majestyk wrote on 2023-02-02, 12:49:

- When the mainboard starts, you can see it go through numerous POST codes, before it finally stops at "3C".

I've read that for AMI "3c" is "Programmable Interval Timer (8254) is being set". But 8254 is in 82c206 that has been already changed

BTW, It's an interesting unusual 386 motherboard. Maybe it was a server or a workstation...

majestyk wrote on 2023-02-02, 12:49:

I recently bought an old full-AT 386 mainboard (SIS "rabbit" chipset) in non-working condition.

Speaker being quiet might be a false lead, don't bet too much on that. If the BIOS hangs (in a loop or due to HW issue) rather then actually report an error you will not get any beeps.

First - nice mobo. But big ones like that often get bent and microcracks can happen. Also do give the bottom layer a good visual inspection, since it's 2nd hand and had parts removed, somebody might have cut traces on the bottom too - simply by dragging the whole PCB around over some sharp edges. That and traces around edges and all mounting holes are always suspect.

Does the 3C code change at all when the RAM is missing? Did you try to reset the mobo manually via reset switch?

Yes it´s a nice - and rare - mainboard and I´d love to see it working again!

I did check for scratches and cuts on both sides of the PCB _very_ carefully. There were no serious findings and I measured all traces when in doubt. It seems there might be just 3 layers here - upper and backside and one layer in between for ground and supply voltages.

When I first tried I had the external battery connected with wrong polarity, since the connector has no blind pin and there´s a "+" marking directly at the connector that in fact belongs to the battery´s "+" pin. This was the case during the tests with both peripheral controllers and I thought they might have been damaged by the reverse polarity voltage but there´s a diode decoupling the external battery connector from the other sources (3.6V battery and PSU 5V). So I think that´s all o.k.

The POST messages seem to be random, everything between "-- --" and a long row of subsequent messages is possible.
When using the older (black and white) AMI BIOS, pressing reset in most cases has no effect on the messages, the "CLK"LED on the POST card goes out as long as reset is pressed, the "reset" LED is lit all the time and this indicates some serious issue. I had mixed up the "reset" and "clk" LEDs all the time because the POST card sits "upside down" when in an ISA slot 🙁
It also seems to not make any difference if RAM is present in BANK0 or BANK3 or not at all.

majestyk wrote on 2023-02-02, 18:36:

Yes it´s a nice - and rare - mainboard and I´d love to see it working again!

If there's a diode then the battery mistake should have no consequences. Random or missing codes would suggest you have a BIOS issue. This can also be reset related but figuring out if all chips receive proper reset signal is more difficult so lets focus on BIOS. I would suspect the ROM is not being accessed properly - like a data line is weak and glitching or address line is stuck or disconnected.

It seems that BIOS chip signals are generated by the 85C320 nearby, check all connections and pins. I see quite a few '245 chips nearby, possibly one of those is connected to the ROM data lines - and these buffers can degrade or even die eventually. Problem is, weak (or slow) but not dead chips will still pass tests in most EPROM programmers capable of testing 74/4000 series. Only dedicated 74 testers can detect such faults, if at all. With no other ideas you might consider desoldering a '245 that's connected to BIOS ROM, and putting a socket in. BTW if you do try to remove old chip without damaging it, ALS are getting harder to find and F is even more rare now.

For now try to test the ROM address pins, preferably with the chip inserted and testing on its pin directly - in case the problem is with socket not making good contact (but from experience, these sockets are not bad unless corroded or having more than 100 removals/insertions).

Obviosuly the perceived BIOS issue could very well start at the CPU socket too (see if you figure out which pin the repaired traces went to, might be a hint) but I'd start with the ROM socket and nearby chips.

I rechecked all traces, connections and solderings yesterday and I´m quite sure everything´s o.k. there.

