Deunan wrote:

U58#5 went to 4.74V? That's pretty much VCC level, shouldn't be possible for a TTL chip, there's something wrong with U58 and that particular signal. Almost looks like it's somehow shorted internally on this particular data line.

... diagram of the internal structure for the LS 245.

page 9 here: http://www.ti.com/lit/gpn/sn54ls245-sp

After all, its analog transistors. Interesting how that propagates through these "digital chips"...

It's always "analog" transistors, even for modern CMOS tech. Unfortunately that's merely an equivalent circuit - and missing the middle part of it as well. You could add just one transistor and a resistor or two to make it work like it's drawn but that wouldn't give you tristate option for example.

Deunan wrote:

U58#5 went to 4.74V? That's pretty much VCC level, shouldn't be possible for a TTL chip, there's something wrong with U58 and that particular signal. Almost looks like it's somehow shorted internally on this particular data line.

OK, retry the test but with 1k resistor. Then also try all the other data pins and measure the voltage. It should be close to 4V but not really anywhere this high. How are your soldering skills, you think you'd be able to replace this chip with a socket? It might be less difficult if you just gently cut the chip out and remove the leftover pins one by one.

For now do the test and I'll see if I can find a diagram of the internal structure for the LS 245. BTW if you would plan on replacing it, and LS was hard to find, then a HCT series would be a good modern substitute.

I can source LS245, my soldering skills are sufficient (I already replaced some DIP40 chips desoldering them with copper braid and soldering in sockets) and I'm going to buy desoldering iron soon too.

As for the test - so I'm supposed to ground O0-O7 pins on the BIOS ROM socket with 1k resistor one by one and then measure opposite side pins (A1-A8) on U58 right?

Well actually not ground but connect to 5V. But doing both of these would be even better, we could test U58 response to both proper L and H inputs, rather than let the inputs float.
But yes, just the O0-O7 pins and measure on A side of U58. I forgot to make this clear, the BIOS socket should be empty for all these tests - but I suppose you figured that out anyway.

Then, if you have some more free time, maybe try testing U58 in the opposite direction. Remove that GND-pin 7 resistor from 8288 socket. This will make U58 A side inputs and the B side (BIOS socket side) outputs. Measure the default voltages on A and B pins to see if the bit 4 sit sill wonky in that direction as well. If it is, it would confirm it's not working as it should.
And then you can also use 1k resistor to pull the pins on A side L/H and test on the socket. Hopefully this test will help us decide to go ahead with U58 replacement or maybe keep looking.

Deunan wrote:

Well actually not ground but connect to 5V. But doing both of these would be even better, we could test U58 response to both proper L and H inputs, rather than let the inputs float.
But yes, just the O0-O7 pins and measure on A side of U58. I forgot to make this clear, the BIOS socket should be empty for all these tests - but I suppose you figured that out anyway.

Yeah, I meant +5V but I will test it with the GND as well. BIOS sockets is empty.

Deunan wrote:

Then, if you have some more free time, maybe try testing U58 in the opposite direction. Remove that GND-pin 7 resistor from 8288 socket. This will make U58 A side inputs and the B side (BIOS socket side) outputs. Measure the default voltages on A and B pins to see if the bit 4 sit sill wonky in that direction as well. If it is, it would confirm it's not working as it should.
And then you can also use 1k resistor to pull the pins on A side L/H and test on the socket. Hopefully this test will help us decide to go ahead with U58 replacement or maybe keep looking.

I will do that. In the mean time I received the second XT board I mentioned earlier. Also MMaximus here XT mobo in AT case: questions about PSU connectors has almost exactly the same board with manual with schematics - hopefully he will share...

Deunan wrote:

Well actually not ground but connect to 5V. But doing both of these would be even better, we could test U58 response to both proper L and H inputs, rather than let the inputs float.
But yes, just the O0-O7 pins and measure on A side of U58. I forgot to make this clear, the BIOS socket should be empty for all these tests - but I suppose you figured that out anyway.

As follows (using 100Ohm resistor):
O-pins shorted to +5V:
Well… this test doesn't make much sense, because all of the U58#2-9 pins are 4,74V at all time, even if none of the O-pins on the ROM BIOS socket are shorted to +5V...

When O-pins are shorted to GND, on the other hand, it pulls corresponding U58#2-9 pins to around 0,1V.

