http://www.bioscentral.com/postcodes/amibios.htm
http://postcodemaster.msdtrack.com/AMI90.shtml

Assuming reading from left to right is correct and that as a google search indicated it shows the current, previous error codes respectively, then the following may be applicable.

04 (04 - 8259 Interrupt Controller Initialization) might suggest that something up with the interrupt controller initialization. Or perhaps the board is not making it to the next code 05 (05 - Chipset Initialization Over, DMA & Interrupt Controller Disabled) indicating a chip set initialization error? I know only a little bit about this stuff, but I'm guessing the interrupt controller and a bunch of other stuff that was probably discrete on the original IBM PC are all integrated into the chip set, so either would point to the same issue.

Based on your first video when the machine does boot you're getting to these codes:
30 - Virtual Mode Memory Test About to Begin
31 - Virtual Mode Memory Test Started

That seems odd if it's referring to virtual 8086 mode which was introduced with the Intel 80386 according to Wikipedia at least. So I don't know it means something else under '286 BIOS' or isn't relevant here. That combined with your on screen message "8042 Gate A20 Error" suggests an issue while testing memory. The 8042 would have been the keyboard controller and a few internet sites suggest the A20 gate stuff is controlled by/integrated into? the keyboard controller. So maybe the keyboard controller is wonky or your keyboard isn't behaving properly?

https://technick.net/guides/hardware/beep_cod … uide_beep_codes
^ there is a beep code for the Gate A20 error, so maybe that's what you're getting.

I'd recommend cleaning the board (q-tip and 90% isopropyl alcohol are decent for spot cleaning) with the aim of removing any dust or gunk. Maybe take it outside and give it a couple blasts of compressed air first? After that sticking it under some sort of magnification to check the solder joints might be a good place to start. And it probably can't hurt to pop out, inspect, and re-seat any socketed chips.

I'm not sure why the CPU would be getting hot. As a matter of idle speculation I'd suspect a chip defect, voltage supply problem, or something odd about the way the board is currently driving it. Normally CPUs make heat simply by operating and those which pre-date power states/throttling likely run at full speed spinning their wheels either on code loops or constant NOPs. However it does seem odd that a 286 would be noticeably hot, much less under so little presumed load.

The BIOS chips (AMI marked, socketed), the keyboard controller (probably the long 40 pin chip labeled 'D Jet key'), and the chipset (HT12/A, surface mounted) seem worth scrutiny in particular given the errors. They, and their sockets (if socketed), should be checked for any bad solder joints or other obvious problems. The fact that it's running at all and you get both POST codes and beeps in addition to screen output suggests that the CPU is at least basically functional and the BIOS chips are still readable at least partially.

P.S.
The chips on the far end (by the 8-bit ISA) all appear to be GoldStar GM71C256-10 which should be 256k DRAM and the other similar chips nearby are OKI M41256A 256k DRAM. So they're technically similar chips. So if it's all main ram you have 256 KB of ram. No idea why they're separated into separate areas other than maybe one was meant to be populated with stock ram and the other designated for upgrades.

Yep, at least one of the OPs faults is centered around the fact that the chipset isn't able to control the A20 line. The A20 line does indeed get controlled by the keyboard controller in 286 based systems.

The A20 line is the address line that lets the chipset choose between the first megabyte of system ram and the second megabyte. On old XT systems, the first megabyte of RAM is all there was (no A20 ), so if you tried to access memory FFFF;0010, you would actually access the very first byte of memory, location 0. In other words, with the segment system, memory would wrap around.

The 286 processor however, if you try to access FFFF:0010, would try to access the memory at location 100000, rather than 000000. To maintain compatibility with that old wraparound the XT had, IBM routed A20 through the keyboard controller and then you would program the keyboard controller to either block A20 or not. If you block reaching it the memory banks, so the memory always saw A20 as 0 on the address bus, the old XT behavior occurred when accessing FFFF:0010. If you enable normal A20 use, then FFFF:0010 would access memory location 10000. This is how MSDOS can load a portion of itself high, by using the memory space that would normally wrap around to 0 on an XT.

To do the memory test, A20 has to be enabled so when the 286 switches to protected mode, it can access all the memory. If it can't enable A20, the system board tends to panic and do what the OP sees. The virtual mode memory test probably refers to the fact that the CPU is being placed into protected mode to do the memory test, which is what one would expect.

