snufkin wrote on 2022-01-28, 19:55:

Ok, so if we use the Award ISA/EISA BIOS Version 4.0 entry from bioscentral, then with no memory installed it gives C1 BE (thank […]
Show full quote

Ok, so if we use the Award ISA/EISA BIOS Version 4.0 entry from bioscentral, then with no memory installed it gives C1 BE (thanks PC@LIVE), so it's completed chipset initialisation and fails on memory size check. Which makes sense. Then with memory installed it fails on 0B 0A, so it's completed initialise keyboard and fails on "CMOS Checksum ; video interface initialised" (video memory and video adapter still to come). 0B often seems to be something CMOS related. Q1&Q2 might be something to do with voltage supply for the CMOS?

If you haven't already got it figured out, any chance of photos showing the top and back of the area around those transistors, and of the flat face of the transistor?

[edit: circuit looks a bit like this (from the 82C206 datasheet:
82C206_BatteryCircuit.jpg
]

Yes this is very much in line with the circuit that I am facing.

Q1 is in fact not shorted but directly related to the CMOS reset jumper JP11.

Edit: or maybe there is a short still, continuity mode vs diode mode is tricking me here. I just have to desolder both Q1 and C42 to be sure.

Q1 is what junctions current from the battery to CMOS RAM.

However, regardless of its jumper position, C42 still reads a short so I will desolder that bad boy first.

Boy it takes time for the CMOS voltage to go down when it is floating. And I think due to this short, there is no means to short this to ground via JP11, and thus never resets properly?

Getting closer 😀

snufkin wrote on 2022-01-28, 19:55:

Ok, so if we use the Award ISA/EISA BIOS Version 4.0 entry from bioscentral, then with no memory installed it gives C1 BE (thank […]
Show full quote

Ok, so if we use the Award ISA/EISA BIOS Version 4.0 entry from bioscentral, then with no memory installed it gives C1 BE (thanks PC@LIVE), so it's completed chipset initialisation and fails on memory size check. Which makes sense. Then with memory installed it fails on 0B 0A, so it's completed initialise keyboard and fails on "CMOS Checksum ; video interface initialised" (video memory and video adapter still to come). 0B often seems to be something CMOS related. Q1&Q2 might be something to do with voltage supply for the CMOS?

If you haven't already got it figured out, any chance of photos showing the top and back of the area around those transistors, and of the flat face of the transistor?

[edit: circuit looks a bit like this (from the 82C206 datasheet:
82C206_BatteryCircuit.jpg

Q1 looks like the PNP in that circuit. So it it's gone short C-E then the battery will be trying to power the board +5V when the board is off.
]

Q1 is labeled KN3906 9403
Q2 is labeled KN3904 9416

Q1 reads C-B voltage drop 0.688V and .OL in the other direction. However E-B voltage drop as -1.5V (.OL in the other direction) ? Battery voltage at play here I guess.

Q2 reads B-E and B-C voltage drop of 0.698V

I'm wondering if JP24 is the CMOS clear? One side (pin 3?) comes from the battery (via CR2), not sure where the other side (pin 1?) goes but I'd guess Ground. Then I guess pin 2 goes off (via Q1) to CMOS stuff stored in the 82C206.

Ah, you said C42 is measuring short? The collector of Q1 attaches to a pad of that 1k resistor R64. The other side R64 connects to C42, and heads off somewhere on the board alongside the traces from the 32kHz crystal. Based on the schematic from earlier, I guess that trace is the PWRSTB# signal to the 206, and C42 is the 0.05uF ceramic capacitor to ground. If that's gone short, then PWRSTB# is held low. The datasheet says:

Power Supply Strobe: An active low input used to establish the condition of the control registers of the real-time clock when po […]
Show full quote

Power Supply Strobe: An active low input used to establish the condition of the control registers of the real-time clock when power is applied to the device. In a PC/AT-based system, it should be tied to the battery back-up circuit. When PSRSTB# and TEST are both low, the 82C206 is not accessible and the following bits in the RTC are cleared to zero:
• PIE (Periodic Interrupt Enable)
• AIE (Alarm Interrupt Enable)
• UIE (Update ended Interrupt Enable)
• UF (Update ended Interrupt Flag)
• IRQF (Interrupt Request status Flag)
• PF (Periodic Interrupt Flag)
• AF (Alarm interrupt Flag)
• Square Wave output enable.

Which doesn't sound like something that wants to be held low if the board is going to work.

Of course, the short might not be the capacitor. Might be a fault in the 206. All the fun of finding shorts.

Just to add a bit more, mostly to help me understand it by working through how I think it works... I think one of CR2 and CR3 is a blocking diode to prevent charging of an external battery and the other is part of the charge regulator for the on board battery (can't make out the traces well enough to be sure). External battery voltage (assumed to 6V I think) arrives at the board edge side of CR5, goes through CR5 and CR4,where it joins the on board battery (I think I can see a trace from JP24->C45->CR4). Battery voltage then touches Collector of Q1, C48 and R64. The fat trace from C48 will likely head off to the 82C206.