Reset signals are o.k at the CPU, UM882B206L and SIS 85C320.
On the ISA bus slots "Reset-DRV" pins are at 5V all the time. Just wondering where this signal is coming from?...
-> It´s coming from Hex-Inverter U46 (SN7406N) pin 12 :

I probed the inputs and outputs with my scope:
Pin 13 - 5V (connected to Pin 2)
Pin 12 - 5V

Pin 11 - 5V
Pin 10 - 3.5V pulsating once per second

Pin 9 - 5V
Pin 8 - 5V (connected to Pin 5)

Pin 6 - 3.5V pulsating once per second
Pin 5 - 5V

Pin 4 - 5V (The speaker is connected to Pin 4)
Pin 3 - 0V

Pin 2 - 5V
Pin 1 - 0V

The ISA reset line is pulled up with 4.7KOhm (resistor array) but for whatever reason it´s not pulled down by the inverter output L=|F when F is high.
When I ground Pin 12 the "rst" LED on the POST card goes out (of course).

The 32KHz RTC oscillator is working, the 14.3181 oscillator as well. Bus clock output at Pin3 of 85C320 is at 4Vpp, at the ISA sockets (Pin B30) it´s at 1Vpp).

There was never any traffic at one side of buffer U45 (74F245) so I replaced the KB-controller that´s connested to it´s T/|R input. Now there IS data traffic and when I pull down ISA reset line the system starts!

majestyk wrote on 2023-02-03, 08:57:

On the ISA bus slots "Reset-DRV" pins are at 5V all the time. Just wondering where this signal is coming from?...

ISA reset is connected to chipset but on earlier mobos it is often routed via some OC gate like the 7406 you have there. This is because ISA cards can drive this pin low too (that's very rarely used like that). So OC or OD output plus pull-up resistor is correct. And BTW most ISA cards will not work without correct reset pulse, or if the signal is asserted at all times.

I might have misunderstood your earlier comment about the reset LED on the POST card - are you saying it is not changing state when reset button is pressed (or briefly on power-up)? If so, that's definitely a problem that needs to be solved. The gate in '06 is between 12 and 13, and if you see a pulse at 12 but not at 13 the '06 is busted. In fact I already suspect it right now because your voltages are wrong, it's a negation so with 5V input it should have ~0V output but it's always nice to have two things point to a problem, less chance of making mistake when measuring or writing things down.

I'd say desolder and replace the '06. They do die, one of the reason of using a separate gate is to protect the chipset ouput from bad ISA cards that have some lines shorted to VCC or GND. If you don't have spare on hand you can easily emulate one gate with small signal NPN transistor (plus say a 470 om resistor to limit the base current). Reset is not timing critical at all, any common NPN transistor will work.

Oops, in fact it´s a 7406N (4069BUP is nearby and I read the wrong chip).
But it´s a HEX-Inverter anyway. I do have some TC4069BUP on stock - do you think I can use one of those?

majestyk wrote on 2023-02-03, 11:21:

Oops, in fact it´s a 7406N (4069BUP is nearby and I read the wrong chip).
But it´s a HEX-Inverter anyway. I do have some TC4069BUP on stock - do you think I can use one of those?

Pinout is the same but 7406 has OC (open collector) outputs, and 4069 has push-pull outputs. It might just seem to work for a while until the output of the 4069 gets damaged when it's trying to output H and something else drives L on that signal (or, worse, some other, more important chip gets it output busted). So no, if you need a temporary fix to continue testing use a single NPN transistor:

pin 13 -> resistor -> NPN base
pin 12 -> NPN collector
GND (pin 7 will do) -> NPN emitter

Resistor is not critical, 100 ohm to 4k7 will probably work fine, 470-1000 ohm is preferred.

Also, the 4069 is CMOS and low drive part, not suitable to be driving multiple TTL inputs on ISA cards. The usual role of 4069 on these older mobos is a low power oscillator circuit for the 32kHz crystal (plus a gate or two might be used for some other purposes like combining reset signal from PWR_GOOD input with something else, but not as a final driver).

EDIT: To make this clear, the transistor fix explained above can be safely added to the mobo with the 7406 in place. No need to desolder it or isolate the busted gate. In fact it would work just as well if the gate was operational, that's the benfit of OC outputs on 7406.

Thanks for the warning! I managed to find a 7406N in the meantime, removed the defectice one, soldered a DIL14 socket in it´s place and inserted the new inverter.