So yes, definitely check over the keyboard controller, it's solder points along with all nearby traces, the chipset as well. there is probably just a bad connection somewhere causing this, as the system is so close to functioning at this point.

Thanks for the replies. I will check this out.

I inspected the motherboard carefully and I don't see ANY mechanical dammage, not a single scratch. Soldering points are all shiny too. When I bought it few months ago it already had the blue barrel of cancer removed so it never had any chemical dammage as well. There is just a light dust, but nothing that would cause the shorts.

Pics: https://imgur.com/a/WY6ylYP

Also here is the full list of the POST codes logged by the diag board (I just bought it and I wasn't sure how to use it when I recorded videos yesterday) from the last one (when it throws the A20 error, to the first one which according to the manual is just a code that the diagnostic board always shows when it starts). Video is initialized when the code 19 is shown. At 22 BIOS message is shown. Also like I said earlier sometimes the POST is stuck on the code 04. And sometimes it doesn't even show that - just ---- is seen on the diag board):
31
30
25
24
23
22
21
20
1A
19
18
17
15
12
11
10
0C
0b
0a
09
08
07
06
05
04
03
None of the tables from http://www.bioscentral.com/postcodes/amibios.htm have the exact same list of codes…

Sometimes LEDs on the keyboard are all lit up after poweron (sometimes they just give me a short blink like they should). I suspect keyboard controller is shot, too bad it's soldered on (it has only a few pins connected to the board on the solder side, I don't know if these PCBs were just two layer or multi layer). I do have shot 386 board with Jetkey keyboard controller (but it may differ from this one slightly, it has v.5.0 marking on it).

I know this will sound weird but use a different keyboard
the a20 gate handler chip is in the keyboard (you could also try changing fast a20 gate in the bios)

edit: didnt read the last post

Davros wrote:

I know this will sound weird but use a different keyboard
the a20 gate handler chip is in the keyboard (you could also try changing fast a20 gate in the bios)

edit: didnt read the last post

Well I can't change anything in the BIOS because it never shows the setup prompt nor the keyboard is working.

I will try the different keyboard, althout it's the oldest one I have (it's BTC 5140 keyboard).

Also I tried to swap BIOS chips from other similar HT12/A motherboard I have but it didn't worked at all - it didn't even tried to start and showed nothing on the diag board...

HanJammer wrote:

I inspected the motherboard carefully and I don't see ANY mechanical dammage, not a single scratch. Soldering points are all shi […]
Show full quote

I inspected the motherboard carefully and I don't see ANY mechanical dammage, not a single scratch. Soldering points are all shiny too. When I bought it few months ago it already had the blue barrel of cancer removed so it never had any chemical dammage as well. There is just a light dust, but nothing that would cause the shorts.

Pics: https://imgur.com/a/WY6ylYP

Also here is the full list of the POST codes logged by the diag board (I just bought it and I wasn't sure how to use it when I recorded videos yesterday) from the last one (when it throws the A20 error, to the first one which according to the manual is just a code that the diagnostic board always shows when it starts). Video is initialized when the code 19 is shown. At 22 BIOS message is shown. Also like I said earlier sometimes the POST is stuck on the code 04. And sometimes it doesn't even show that - just ---- is seen on the diag board):

...

None of the tables from http://www.bioscentral.com/postcodes/amibios.htm have the exact same list of codes…

Sometimes LEDs on the keyboard are all lit up after poweron (sometimes they just give me a short blink like they should). I suspect keyboard controller is shot, too bad it's soldered on (it has only a few pins connected to the board on the solder side, I don't know if these PCBs were just two layer or multi layer). I do have shot 386 board with Jetkey keyboard controller (but it may differ from this one slightly, it has v.5.0 marking on it).

No obvious sign of mechanical damage is not 100% proof that things will work, sadly. I picked up a 486 board that I couldn't even get to show anything on the screen. Except for briefly flashing the keyboard lights it never seemed to do anything at all. I know the ISA video card I used wasn't broken because I tested it on an old P3 Gateway board and it worked okay (albeit with some odd jitter/artifacting?). And that was before I smashed a glass diode trying to snip through the battery leads. I'm not sure what's wrong on that one, because it could as easily be the CPU, BIOS, keyboard controller, ram, or some other support circuitry.