Separately from the that... R57 and R58 form a potential divider between PSU +5V and GND to the Base of Q2. When the board is off then no current flows and Q2 is off. If Q2 is off then no current can flow out of the Base of Q1, so that's off as well. No current can flow from the Collector-Emitter, so the battery is isolated from the rest of the board and only has to run the 32kHz oscillator and the 82C206. When the PSU is on then current can flow in to the base of Q2 and out of the Emitter to Ground. That allows current to flow from the Collector to the Emitter. That current comes through R56, which comes in through the Emitter of Q1 (now at +5V from the PSU) and out of the Base of Q1. That allow current to flow from the Emitter to the Collector (where the battery voltage is...), so the 206 can now be powered by the PSU and the battery can be charged up (with the external battery protected by CR5 and CR4).

Fingers crossed it is C42 at fault. The resistance to ground of the thin trace that connects to C42 and R64 needs to be not short.

snufkin wrote on 2022-01-29, 00:39:

I'm wondering if JP24 is the CMOS clear? One side (pin 3?) comes from the battery (via CR2), not sure where the other side (pin […]
Show full quote

I'm wondering if JP24 is the CMOS clear? One side (pin 3?) comes from the battery (via CR2), not sure where the other side (pin 1?) goes but I'd guess Ground. Then I guess pin 2 goes off (via Q1) to CMOS stuff stored in the 82C206.

Ah, you said C42 is measuring short? The collector of Q1 attaches to a pad of that 1k resistor R64. The other side R64 connects to C42, and heads off somewhere on the board alongside the traces from the 32kHz crystal. Based on the schematic from earlier, I guess that trace is the PWRSTB# signal to the 206, and C42 is the 0.05uF ceramic capacitor to ground. If that's gone short, then PWRSTB# is held low. The datasheet says:

Power Supply Strobe: An active low input used to establish the condition of the control registers of the real-time clock when po […]
Show full quote

Power Supply Strobe: An active low input used to establish the condition of the control registers of the real-time clock when power is applied to the device. In a PC/AT-based system, it should be tied to the battery back-up circuit. When PSRSTB# and TEST are both low, the 82C206 is not accessible and the following bits in the RTC are cleared to zero:
• PIE (Periodic Interrupt Enable)
• AIE (Alarm Interrupt Enable)
• UIE (Update ended Interrupt Enable)
• UF (Update ended Interrupt Flag)
• IRQF (Interrupt Request status Flag)
• PF (Periodic Interrupt Flag)
• AF (Alarm interrupt Flag)
• Square Wave output enable.

Which doesn't sound like something that wants to be held low if the board is going to work.

Of course, the short might not be the capacitor. Might be a fault in the 206. All the fun of finding shorts.

Just to add a bit more, mostly to help me understand it by working through how I think it works... I think one of CR2 and CR3 is a blocking diode to prevent charging of an external battery and the other is part of the charge regulator for the on board battery (can't make out the traces well enough to be sure). External battery voltage (assumed to 6V I think) arrives at the board edge side of CR5, goes through CR5 and CR4,where it joins the on board battery (I think I can see a trace from JP24->C45->CR4). Battery voltage then touches Collector of Q1, C48 and R64. The fat trace from C48 will likely head off to the 82C206.

Separately from the that... R57 and R58 form a potential divider between PSU +5V and GND to the Base of Q2. When the board is off then no current flows and Q2 is off. If Q2 is off then no current can flow out of the Base of Q1, so that's off as well. No current can flow from the Collector-Emitter, so the battery is isolated from the rest of the board and only has to run the 32kHz oscillator and the 82C206. When the PSU is on then current can flow in to the base of Q2 and out of the Emitter to Ground. That allows current to flow from the Collector to the Emitter. That current comes through R56, which comes in through the Emitter of Q1 (now at +5V from the PSU) and out of the Base of Q1. That allow current to flow from the Emitter to the Collector (where the battery voltage is...), so the 206 can now be powered by the PSU and the battery can be charged up (with the external battery protected by CR5 and CR4).

Fingers crossed it is C42 at fault. The resistance to ground of the thin trace that connects to C42 and R64 needs to be not short.

You are definitely right about JP24 being the CMOS reset jumper. Most impressive how you are able to derive all this from the crude pictures. Will provide a more thorough update later, but I can say as much as R64 goes to OPTi 82C206 chipset. So this is what maintains the CMOS RAM voltage.

Now with JP24 in CMOS reset position I get sane diode mode readings from Q1 (PNP one). But the Q2 now reads 0.7V B-C, open C-B, 0.7V B-E and 1.9V E-B (should be open) in circuit, so this is my main culprit right now. I will desolder it and measure it out of circuit.

C42 seems ok (no short). And the input to OPTi 82C206 is not shorted to ground.

You are completely right about the voltage divider and Vcc coming into R57.