Deunan wrote:

Then, if you have some more free time, maybe try testing U58 in the opposite direction. Remove that GND-pin 7 resistor from 8288 socket. This will make U58 A side inputs and the B side (BIOS socket side) outputs. Measure the default voltages on A and B pins to see if the bit 4 sit sill wonky in that direction as well. If it is, it would confirm it's not working as it should.
And then you can also use 1k resistor to pull the pins on A side L/H and test on the socket. Hopefully this test will help us decide to go ahead with U58 replacement or maybe keep looking.

OK, so again (1k resistor this time):

Tests without 8288 resistor (the one between GND and pin 7, the other one in place):
U58#1: 3,8V
U58#2-9 pins are still 4,74V (all of them).
U58#11-18: ~3,8V
U58#19: 0,02V

With 1k resistor between GND and O-pins of the BIOS ROM socket voltage on corresponding A-pins of the U58 stay at 4,74V, while corresponding B-pins change to around 3,2-3,3V.
With 1k resistor between +5V and O-pins of the BIOS ROM socket voltage on corresponding A-pins of the U58 changes to around 4,74V, while corresponding B-pins stay at 4,74V.

Tests with 8288 resistor (the one between GND and pin 7, and the other one):
U58#1: 0,11V
U58#2-9 pins are still 4,74V (all of them).
U58#11-18: ~1,75-2,00V (those values slowly go down over time).
U58#19: 0,22V

With 1k resistor between GND and O-pins of the BIOS ROM socket voltage on corresponding A-pins of the U58 changes to around 0,11V, while corresponding B-pins are around 0,25-0,46V.
With 1k resistor between +5V and O-pins of the BIOS ROM socket voltage on corresponding A-pins of the U58 stays at around 4,74V, while corresponding B-pins change to 4,74V.

Above tests are repeatable.

Hm. So the test with the resistor has different result now than it had last time when you measured voltages on 3 of the '245 chips. And the only thing different is the voltages on B side were a bit higher now (you said those slowly go down so chances are last time it took you more time to measure U23 before you moved to U58).

The A side of U58 is connected to ISA bus and possibly there are pull-up resistors there. So check that - measure resistance from pins 2-9 to 5V. If there are pull-ups it would explain the unusually high output voltages. BTW did you have any ISA card inserted during this test? Because that also could be the culprit.

If it's not pull-ups then I wonder if a faulty tri-state circuitry could produce such results. In that case I'd replace U58 just because it's acting so strangely. While it seems to pass the logic level almost as good as a CMOS chips, this should not happen and possibly it only looks like that when the switching is very slow - and would not actually work properly at MHz frequency.

Most EPROM burners can test 74' series chips but these are not electrical test, merely functional and also the signal switching speed is not that high. But if you can get U58 out of the mobo with pins more or less intact, you could test it. I'd still replace it anyway.

A small update:

1) Forgot to to clarify that for tests without the pin 7 resistor in 8288 socket the test on U58 should be reversed. That is, in this scenario you measure voltage on the BIOS socket and apply resistor H/L signal on the A side - so, effectively, on the ISA slot because of how it's connected. But for now just test the default voltages and presence of pull-ups.

2) I found a proper internal schematic of a TTL chip with tri-state outputs: the '125 and '126:

I've had some fun figuring out how the floating works, it's more complicated than a modern CMOS. Figured somebody might be interested as well.

Deunan wrote:

Hm. So the test with the resistor has different result now than it had last time when you measured voltages on 3 of the '245 chips. And the only thing different is the voltages on B side were a bit higher now (you said those slowly go down so chances are last time it took you more time to measure U23 before you moved to U58).

The test posted on 2019-4-18 @ 10:25 where I measured U23 U58 and U66? Well yes, but those tests were performed with BIOS EPROM in socket and once I put it back right now - then voltages again dropped on the pins 2-9 to around 1,4V and pin 5 is close to 0,1V. Once EPROM is removed from it's socket - voltages on the pins 2-9 go up to 4,74V.

Deunan wrote:

The A side of U58 is connected to ISA bus and possibly there are pull-up resistors there. So check that - measure resistance from pins 2-9 to 5V. If there are pull-ups it would explain the unusually high output voltages. BTW did you have any ISA card inserted during this test? Because that also could be the culprit.