Unfortunately without more intimate knowledge of 286 board design and ICs of the era. especially those on your board, plus proper test equipment like an oscilloscope and a logic analyzer, determining the exact problem or even completely ruling out possible ones is going to be nearly impossible. Without either trial and error is basically what you're left with. In any case electrical flow can be a funny business and wires, traces, joints, etc can oxidize. And honestly there might be something wrong with one of the other ICs resulting in unexpected behavior that isn't correct.

The reason the tables vary is likely that information is incomplete, but they also refer to different versions of the BIOS which may have dropped some codes and/or included new ones. I was looking primarily at the first table with numerical codes and they seemed to line up ok (19 is "Attempting to pass control to video ROM at C0000h" and 22 is "The power-on message is displayed"). Presumably some codes are common across the entire line of AMI BIOSes, but I don't know if there is a list out on the internet for your specific version. Unfortunately I have no idea what codes your board should or shouldn't show, but the ones that do are a lot better clue than just describing what happens.

http://www.dewassoc.com/support/bios/errors/k … roller_fail.htm

The above link suggests the same troubleshooting as the other guy, swap keyboards if possible. It also points out that the keyboard and controller chip (on motherboard) talk via serial and implies that a poor connection might also cause problems. It seems likely, but not a guarantee that the keyboard controllers are compatible, particularly if they are the same brand. If you wanted to go with trying to swap them, I'd say try reflowing the solder on the connections to the old one first to see if that helps any as getting it out is going to be a fair bit of work. Ultimately you can certainly try a swap if sure it's shot, but without knowing how it failed you might just be dooming the next chip you connect up too...

Do you have any way to verify that your keyboard works?

Is this what you have?
http://www.btc.com.tw/pdf/5140.pd
https://deskthority.net/wiki/BTC_51_series

I don't see an AT/PS2 switch anywhere, so I'm not sure what distinguishes an AT connector version from a PS/2 connector version or if the protocols/behavior are equivalent.

Any chance of a full picture of the rear and maybe a close up near the keyboard port (on the back side)?

P.S.
If you can work out what how the keyboard controller is connected to the other circuitry, you might be able to test some of the traces' for continuity with a multi-meter. It may also be worth checking the voltages at chips when powered (not sure where to stick the ground probe, so maybe someone else can help you there) to make sure they are getting the correct voltages. The supply pin (Vss, Vdd?), which you can find on a datasheet, should probably be somewhere between 4.5 and 5.5 volts on most of the 74xxx logic.

Also since talking to the keyboard controller happens to be a read/write on the main system bus from/to a particular address, at least afaik, it may well be that some chip in between has bitten the dust and/or isn't working right... I have no idea how you'd test it, but it's definitely worth a check if you can find a way. The 74x245 (DM74LS245N) there right next to the keyboard controller seems like a good place to start looking since it's an octal bus buffer/transceiver and it appears to be connected to the keyboard controller on the top side of the board.

Just looking closely at your pics makes me think this is mostly a 2-sided board, but I wouldn't swear to it since I'm very much not a pcb designer/electrical engineer/etc.

Istarian wrote:

Do you have any way to verify that your keyboard works?

Yeah, I have pleny other motherboards and computers I use it with on a daily basis.

Istarian wrote:

Is this what you have? http://www.btc.com.tw/pdf/5140.pd https://deskthority.net/wiki/BTC_51_series I don't see an AT/PS2 switch […]
Show full quote

Is this what you have?
http://www.btc.com.tw/pdf/5140.pd
https://deskthority.net/wiki/BTC_51_series
I don't see an AT/PS2 switch anywhere, so I'm not sure what distinguishes an AT connector version from a PS/2 connector version or if the protocols/behavior are equivalent.

Any chance of a full picture of the rear and maybe a close up near the keyboard port (on the back side)?

Yes. It doesn't have PS/2 at all. Just regular 5din AT connector. I bought this keyboard in 1997. No switches on it. AT is electrically compatible with PS/2 anyway.