The thick trace from C48 goes straight to the ISA port, does it need battery backup?

Edit: this is obviously Vcc to ISA bus 😀

I have no idea about JP11 next to the BIOS ROM, maybe something about oscillator source?

Dropping some notes I have gathered so far:

LocalBus wrote on 2022-01-29, 16:12:

You are definitely right about JP24 being the CMOS reset jumper. Most impressive how you are able to derive all this from the crude pictures.

It's mostly a case of taking the photos you've provided, using a tool to align them all (I just use GIMP), top and bottom, then seeing where the traces go. Each photo can be its own layer, and I can switch layers on and off. So if a trace on the top surface goes to a via then I can just turn off the top layer and pick up the via on the bottom surface. As far as I know motherboard designers try to keep inner layers for grounds and power planes, so most signal traces can be followed, unless they go under components.

For instance, I can see that the thick trace from C48 goes to a via next to the ISA slot, but on the bottom layer it heads under the slot and out the other side, going to that 4069 chip. I can look up (or sometimes remember) that that's a bunch of NOT gates/inverters. It's right by the 32kHz crystal, which is used for the CMOS clock. So it's a fair guess that it's involved in generating the clock signal, which the battery will have to keep running, by feeding the output of a NOT gate back in to its own input via the crystal. At which point I could start looking for traces to confirm that. The trace disappears from the bottom layer, so I'd expect to see a thick trace emerging from underneath the 4096 and head in the direction of the 206.

Will provide a more thorough update later, but I can say as much as R64 goes to OPTi 82C206 chipset. So this is what maintains the CMOS RAM voltage.

The side of R64 nearest Q1 should go to the Vcc pins of the 206 (pin 26 and 78). The side closest the board mounting hole should go to that PSRSTB# (pin 7). One thing worth doing is checking what the voltage is relative to ground on both sides of R64.

Now with JP24 in CMOS reset position I get sane diode mode readings from Q1 (PNP one). But the Q2 now reads 0.7V B-C, open C-B, 0.7V B-E and 1.9V E-B (should be open) in circuit, so this is my main culprit right now. I will desolder it and measure it out of circuit.

Don't forget that the Base of Q2 is between the voltage divider formed by R58 (to Ground?) and R57 (to +5V?). And that the Emitter of Q2 is probably connected directly to Ground. So the E-B junction is in parallel with R58 (10k). So your diode test will show a current flowing, but it's probably not through Q2.

C42 seems ok (no short). And the input to OPTi 82C206 is not shorted to ground.

What were you measuring when you found a short across C42 before?

I have no idea about JP11 next to the BIOS ROM, maybe something about oscillator source?

Dunno. I'd start with guessing something involving programming voltage, maybe a size select?

Since you're getting some POST codes I'm figuring that accessing the BIOS is ok, so not too worried about that.

Quick skim of the notes look ok. It can make it clearer if you start drawing out symbols. I find it easier to grasp what connects to what by being able to see the lines connecting devices, and what any voltages between two points are by drawing arrows alongside them. You can take that schematic I posted earlier of the 206 charge circuit and you should be able to number all the components in that schematic with the component numbers from this board. E.g. R57 is the 51k resistor going from the base of the NPN transistor to Vcc.

For the moment, with the board off (but battery in place) I'd check the voltages (relative to ground) either side of R64, then measure them again with the board on.

Something possibly odd. Just skimming the Opti 206 datasheet and I can't see that it supports a low voltage supply for battery standby. The working voltage is given as 4.75 - 5.25V. So if R64 nearest Q1 is below that then that could cause problems, and the on board battery is around 3.6V?

Thinking back to your POST card results, one possible interpretation of the 0B fault was "CMOS checksum; video interface initialised". Maybe the BIOS is corrupted, or there's a problem with the video card? Hopefully Jan will be able to help with the BIOS once you get your TL-866.

snufkin wrote on 2022-01-29, 23:54:

It's mostly a case of taking the photos you've provided, using a tool to align them all (I just use GIMP), top and bottom, then […]
Show full quote

LocalBus wrote on 2022-01-29, 16:12:

You are definitely right about JP24 being the CMOS reset jumper. Most impressive how you are able to derive all this from the crude pictures.

It's mostly a case of taking the photos you've provided, using a tool to align them all (I just use GIMP), top and bottom, then seeing where the traces go. Each photo can be its own layer, and I can switch layers on and off. So if a trace on the top surface goes to a via then I can just turn off the top layer and pick up the via on the bottom surface. As far as I know motherboard designers try to keep inner layers for grounds and power planes, so most signal traces can be followed, unless they go under components.

For instance, I can see that the thick trace from C48 goes to a via next to the ISA slot, but on the bottom layer it heads under the slot and out the other side, going to that 4069 chip. I can look up (or sometimes remember) that that's a bunch of NOT gates/inverters. It's right by the 32kHz crystal, which is used for the CMOS clock. So it's a fair guess that it's involved in generating the clock signal, which the battery will have to keep running, by feeding the output of a NOT gate back in to its own input via the crystal. At which point I could start looking for traces to confirm that. The trace disappears from the bottom layer, so I'd expect to see a thick trace emerging from underneath the 4096 and head in the direction of the 206.