The system is starting now, the ISA bus RESET issue is fixed, reset switch is also working now and speaker sound is working.

During memory test I can abort testing by hitting "Esc", I can reboot by hitting "Cntrl + Alt + Del", but I´m not able to enter BIOS setup by hitting "Del". The message "Press Del. to enter setup..." disappears and the system stops / hangs. POST code is "7F 7E".

I tried several BIOSes for this platform to no avail.

Weird, usually BIOS will demand to enter setup if CMOS checksum is incorrect, not continue booting. Perhaps the message is misleading, did you try other key combinations to enter BIOS?
BTW, did you install any ISA cards that carry their own BIOS (like SCSI for example)? Other than VGA, that is already initialized if you have screen output.

I have an old Biostar 386 BIOS chip (later, coloured version) that allowes me to enter setup and reports checksum errors, but that BIOS doesn´t work completely correct, it´s probably for a different chipset.
There´s only the ISA VGA card populated at the moment.

Well, depending on the date of that AMI BIOS (pre- or post-1990) the 7F might refer to BIOS extensions on add-on cards or attempting to enter the BIOS menu (which you are trying). So since there no other cards with ROM present I would assume the BIOS hangs while trying to enter/show its menu. The question is why would that be - either wrong BIOS for this mobo or perhaps it's corrupted.

All I can think of at the moment is to try different video cards, in case there is some sort of incompatibility, and to look for a different BIOS (can be different mobo but with same chipset). But for that you'll need EPROM programmer of some sort and a fresh EPROM or Flash-based emulator chip.

O.k., everything´s fine now.
I tried the BIOS of the Chaintech 325S (April 9th 1990) and managed to enter BIOS once so I could configure everything. Then I connected a floppy drive and populated 12MB RAM and 8 DIL 8Kx8 cache chips.

Phew, this wasn´t one of the easy ones, thanks for the help and advice!!

majestyk wrote on 2023-02-03, 15:54:

O.k., everything´s fine now.

Great! Do you have 4MiB sticks to see if the mobo will accept them? If so there is a chance it could accept up to 64MiB, a rarity for 386 (and early one too). Though obviously just because 16MiB is recognized in one bank doesn't mean all banks can be populated like that - and I'm assuming you don't have that many 4M sticks or you'd have already tried that.

I do have 16 4MB sticks, but 4MB (9-chip) sticks are being detected as 1MB ones (3-chip ones are not working at all).
So 16MB it is on this Modula Tech.

I think it´s a BIOS limitation, the "Rabbit" chipset can handle 32MB. The later (coloured) 386 AMI BIOSes for this chipset don´t work on this board, while in turn the early b/w ones do work on later "Rabbit" mainboards.

Actually the "rabbit" on the ASUS ISA-386c MH doc says 64Mb using 8x4mb simms on board and 8x4mb simms on the mem expansion card. Some of the others at TRW show max of 16Mb and some show 32Mb....
Very few show using 4mb simms at all...

This one supports 32 MB:
Troublesome mystery 386 "rabbit" mainboard

I still havent managed to get 256K cache working though...
In the SIS datasheets it says this chipset sepports 16 MB.
Maybe I will give the ASUS BIOS a try.
Edit: The ASUS BIOS doesn´t work on this mainboard just like all next generation 386 AMI BIOSes.

The "not being able to enter Setup" error was back with a vengeance this morning accompanied by additional strange effects like beeps and hissing from the speaker during POST.
The situation improved, when "JP8" was being pushed to the side during POST. So I reflowed the nearby 82C310 chip suspecting some deterioreted solderings there. But this didn´t help.
A further examination revealed 4 damaged traces between this chip and the TAG chips. There were several infinitesimal small dots and cuts hardly visible even under a good magnifying glass. Not the usual visible cuts and scratches.
I had to repair all 4 of the traces with wires to (finally!?) restore full function.

This reminds me of a broken via I had on one mobo, the crack was on the bottom layer but every time I put the PCB upside down to test it, it would bend differently ever so slightly and start making proper contant. Took some time to find.
Anyway, I had a question - I see you like to protect your rework with some sort of transparent lacquer. What do you use? Something pro or nail polish?