Istarian wrote:

P.S.
If you can work out what how the keyboard controller is connected to the other circuitry, you might be able to test some of the traces' for continuity with a multi-meter. It may also be worth checking the voltages at chips when powered (not sure where to stick the ground probe, so maybe someone else can help you there) to make sure they are getting the correct voltages. The supply pin (Vss, Vdd?), which you can find on a datasheet, should probably be somewhere between 4.5 and 5.5 volts on most of the 74xxx logic.

That's a good idea. I will look at the chip references in the internet and try testing the voltages. I will connect the ground probe to the PSU ground lead. It should have common ground anyway.

As it says on bioscentral for AMI BIOS Post Codes (After April 1990):
04 Passed keyboard controller test with and without mouse

I think this mainboard was made after year 1992 week 10 (date code 9210 written on the chipset)

So yes, the keyboard controller or its circuits might be damaged.

PS Did you try to start it without a keyboard ? without RAM ?

Had this A20 gate error on a 386sx motherboard, in the end the fault was defective KBC. My advice is to try and swap it, but if you care about your board, don't try this unless you have proper desoldering tools.

quicknick wrote:

Had this A20 gate error on a 386sx motherboard, in the end the fault was defective KBC. My advice is to try and swap it, but if you care about your board, don't try this unless you have proper desoldering tools.

All IMHO (in theory). If it was known for sure that exactly KBC is bad then removing of KBC could be done just with nippers (to cut pin by pin near the IC package). After that it would be easy to remove the remains of pins and than to clean holes from solder (one by one) with a soldering iron and a toothpick.

I have hot air rework station and I guess I can remove the controller and install socket for new controller instead. Sourcing the proper KBC is other thing though… I have two mobos (286 OCTEK FOX II with AMI KBC) and some other 386 mobo with JetKey KBC. Still I'm not sure if it's required for the KBC to be exactly the same as the one I'm replacing...

Just going to say you might as well check continuity between the KBC and the next chip or component on every connected pin first, just to be sure you've ruled out a simple break in the trace

In my experience, the keyboard BIOS chips were all the same - I tested out my JetKey with a 286, 386sx. and 386DX boards and they all worked.

I bought some proper soldering (or rather - de-soldering) supplies... I'm going to desolder the KBC and solder a DIP40 socket in it's place. Same with the HM6818P chip next to it.

So I fixed it.
I desoldered the KBC (which wasn't that easy - I had to use the pump and braid several times on each row of pins), I soldered a new DIP40 socket, installed the KBC from another HT12 motherboard (pretty much dead one). And voila! 😁

https://youtu.be/w33fLORxsWg

Behold my sloppy soldering joints:

I will certainly have some good time playing around with this motherboard. I love 286s!
Thanks for the help!!!

That is awesome.

Did you end up trying the original keyboard controller in the replacement socket? In case it was a bad solder connection vs a bad chip?

Edit: now it is running, can you do a checkit 3.0 score on it, want to compare with how mine runs.

canthearu wrote:

That is awesome.

Did you end up trying the original keyboard controller in the replacement socket? In case it was a bad solder connection vs a bad chip?

Edit: now it is running, can you do a checkit 3.0 score on it, want to compare with how mine runs.

Yes, I tried original KBC in the socket and it shows exactly the same behaviour as before. One of the pins on it is a bit shorter, but it's not so short it wouldn't come in contact with the socket or solder joint...

Yes I will "checkit" it soon. I have another two working 286 motherboards (one is Suntac with 12MHz Intel and one is 103S with 16MHz AMD) and another three are on the way to me (unfortunatelly one was lost during shipment and I don't think DHL will ever find it 😒 ).

So I made some more tests today, and it seems it's not fully fixed yet. The problem is CPU gets really hot (almost too hot to touch it). So it works well for a while, boots DOS, runs programs for a while and when it gets too hot it just hangs. Even keyboard won't react. When I reset it or turn off the computer and turn it on again - nothing happens. Blank screen. When I allow it to cool down for some time - then it works again like nothing happend. Any guesses what can cause this behaviour? It was getting hot like that before I replaced KBC as well. It's 16MHz Siemens CPU. I would rather not install water cooling on it ;D

Also it seems motherboard doesn't see RAM above 640kB and below 1MB (UMB) - it's not even detected by BIOS and tested during POST. I'm not sure if it's standard behaviour on HT12/A.