Will provide a more thorough update later, but I can say as much as R64 goes to OPTi 82C206 chipset. So this is what maintains the CMOS RAM voltage.

The side of R64 nearest Q1 should go to the Vcc pins of the 206 (pin 26 and 78). The side closest the board mounting hole should go to that PSRSTB# (pin 7). One thing worth doing is checking what the voltage is relative to ground on both sides of R64.

Now with JP24 in CMOS reset position I get sane diode mode readings from Q1 (PNP one). But the Q2 now reads 0.7V B-C, open C-B, 0.7V B-E and 1.9V E-B (should be open) in circuit, so this is my main culprit right now. I will desolder it and measure it out of circuit.

Don't forget that the Base of Q2 is between the voltage divider formed by R58 (to Ground?) and R57 (to +5V?). And that the Emitter of Q2 is probably connected directly to Ground. So the E-B junction is in parallel with R58 (10k). So your diode test will show a current flowing, but it's probably not through Q2.

C42 seems ok (no short). And the input to OPTi 82C206 is not shorted to ground.

What were you measuring when you found a short across C42 before?

I have no idea about JP11 next to the BIOS ROM, maybe something about oscillator source?

Dunno. I'd start with guessing something involving programming voltage, maybe a size select?

Since you're getting some POST codes I'm figuring that accessing the BIOS is ok, so not too worried about that.

Quick skim of the notes look ok. It can make it clearer if you start drawing out symbols. I find it easier to grasp what connects to what by being able to see the lines connecting devices, and what any voltages between two points are by drawing arrows alongside them. You can take that schematic I posted earlier of the 206 charge circuit and you should be able to number all the components in that schematic with the component numbers from this board. E.g. R57 is the 51k resistor going from the base of the NPN transistor to Vcc.

For the moment, with the board off (but battery in place) I'd check the voltages (relative to ground) either side of R64, then measure them again with the board on.

Something possibly odd. Just skimming the Opti 206 datasheet and I can't see that it supports a low voltage supply for battery standby. The working voltage is given as 4.75 - 5.25V. So if R64 nearest Q1 is below that then that could cause problems, and the on board battery is around 3.6V?

Thinking back to your POST card results, one possible interpretation of the 0B fault was "CMOS checksum; video interface initialised". Maybe the BIOS is corrupted, or there's a problem with the video card? Hopefully Jan will be able to help with the BIOS once you get your TL-866.

Yes never underestimate the power of image manipulation 😀

Ok C42 is beeping in continuity when the CMOS battery is in circuit (I.e. JP24 Pin-2 + 3 position) - but only in one direction, so I guess from a continuity point of view it essentially see the battery in parallel with C42 (via CR2). Does this make sense?

With JP24 in CMOS reset jumper position, it does not show any continuity over C42.

R64 shows 3.605V relative to GND on either sides when battery is in circuit, otherwise 0V.

JP11 (Pin-2) goes to BIOS ROM pin-1 which have something to do with A14 ?

https://ibm.retropc.se/ic/MK38000.png

Maybe old and misleading? Other references just point out Vpp for Pin-1. Only relevant when programming?

JP11 (Pin-3) goes straight to +12V rail on the mainboard AT connector.

JP11 (Pin-1) goes to battery voltage 3.6V

So depending on position you set different levels on Pin-1 on the BIOS ROM. Question is which position that is correct? It was set to 12V when I got it.

Regarding Q2 E-B reading in continuity mode, it sees the voltage divider resistance of 8.42 kOhm so I guess the 1.9V diode mode readings in this direction can be attributed to that.

Before I fire it up again I would really like know what is the purpose of JP11 depending on position (A14 pin on BIOS ROM), 12V vs 3.6V (or rather 3.3V after the diode).

Yes the CMOS battery is a three cell NiMH (originally NiCd) which gives 3.6V. I would guess this is pretty standard. Also known as Varta/3 battery.

LocalBus wrote on 2022-01-30, 11:10:

Ok C42 is beeping in continuity when the CMOS battery is in circuit (I.e. JP24 Pin-2 + 3 position) - but only in one direction, so I guess from a continuity point of view it essentially see the battery in parallel with C42 (via CR2). Does this make sense?

With JP24 in CMOS reset jumper position, it does not show any continuity over C42.

