Nexxen wrote on 2024-05-23, 11:30:

Do you have a complete sequence with codes?

No, I'm too busy (and lazy) to do a full analysis on the code, I just went ahead a bit to see what is output next. Starting with KBC (and RTC) the self tests get complicated and harder to follow for OK/NG, which is not helped by the fact that RAM refresh is still uninitialized at this point so the code is bit more tricky as it can only use CPU registers.

Nexxen wrote on 2024-05-23, 11:30:

I had a similar case I'm still trying to solve and had codes up to 08-09, randomly stopping before those at different points.

Anyway, if it issues codes could it be a 74xxx dead in between?

If your mobo has a 74F245 driving the lower 8-bit of the ISA bus (usually along RTC and KBC, though sometimes these are separate busses) then consider desoldering it and putting a socket in, if possible, to test a replacement chip. These '245 die more often than other chips - exactly because they drive ISA bus and not all cards are well-behaved (some could be plain broken and shorted somewhere). That being said I had a case of a mobo with a micro-craked via that delivered reset signal to ISA slots. That was very random, would go away if the mobo was out of the case and upside down for diagnosis, and in general gave me a lot of grief.

BTW that '245 should be replaced with another F chip, and if that's hard to find then something like ALS or better yet ACT. Slower series might just not cut it in this application.

Deunan wrote on 2024-05-23, 13:15:

No, I'm too busy (and lazy) to do a full analysis on the code, I just went ahead a bit to see what is output next. Starting with […]
Show full quote

Nexxen wrote on 2024-05-23, 11:30:

Do you have a complete sequence with codes?

No, I'm too busy (and lazy) to do a full analysis on the code, I just went ahead a bit to see what is output next. Starting with KBC (and RTC) the self tests get complicated and harder to follow for OK/NG, which is not helped by the fact that RAM refresh is still uninitialized at this point so the code is bit more tricky as it can only use CPU registers.

Nexxen wrote on 2024-05-23, 11:30:

I had a similar case I'm still trying to solve and had codes up to 08-09, randomly stopping before those at different points.

Anyway, if it issues codes could it be a 74xxx dead in between?

If your mobo has a 74F245 driving the lower 8-bit of the ISA bus (usually along RTC and KBC, though sometimes these are separate busses) then consider desoldering it and putting a socket in, if possible, to test a replacement chip. These '245 die more often than other chips - exactly because they drive ISA bus and not all cards are well-behaved (some could be plain broken and shorted somewhere). That being said I had a case of a mobo with a micro-craked via that delivered reset signal to ISA slots. That was very random, would go away if the mobo was out of the case and upside down for diagnosis, and in general gave me a lot of grief.

BTW that '245 should be replaced with another F chip, and if that's hard to find then something like ALS or better yet ACT. Slower series might just not cut it in this application.

This was instructive, thanks.

Thanks for the help and your time Deunan,

I check data/bus connections:
- kbc and rtc: connect to special address bus of HT101 chip
- roms : address lines connect to special address bus of HT101 chip, data lines connec to HT102 (near cpu)

I don't found problems between rtc,kbc,roms to chipset lines.

I'll try to check sn74ls245an but in this board is connected to high address lines A17-A23 of 16bits ISA.

An useful datasheet to check lines
https://theretroweb.com/chipset/documentation … 0d411422518.pdf

ggalvan wrote on 2024-05-24, 06:56:

I don't found problems between rtc,kbc,roms to chipset lines.

That '245 of yours should not be causing problems this early in the boot process. Devices like RTC and KBC are 8-bit only and addresses via I/O so the upper part of the ISA bus doesn't really matter to them.

Do you have the ability to also write to EPROM chips? I could prepare some test "BIOSes" that will poke around RTC and KBC and display some codes. Test the busses, compare write and read, etc. This might be more informative than the silly tests that original BIOS does. However that very rarely gives any answers on the first try (or even second), usually multiple programs need to be tried to narrow down the search area. So you'll need something to reprogram EPROMs without too much trouble, and also I would suggest getting a Flash based EPROM emulator chip - unless you have a UV eraser lamp and a bunch of EPROM chips on hand already.

Deunan wrote on 2024-05-24, 09:02:

ggalvan wrote on 2024-05-24, 06:56:

I don't found problems between rtc,kbc,roms to chipset lines.

That '245 of yours should not be causing problems this early in the boot process. Devices like RTC and KBC are 8-bit only and addresses via I/O so the upper part of the ISA bus doesn't really matter to them.

Do you have the ability to also write to EPROM chips? I could prepare some test "BIOSes" that will poke around RTC and KBC and display some codes. Test the busses, compare write and read, etc. This might be more informative than the silly tests that original BIOS does. However that very rarely gives any answers on the first try (or even second), usually multiple programs need to be tried to narrow down the search area. So you'll need something to reprogram EPROMs without too much trouble, and also I would suggest getting a Flash based EPROM emulator chip - unless you have a UV eraser lamp and a bunch of EPROM chips on hand already.

I know about a bios made by Deksor that only displays a "01" to check whether a board is dead or not.
To help understanding, you'd create and image that goes through some codes that will work "if" RTc, keyb, ... are connected and functional? That would help.

The BIOS I traced on a 286 board is connected to A30 through A16 on the ISA bus (Addresses 1 to 15) on 8-bit XT bus. POST cards work on that.
@ OP: are A30-16 of ISA bus connected to BIOS chips?

I've made a few test "BIOSes" in the past. As I see it there is little point in making a big combo tester if you don't know how badly the mobo is messed up, and that goes double for remote testing that somebody else will be doing. So I prefer to make simple test programs and test things one by one - takes longer but less chance of getting wrong results. PM me if you are interested in that for your own mobo, I don't want to derail this thread.