No ISA cards during all of the tests in the last 7 days or so (I think I removed POST CARD very early - like after first two tests). Resistance between Vcc pin and U58#2-9 pins is consistent and close to around 4,7k.

Deunan wrote:

If it's not pull-ups then I wonder if a faulty tri-state circuitry could produce such results. In that case I'd replace U58 just because it's acting so strangely. While it seems to pass the logic level almost as good as a CMOS chips, this should not happen and possibly it only looks like that when the switching is very slow - and would not actually work properly at MHz frequency.

Most EPROM burners can test 74' series chips but these are not electrical test, merely functional and also the signal switching speed is not that high. But if you can get U58 out of the mobo with pins more or less intact, you could test it. I'd still replace it anyway.

I will desolder it later next week (I want it to do it properly and when I was replacing the caps I found solder on this board hard to work with, so I need desoldering iron to do this).

Deunan wrote:

A small update:

1) Forgot to to clarify that for tests without the pin 7 resistor in 8288 socket the test on U58 should be reversed. That is, in this scenario you measure voltage on the BIOS socket and apply resistor H/L signal on the A side - so, effectively, on the ISA slot because of how it's connected. But for now just test the default voltages and presence of pull-ups.

Well, I suspected I'm doing something wrong… anyway - voltages on the BIOS ROM socket are as follow (without 8288 pin7-GND resistor):
#11: 3,78V
#12: 3,76V
#13: 3,79V
#15: 3,79V
#16: 4,73V
#17: 3,79V
#18: 3,80V
#19: 3,80V

ISA A2-9 pins are 4,75V so no sense in connecting them to +5V as they are already at this voltage.

Connecting A2-9 ISA pins to GND using 1k resistor pulls them down to 0,86V and corresponding BIOS ROM socket pins to ~0,11-0,13V.

Second XT motherboard I bought last wednesday seems to work just fine! It's some sort of ERSO XT clone and it's overal layout and logic chip count is similar to the board we are trying to fix. Manual for this board contains some schematics, so maybe you will find them useful http://minuszerodegrees.net/manuals/Turbo-XT% … ion%20Guide.pdf. I can also remove the CPU and test the CPU from the other board (unfortunatelly it seems CPU on this 'new' board is hard to remove).

HanJammer wrote:

The test posted on 2019-4-18 @ 10:25 where I measured U23 U58 and U66? Well yes, but those tests were performed with BIOS EPROM in socket and once I put it back right now - then voltages again dropped on the pins 2-9 to around 1,4V and pin 5 is close to 0,1V. Once EPROM is removed from it's socket - voltages on the pins 2-9 go up to 4,74V.

Oh, so you did have it in the socket back then. That might explain the different behaviour of the bit 4 then. A few more tests - remove BIOS and measure voltage on the address and control pins of the socket. With both 8288 resistors. Let's see what byte is being addressed (should be ADR [19:0] = 0xfffff but due to the multiplexing of low address byte with data it's a bit wild without the '373 latching).

HanJammer wrote:

Resistance between Vcc pin and U58#2-9 pins is consistent and close to around 4,7k.

Pull-ups, that explains the 4.74V then. Again, a pity, because U58 might be working properly after all. Forget desoldering for now, get an erasable replacement for the BIOS chip. If you haven't yet, also try swapping the CPU and 8288 from the working mobo to this one, and the other way around as well (so that you'll know if the chips are bad or the mobo).

HanJammer wrote:

#11: 3,78V #12: 3,76V #13: 3,79V #15: 3,79V #16: 4,73V #17: 3,79V #18: 3,80V #19: 3,80V […]
Show full quote

#11: 3,78V
#12: 3,76V
#13: 3,79V
#15: 3,79V
#16: 4,73V
#17: 3,79V
#18: 3,80V
#19: 3,80V

Wait, pin 19 is H? Hm, why. Can you confirm that, and also maybe try to find what it's connected to then?
EDIT: Or did you just skip pin 14 while writing down results, and actually it's pin 18?

Deunan wrote:

Oh, so you did have it in the socket back then. That might explain the different behaviour of the bit 4 then.

Yes, I removed it since "2019-4-18 @ 11:01 " post where you asked to measure the resistance of the BIOS ROM data pins. Earlier it was in the socket.