Sounds plausible. With JP24 1-2 will connect Ground of pin 1 (?) to pin 2, which connects to the Q1 side of R64. Other side of R64 connects to C42. The other side of C42 connects to Ground. So measuring across C42, current has to go through the 1k R64, so they're connected but not shorted. With the battery in place things are more complicated. The meter will try to drive a current between the two probes and measure what voltage is produced. If there's no voltage then there's no resistance (a short), if there's a voltage then there's at least some resistance between the two points. So if the battery is holding one side of C42 at 3.6V, and the other side is at 0V, then the meter will give the wrong answer for resistance. One way around, it'll try to drive a small current, measure 3.6V and assume that's a resistance and not connected. The other way around it'll try to drive a small current, measure the voltage as -3.6V, so a -ve resistance, and show them as connected.

I find it causes fewer problems and confusion if any voltage sources are removed before attempting to measure any resistance/continuity/diode testing.

R64 shows 3.605V relative to GND on either sides when battery is in circuit, otherwise 0V.

So that's below what the datasheet allows for the 206. Which may not be a problem since plenty of boards use the same set up. But it's not great. Putting a 6V battery pack on the external battery connector might be worth trying. You don't have to remove the board battery to do that (although removing it is probably a good thing to do anyway).

JP11 (Pin-2) goes to BIOS ROM pin-1 which have something to do with A14 ?
JP11 (Pin-3) goes straight to +12V rail on the mainboard AT connector.
JP11 (Pin-1) goes to battery voltage 3.6V

Would you mind carefully lifting the label on the ROM and finding out what the part is? There are some different ROM pinouts that might make a difference. If the part is something like a 27FS010 then pin 1 is a Vpp pin for doing chip erase and program, and the jumper is to enable or disable updating the BIOS. In that case, that jumper wants to be not 12V. It's probably not a problem, the chip isn't erased just by applying 12V to that pin, other stuff has to happen as well. And if the chip had been erased then you wouldn't get anything from the POST card. But it'll be safer with it not at 12V.

datasheet for a 27FS010: https://www.mouser.co.uk/datasheet/2/176/71152-1102455.pdf

Odd thing for me is that you get the battery voltage there. I can't think why it would need to be powered when the board is off. If you haven't already (I might have missed this), could you remove Q1 and if your meter has a transistor checker then give it a test. That transistor is supposed to isolate the battery power bits of the board (RTC, CMOS) from the rest of the board.

Some updates. I desoldered Q2 and measured it in diode mode (my DMM does not have the transistor measurement feature) and it looks good for both B-E and B-C. Nothing in reverse.

I also desoldered the battery just to rule it out of the equation and now C42 does not read continuity in any direction with my DMM.

R64 reads 5V when the board is powered.

Pin-1 on the BIOS ROM receives 5V as well with JP11 in 1-2 position (2-3 position is 12V). It reads 0V when powered off so no battery voltage present here; measured before I desoldered obviously 😉

I fired it up and no difference. Even tested with an ISA VGA card (Trident) but I guess it doesn't reach that far in the POST sequence.

Still 0B 0A.

In an attempt to rule out any shorts I picked up my recently acquired FLIR One camera attachment to my smartphone, and I found something interesting. The OPTi 82C206 does not read any heat signature what so ever. So I think there is a power issue to this chip or possibly something with clock signal so it doesn't really enable?

If it was shorted it would have been the opposite, so that is at least good news 😅

I think I found a datasheet for the 82C206 so I need to go over all the inputs.

Looking at the datasheet there is something about "power good signal" on Pin-6 PWRGD, does this come straight from the PSU in the case of an old school AT PSU? Who knows in what shape it is in 🤔 I quickly measured its outputs with just a simple fan connected and +5V was good, but 12V was more like 11.2V.

Edit: just found the orange lead on the mainboard AT connector. It gives +5V from the PSU. I have now traced this all the way to the 82c206 via a SN7407N buffer/driver.

Time to order one of those ATX-to-AT adapters!

Edit: it does seem to receive Vcc voltage as supposed to via pin 26 and 78 respectively.

I think I have also managed to trace the 32kHz clock signal from the Xtal via TC4069 and finally to pin 75.

SYSCLK seems to go into pin 15 but I have not yet traced it all the back to the Chrontel clock generator.

Next up is to just go over all the pads/joints with some fresh solder. With 100 pins there might very well be one cold joint after all these years.

Edit: just found a quite comprehensive list of POST codes here -> http://mrbios.com/techsupport/award/postcodes.htm

Late Award BIOS (4.5x-non PnP)

0B :
Test CMOS RAM Checksum if bad or Insert key depressed; load defaults.

Makes me wonder if it is as easy as holding down a key to force it loading the defaults since the ones in CMOS RAM are obviously no good.

Hang in there!

Might be a daft question, but did you put Q2 back in? And have you tried removing and checking Q1? The battery voltage should be permanently connected to the 206, and the board +5V connects to the 206 through Q1. Q1 only turns on when board +5V is present, and it turns off when +5V turns off. That means it should stop the battery trying to power everything connected to +5. Since you saw 3.6V on J11, and I wasn't expecting to see that powered by the battery, I'm wondering if perhaps Q1 is not able to isolate between the battery and board supplies.

