
Unfortunatelly do not have a osciloscope yet.
Just a hypothesis - the part of the baord where this circuitry is had been hit by the battery leakege. It has continuity and all the broken traces are fixed - including PWR good line. But isn ´t it possible the trace is somhow week or distrubed
in the way my multimeter reads continuity but the real square signal is not passing in the right way? (sorry for maybe a stupid question) but there are not many parts there...Or mybe one of the 74xxx circuits is marginal. Two of them were hit by the battery as well and are now socketed after the repair...

The problem is it is quite a bit rarer board and I have no running board to compare. And there is obviosly no datasheet for the chipset...this is basically a one chip solution. But seeing the board running at 10Mhz differently makes me think there are no really broken traces or fryied ICs..but something might be marginal...

Basically the button is connected like that and interacts with the PWR-GOOD. The signal goes from the resistors to the chipset and other Ics - need to investigate more - but its def. reaching those otherwise no way the machine can boot! (and sorry for my a bit stupid drawing)

H.W.Necromancer wrote on 2023-08-09, 14:04:

Basically the button is connected like that

Check again because I'm pretty sure that diode should be connected backwards. You might want to try and replace this diode, assuming it's a standard silicon like 1N4148, with something Schottky. Reason being it'll have lower voltage drop and drive the line closer to GND.

Deunan wrote on 2023-08-09, 14:44:

H.W.Necromancer wrote on 2023-08-09, 14:04:

Basically the button is connected like that

Check again because I'm pretty sure that diode should be connected backwards. You might want to try and replace this diode, assuming it's a standard silicon like 1N4148, with something Schottky. Reason being it'll have lower voltage drop and drive the line closer to GND.

Thanks! Maybe I have just messed up the diagram. Just to mention that diode has not been corroded hence no repairs in that part.

BTW How much critical is the value of that capacitor? The original seems to be OK and has exactly 270pF. Now there is a replacement - it was a blind shot a bit. The replacement reads 300 (330 but it has some tolerance). It seems to be no change..

Deunan wrote on 2023-08-09, 14:44:

H.W.Necromancer wrote on 2023-08-09, 14:04:

Basically the button is connected like that

Check again because I'm pretty sure that diode should be connected backwards. You might want to try and replace this diode, assuming it's a standard silicon like 1N4148, with something Schottky. Reason being it'll have lower voltage drop and drive the line closer to GND.

This is the real situation.

Ok I think the theory it in this
https://en.m.wikipedia.org/wiki/Power-on_reset
It looks like either the timing or the shape of the signal is somehow screwed up. It would explain why the behavior is slightly different with the two different PSUs..(but both known good)

That cap is there only to somewhat debounce the switch, not critical - and while it will also affect POWER GOOD signal, the time constant of 270 pF and 470 Ohm is about 127 ns if my napkin math is corrrect. That should not matter for POWER GOOD if the chipset was designed correctly and it has Schmidt trigger type input for it.

If your other PSU doesn't work properly it is due to it's own RC timer for it being wrong. It might also be related to excessive ripple noise on the 5V line, but for most PSUs the timing circuit is pretty dumb - it'll just follow 5V line, noise or not.

Deunan wrote on 2023-08-09, 18:10:

That cap is there only to somewhat debounce the switch, not critical - and while it will also affect POWER GOOD signal, the time constant of 270 pF and 470 Ohm is about 127 ns if my napkin math is corrrect. That should not matter for POWER GOOD if the chipset was designed correctly and it has Schmidt trigger type input for it.

If your other PSU doesn't work properly it is due to it's own RC timer for it being wrong. It might also be related to excessive ripple noise on the 5V line, but for most PSUs the timing circuit is pretty dumb - it'll just follow 5V line, noise or not.

Thank you. I am sniffing around the circuit. But so far nothing obvious. Just to note when the 270pF capacitor had been missing - corroded trace - it was not possible to start the board at 20Mhz at all (in most of the cases) it was dead "----´ Nothing. So it seems for this particular board it is somewhat "critical". (?)

Usually a cap too big will slow the voltage raise too much and that causes problems. In your case missing cap (~0pF) was a problem.
Well, nothing is stopping you from testing a bigger cap the. 510pF, or just add another 270 or so at the back of the existing one for testing purposes and see what that does.

Deunan wrote on 2023-08-09, 19:55:

Usually a cap too big will slow the voltage raise too much and that causes problems. In your case missing cap (~0pF) was a problem.
Well, nothing is stopping you from testing a bigger cap the. 510pF, or just add another 270 or so at the back of the existing one for testing purposes and see what that does.

Good plan. I have a handful of some capacitors for that purpose. I can try stick one to the bottom. I think the worst scenario is it just will not work.
The board went through some hell. I know zero about its history. (except my friend got it form another guy who bought it at a flea market in Bulgaria. It came with a 10Mhz CPU, some random memory and an 40Mhz OSC (100% added later and not factory). BatteryCorroded, bad shape. Anything is possible 😂)
--
Thank you very much for your help here. One of my assumtions is that maybe my fix of the powergood trace shows continuity on my dmm but may be too bad for the signal (?) . Is that possible? The resistance on the traces is ok.,1 to to 2 ohms and the continuity is there.
-- The pwr-good signal goes through some resistors ( those read good) and goes right to the chipset and to the 74C04N. That chip had not been hit by the battery stuff and all the parts there seems to be Ok.
--
Is there any way how I can issue the reset by software right from DOS? It would be interesting to compare the behavior.

You an ask KBC to reset the system, there might be an int 15h function for that as well (rather then doing direct I/O). I seem to recall DOS 3 had a program for easy rebooting the system from command line, but I might be confusing it with some non-PC machine...

A thing that was happening on a later board I was just playing with. Is that until issues with bad connections at memory slots and possibly also CPU were dealt with and it was running properly, the reset by switch or shorting motherboard pins would not work. I don't know why that is exactly, maybe the actual reset pin of the CPU was floating, but it could mean that you still have glitches elsewhere that are causing reset issues.

EDIT : hallo guys, just to conclude - the board has been fixed.
- I have replaced all the memory sockets
- I have found some bad chips in the original memory set
- I have fixed a bad trace leading to a ceramics capacitors (later on replaced the capacitor as well) which is responsible for "de-bouncing" the signal from the reset button
After all the issues ,corrisoin etc. had been fixed the board had been running just fine with some small issues - problems when "bios cachable" and "video bios cachable" options were set to ON.
And there were some issues if reseted with the buttonand with turob on - sometimes it was necessary to press the reset 3times or so for the board to start at 20Mhz.
Becouse I have installed a socket for the oscilator at the beginning I have noticed there is a silk screen "16mhz" printed on the board. I have found some 33.33Mhz OSC, no 32 available.
After I put this OSC there the board became fully stable under all circumstances with no above mentoned issues. Becouse the 20Mhz OSC was not original . obviosly hand soldered - my assumption is not all the chipsets and boards
produced in this line were suitable for 20Mhz. (maybe they were binning the chipsets same as CPUs)
Anyway from a dead rust crusty board I have got to a fully functional nice and fast 16.something (-: 286. Which is cool. The board is now very reliable.
--
In case you are fighting with the same board it can help you.