Deunan wrote:

A few more tests - remove BIOS and measure voltage on the address and control pins of the socket. With both 8288 resistors. Let's see what byte is being addressed (should be ADR [19:0] = 0xfffff but due to the multiplexing of low address byte with data it's a bit wild without the '373 latching).

OK. I will do this a bit later today or tomorrow.

Deunan wrote:

Pull-ups, that explains the 4.74V then. Again, a pity, because U58 might be working properly after all. Forget desoldering for now, get an erasable replacement for the BIOS chip. If you haven't yet, also try swapping the CPU and 8288 from the working mobo to this one, and the other way around as well (so that you'll know if the chips are bad or the mobo).

I have some UV erasable 27C256s. I have no means to erase them for now, but I can spare them. I have doubts if I can replace 27C64 with one (will it be sufficient to fill them with the data copy-pasting it at address 2000 then 4000 and 6000?)

I can swap CPU, but not the 8288 (the one on the other mobo is not on the socket and I would rather not desolder it from the working motherboard unless it's really necessary.

Deunan wrote:

HanJammer wrote:

#11: 3,78V #12: 3,76V #13: 3,79V #15: 3,79V #16: 4,73V #17: 3,79V #18: 3,80V #19: 3,80V […]
Show full quote

#11: 3,78V
#12: 3,76V
#13: 3,79V
#15: 3,79V
#16: 4,73V
#17: 3,79V
#18: 3,80V
#19: 3,80V

Wait, pin 19 is H? Hm, why. Can you confirm that, and also maybe try to find what it's connected to then?
EDIT: Or did you just skip pin 14 while writing down results, and actually it's pin 18?

I skipped pin 14 intentionally as it's GND. I will repeat the test.

Deunan wrote:

A few more tests - remove BIOS and measure voltage on the address and control pins of the socket. With both 8288 resistors. Let's see what byte is being addressed (should be ADR [19:0] = 0xfffff but due to the multiplexing of low address byte with data it's a bit wild without the '373 latching).

BIOS ROM socket, EPROM removed, pin 4 and 7 8288 resistor in place:
#10 A0: 3,45V
#9 A1: 3,45V
#8 A2: 3,44V
#7 A3: 3,45V
#6 A4: 3,41V
#5 A5: 3,44V
#4 A6: 3,44V
#3 A7: 3,44V
#25 A8: 3,44V
#24 A9: 3,43V
#21 A10: 3,43V
#23 A11: 3,44V
#2 A12: 3,40V
#11 O0: 1,77V
#12 O1: 1,77V
#13 O2: 1,76V
#15 O3: 1,76V
#16 O4: 3,20V (floating, slowly going down)
#17 O5: 1,54V
#18 O6: 1,54V
#19 O7: 1,54V
#22 /OE: 0,10V

Deunan wrote:

HanJammer wrote:

#11: 3,78V #12: 3,76V #13: 3,79V #15: 3,79V #16: 4,73V #17: 3,79V #18: 3,80V #19: 3,80V […]
Show full quote

#11: 3,78V
#12: 3,76V
#13: 3,79V
#15: 3,79V
#16: 4,73V
#17: 3,79V
#18: 3,80V
#19: 3,80V

Wait, pin 19 is H? Hm, why. Can you confirm that, and also maybe try to find what it's connected to then?
EDIT: Or did you just skip pin 14 while writing down results, and actually it's pin 18?

Again BIOS ROM socket voltages, 8288 pin7-GND resistor removed:
#10 A0: 3,46V
#9 A1: 3,46V
#8 A2: 3,44V
#7 A3: 3,45V
#6 A4: 3,42V
#5 A5: 3,44V
#4 A6: 3,45V
#3 A7: 3,46V
#25 A8: 3,46V
#24 A9: 3,46V
#21 A10: 3,45V
#23 A11: 3,46V
#2 A12: 3,43V
#11 O0: 3,83V
#12 O1: 3,80V
#13 O2: 3,82V
#15 O3: 3,82V
#16 O4: 3,83V
#17 O5: 3,82V
#18 O6: 3,83V
#19 O7: 3,83V
#22 /OE: 3,82V

So confirmed.
Curious thing, that #16 is no longer 4,74V. But it maybe somehow connected to the erratic behaviour I observed earlier with beeps/POST CARD earlier (sometimes it was going crazy, and sometimes it just didn't done anything).