Things I'd check with the board off and no battery fitted:
Resistance from R64 side nearest Q1 to pin 26 & 78 on the 206. They should be short (Vcc)
Resistance from R64 side nearest mounting hole to pin 7 on the 206. That should also be short.
Resistance from 206 pin 3,52,77 to Ground (use one of the mounting hole surrounds). Should also be short.
Resistance from one of the +5 AT pins to R64 side by Q1. This should be open circuit (blocked by Q1 when the board is off).

I can't see where the PWRGD trace goes. It looks like it's on the back of the board and goes under the ISA slot 4th from board edge, but then the photos of back of board run out. It may be buffered so it can drive multiple things. If pin 6 of the 206 doesn't go high then it will stay in low power mode, drawing microamps, so wouldn't heat up noticeably compared to the other components.

Nice thermal images. Look at the regulator by the BIOS burn, and the clock buffer.

Very beautiful those thermal images, I think they can be useful but only if you can highlight strange temperatures, that is higher than normal.
Now I do not know exactly what it is to not start the PC, sure that strangely cold Chip could make us suspect a failure there.
It would be useful to check if the CPU and RAM voltages are present, in the SKT 4 there should be + 5V, in the SKT 5 instead about + 3.3V, for the RAM I imagine you will find + 5V, if some are missing it is normal that it will not start.
Obviously I apologize if you have already done these checks, but it can help you locate the possible fault if any voltage is missing.
The use of the ISA card said that nothing changes, but it would be necessary to check the PINs if they are powered, it is not that the POST CARD LED highlights if there is no power, it only shows that in that PIN there is, but it may not be there in the others.
Finally I wonder if the CPU warms up, or not (?), If it doesn't warm up it is obvious that the PC will not start, and at that point there are two possible things or the CPU is not working or there is some power problem.

PC@LIVE wrote on 2022-01-30, 21:09:

Very beautiful those thermal images, I think they can be useful but only if you can highlight strange temperatures, that is high […]
Show full quote

Very beautiful those thermal images, I think they can be useful but only if you can highlight strange temperatures, that is higher than normal.
Now I do not know exactly what it is to not start the PC, sure that strangely cold Chip could make us suspect a failure there.
It would be useful to check if the CPU and RAM voltages are present, in the SKT 4 there should be + 5V, in the SKT 5 instead about + 3.3V, for the RAM I imagine you will find + 5V, if some are missing it is normal that it will not start.
Obviously I apologize if you have already done these checks, but it can help you locate the possible fault if any voltage is missing.
The use of the ISA card said that nothing changes, but it would be necessary to check the PINs if they are powered, it is not that the POST CARD LED highlights if there is no power, it only shows that in that PIN there is, but it may not be there in the others.
Finally I wonder if the CPU warms up, or not (?), If it doesn't warm up it is obvious that the PC will not start, and at that point there are two possible things or the CPU is not working or there is some power problem.

Thanks for the pointers. The CPU heats up allright and yes, I have confirmed that I have 3.3V from the voltage regulator. Sure it is worth going over all the expected voltages, but I'm pretty confident that the CPU and RAM are just fine, or otherwise would have stopped at earlier POST CODE. Would I even get to 01 without a working CPU?

snufkin wrote on 2022-01-30, 20:27:

Might be a daft question, but did you put Q2 back in? And have you tried removing and checking Q1? The battery voltage should b […]
Show full quote

Might be a daft question, but did you put Q2 back in? And have you tried removing and checking Q1? The battery voltage should be permanently connected to the 206, and the board +5V connects to the 206 through Q1. Q1 only turns on when board +5V is present, and it turns off when +5V turns off. That means it should stop the battery trying to power everything connected to +5. Since you saw 3.6V on J11, and I wasn't expecting to see that powered by the battery, I'm wondering if perhaps Q1 is not able to isolate between the battery and board supplies.

Things I'd check with the board off and no battery fitted:
Resistance from R64 side nearest Q1 to pin 26 & 78 on the 206. They should be short (Vcc)
Resistance from R64 side nearest mounting hole to pin 7 on the 206. That should also be short.
Resistance from 206 pin 3,52,77 to Ground (use one of the mounting hole surrounds). Should also be short.
Resistance from one of the +5 AT pins to R64 side by Q1. This should be open circuit (blocked by Q1 when the board is off).

I can't see where the PWRGD trace goes. It looks like it's on the back of the board and goes under the ISA slot 4th from board edge, but then the photos of back of board run out. It may be buffered so it can drive multiple things. If pin 6 of the 206 doesn't go high then it will stay in low power mode, drawing microamps, so wouldn't heat up noticeably compared to the other components.

Nice thermal images. Look at the regulator by the BIOS burn, and the clock buffer.

Relevant question I would put it, but yes, I put Q2 back in circuit. I am yet to remove Q1 and do the same measurements.

I will go over the points you mentioned, I sure hope it is as simple as something power related.

What does the "load defaults" really mean? I would guess fetching those from the BIOS ROM, and if this fails, well then it goes to a halt. Is this fetching taking place directly by 82C206? Assuming some address/data lines connected to the ROM.

Haha yes the voltage regulator sure have a tough life, surprised that it still works 😅

LocalBus wrote on 2022-01-30, 21:46:

PC@LIVE wrote on 2022-01-30, 21:09:

Very beautiful those thermal images, I think they can be useful but only if you can highlight strange temperatures, that is high […]
Show full quote

Very beautiful those thermal images, I think they can be useful but only if you can highlight strange temperatures, that is higher than normal.
Now I do not know exactly what it is to not start the PC, sure that strangely cold Chip could make us suspect a failure there.
It would be useful to check if the CPU and RAM voltages are present, in the SKT 4 there should be + 5V, in the SKT 5 instead about + 3.3V, for the RAM I imagine you will find + 5V, if some are missing it is normal that it will not start.
Obviously I apologize if you have already done these checks, but it can help you locate the possible fault if any voltage is missing.
The use of the ISA card said that nothing changes, but it would be necessary to check the PINs if they are powered, it is not that the POST CARD LED highlights if there is no power, it only shows that in that PIN there is, but it may not be there in the others.
Finally I wonder if the CPU warms up, or not (?), If it doesn't warm up it is obvious that the PC will not start, and at that point there are two possible things or the CPU is not working or there is some power problem.

Thanks for the pointers. The CPU heats up allright and yes, I have confirmed that I have 3.3V from the voltage regulator. Sure it is worth going over all the expected voltages, but I'm pretty confident that the CPU and RAM are just fine, or otherwise would have stopped at earlier POST CODE. Would I even get to 01 without a working CPU?

Well I don't know for sure, but according to what I know, usually the no boot problem is caused by a memory problem.
Then it depends if the CPU is completely broken, I don't think the POST CARD displays any code, if it is malfunctioning it could display some codes, sometimes it starts normally but afterwards it starts to give problems in loading the OS, here you do not get to see many codes (I guess).
Being a fairly old card, there could be problems with the capacitors, they could create strange voltages that prevent starting, it depends on the type present in the MB (electrolytic or tantalum), it can happen that even if visually perfect, in reality they are not, and therefore must be replaced.

LocalBus wrote on 2022-01-30, 21:55:

What does the "load defaults" really mean?

most likely a global flag making BIOS use default (embedded in the code) settings.

LocalBus wrote on 2022-01-30, 21:55:

I would guess fetching those from the BIOS ROM, and if this fails, well then it goes to a halt.

BIOS code is in BIOS ROM, executing BIOS means ROM access is working (unless ROM chip contents are corrupted). There is potential for one of address lines not working between CPU and ROM chip, but its very unlikely since BIOS starts execution at the very end and then jumps all over the place.
https://web.archive.org/web/20120905135825/ht … 32300699?pgno=3
The very first bytes fetched from BIOS ROM chip are the last 16 ones (0xFFFFF0 F000:FFF0) followed by Far jump to somewhere in the middle of BIOS binary.

LocalBus wrote on 2022-01-30, 21:55:

Is this fetching taking place directly by 82C206? Assuming some address/data lines connected to the ROM.

82C206 is not being used while accessing BIOS ROM. It contains a bunch of peripheral chips hanging off ISA bus (interrupts, timers, DMA, RTC, finally CMOS SRAM) + ISA bus control stuff. ISA cards will most likely not even work without working 82C206.
BIOS usually connects directly to main chipset (CPU bus over a buffer + byte order arbitration) in parallel with keyboard controller. You can get the general idea by looking at example open source system using 82C495XLC + 82C206 https://alexandrugroza.ro/microelectronics/sy … sbmc/index.html

"B Test CMOS RAM Checksum Test CMOS" combined with cold 206 suggests it not being powered up at all = BIOS is unable to read/write it. Its possible there is a verify routing here reading back just written bytes and this is where it hangs. Once you dump the bios it will be possible to decompile it and look exactly at the part that hangs based on the last POST code.

LocalBus wrote on 2022-01-30, 21:55:

Edit: just found the orange lead on the mainboard AT connector. It gives +5V from the PSU. I have now traced this all the way to the 82c206 via a SN7407N buffer/driver.
Edit: it does seem to receive Vcc voltage as supposed to via pin 26 and 78 respectively.

good, was about to suggest that.
I find it weird that pin 6 PWRGD is high - this is a perfect candidate for a fault resulting in 82c206 staying cold in standby disconnected from the bus
I see you already found http://www.bitsavers.org/components/opti/data … eets/82C206.pdf
Now check if pin 7 PSRSTB is high when powered up
pin 10 RESET seems to only affect DMA and Interupt controllers, but worth checking anyway if it isnt stuck high when powered up

LocalBus wrote on 2022-01-30, 21:55:

Time to order one of those ATX-to-AT adapters!

why?

snufkin wrote on 2022-01-29, 23:54:

Something possibly odd. Just skimming the Opti 206 datasheet and I can't see that it supports a low voltage supply for battery standby. The working voltage is given as 4.75 - 5.25V. So if R64 nearest Q1 is below that then that could cause problems, and the on board battery is around 3.6V?

Datasheet lists maximum/operating supply and standby current. Doesnt list standby voltage, probably because everything was already standardized on ~3V cells.

>The PWRGD pin is provided on the device to protect the contents of RAM and the real-time clock. It is also used to reduce
power consumption whenever the system is powered down.

rasz_pl wrote on 2022-01-31, 06:49:

LocalBus wrote on 2022-01-30, 21:55:

Time to order one of those ATX-to-AT adapters!

why?

Just to rule out any PSU issues, but yes this is far-fetched. More of something to use in the long run.

Thanks for all the comments and resources/links.

I understand now that the 82C206 is quite passive to begin with and it is getting evident that the problem is when BIOS tries to read/write the CMOS SRAM, which most likely is not powered up, at least not for external access.

In fact, when probing around for GND <-> Vcc voltage/potential on pins 26 and 78 something happened when I touched pin 78: the POST CODE display quickly stepped to FF xx (it startled me since I thought I shorted pin 77 (GND) and pin 78 Vcc) so I pushed reset switch to see that I was back at 0B 0A - thus I didn't recall the last two digits. Now looking at the POST CODE references, this indicates that it reached boot state (FF) ?

All in all, this further points at bad solder joints so I will go over them all tonight.

Tomorrow, hopefully the TL866 will arrive so I can dump the BIOS ROM.

Small update. Resoldered the suspect pins 75 - 80 on the 82C206 but no difference.

Then I started pressing down on various socketed ICs and found that the main clock buffer was not properly seated. Problem is that this just made things worse 🙃 Went downhill from 0B 0A to 00 06 to -- -- so I guess we have another culprit then. Bad clock signal to some parts of the board since the SN74ABT245N is duplicating them (both ISA clock and FSB clock goes through it, where the Chrontel is responsible for generating both).

Went on a crusade and lifted everything that is socketed that I had not touched before, such as cache chips.

Maybe it is as silly as oxidation and reseating of components.

Before I make matters worse I will dump the BIOS ROM...

For those who wonder, I did source a separate 66.666 MHz oscillator and plugged it in the vacant slot.

This allows me to either use the Chrontel clock generator or the external oscillator by changing position of JP13 (currently using Chrontel but both works):

Over and out!

There is 486 board ( zida 4dps clone) that has a similar model name - "PCI400-4" . https://www.petershipkov.com/temp/retro_pc_im … ms_pci400-4.jpg

May be it's same manufacturer, "Taken" ?

Chkcpu wrote on 2022-01-28, 12:55:

@LocalBus, If you have an (E)EPROM programmer, can you make a BIOS copy from the EPROM chip?
I would love to poke around in the Award BIOS from this unique board, and I should be able to tell you exactly what the BIOS is doing in POST 0A. 😉

Jan

Ok, I think I managed to succeed with the BIOS ROM dump (Intel D27C010J EEPROM) - 128 kB in size:

Let me know if it is valid, I did run verify a couple of times. Thanks!

snufkin wrote on 2022-01-30, 12:17:

Would you mind carefully lifting the label on the ROM and finding out what the part is?

Here goes, out of necessity I needed to lift it 😀

overdrive333 wrote on 2022-02-01, 15:52:

There is 486 board ( zida 4dps clone) that has a similar model name - "PCI400-4" . https://www.petershipkov.com/temp/retro_pc_im … ms_pci400-4.jpg

May be it's same manufacturer, "Taken" ?

Interesting, thanks! Certainly looks similar and shares the same naming convention. "Taken" 😂

Soon we have mapped out this board pretty well I think!

LocalBus wrote on 2022-02-01, 17:00:

Ok, I think I managed to succeed with the BIOS ROM dump (Intel D27C010J EEPROM) - 128 kB in size: […]
Show full quote

Chkcpu wrote on 2022-01-28, 12:55:

@LocalBus, If you have an (E)EPROM programmer, can you make a BIOS copy from the EPROM chip?
I would love to poke around in the Award BIOS from this unique board, and I should be able to tell you exactly what the BIOS is doing in POST 0A. 😉

Jan

Ok, I think I managed to succeed with the BIOS ROM dump (Intel D27C010J EEPROM) - 128 kB in size:

AWARD_ISA_PCI_586_non_PNP_SN_013870745_1994.zip

Let me know if it is valid, I did run verify a couple of times. Thanks!

Yes, the BIOS ROM dump is a valid 128KB Award v4.50G non-PnP / PCI BIOS. So no corruption here. 😀
BIOS-ID: 07/18/94-OPTI-571/822-2A5ULT40-00
Revision: (2a5ulT40) 256K/512K/1024K CACHE VERSION

Give me some time to analyze the POST routines to see what's going on.

Jan