Pin 19 on the BIOS ROM socket is connected to U54 pin 18 and U58 pin 18.
Also I found it's connected to U19 '74 pin 12.
#9 Q pin of this U19 '74 flip-flop is connected to U28 '08 pin 9.
U28 '08 pin 8 -> CPU pin 17.
U28 '08 pin 10 -> U6 '10 pin 8.
U6 '10 pin 9 -> U11 '74 pin 8.
U6 '10 pin 10 -> U14 '00 pin 8
U6 '10 pin 11 -> U5 '00 pin 3
and so on...
should I trace it further?

No need to trace pin 19 any further. Actually it's another mistake on my part, I thought you tested U58 B-side rather than socket. But with this last test I've re-read everything and it falls into place. Pin 16 that was wonky on last test is O4, and here again it's different value than the rest (although falling down now).

I actually have a theory now. A bit of educated guessing was involved but it's not improbable, and it's time we stopped this merry chase around the mobo and actually tried some repairs.

My guess is U58 bit 4 is faulty, but only in the B->A direction and only partially. It can drive to L but not to H - at least, not fast enough to work at CPU speed. So we don't see that on the test because of the pull-ups on A side, but a 4k7 resistor will not supply enough current to reliably drive other inputs on fast L->H transitions.

The low (~1.4V) voltages we got on previous U58/U23 tests were actually not low at all. It's not possible to see it without a scope but I bet there were actually semi-random fast transitions between L and H and the meter simply showed a mean value. That's because you confirmed the BIOS chip was in the socket during that test. Why this happened? Simple, our tests are a bit crude - we are not properly latching the address lines but rather using the '373 transparent mode. This way the 8 data bits get output on the lowest address lines at all times, since it's the same CPU pins.

So let's say the mobo started up and defaulted to 0xfffff address, this would read the last BIOS ROM value which is 0xe3. This value is ouput to data bus, then transferred to address lines A0-A7 near the CPU, and fed back to BIOS chip. The address now is 0xfffe3. In BIOS ROM at 0x1fe3 the value is 0x00, which is now being output to the data bus. And the cylcle repeats, at the maximum speed the ROM is able to output a new value. It's not as neat in real system due to the fact that other chips, and ROM itself, might have sligtly different latency per each data line but the result is more or less the same - oscillations (as long as the BIOS data in last 256 bytes is not the same as the lowest address bits).

There was one difference to that model in the tests though, the line O4 was low at all times. Why would it be, the BIOS data has both 1 and 0 as bit 4, unless the U58 was not able to drive this line fast enough so it was pretty much always L.

TL;DR: It all points out to bit 4 on U58, it has to be replaced to make sure, we cannot test it any better while in motherboard. But, if my theory as to the exact failure mode is correct, it might just be possible to boot the mobo by piggy-backing the faulty U58 with a working chip. Though this doesn't always work - depends on the actual problem with the original chip. But is simple to try.

Deunan wrote:

TL;DR: It all points out to bit 4 on U58, it has to be replaced to make sure, we cannot test it any better while in motherboard. But, if my theory as to the exact failure mode is correct, it might just be possible to boot the mobo by piggy-backing the faulty U58 with a working chip. Though this doesn't always work - depends on the actual problem with the original chip. But is simple to try.

I will replace U58. I already ordered desoldering gun which make this job cleaner, and I will order some '245s as well (I guess I can use 245NC as a direct replacement to 245N, can I?).

For now let's have some brake.

And again thanks for this explanation.

I just tested the CPU from the broken motherboard on the 'new' working one and it works just fine.
I will order parts today so I will get back with results once U58 is replaced...

OK, back to work. I received my desoldering station today and desoldered U58 right away. I soldered in a DIP20 socket. Now, I forgot I have some old dammaged 286 motherboard for parts and I found a lot of 74 logic chips on it - 74F245Ns among others. The question - is it a direct replacement for 74S245N I just desoldered?
Also I tested this U58 chip I desoldered in my TL866II Plus but test came out OK.

I ordered some S245Ns, S373Ns, S244Ns, Siemens SAB8288, some 14,31818 oscillators, 28C64s too but they will arrive probably the next week...

Congrats to your new solder sucker 😀

74F should normally work fine. F means "fast", being near to S (Schottky) in speed, but nearer to LS (Low power Schottky) in consumption.

Did you test the chips in U23 and U66 too?

retardware wrote:

Congrats to your new solder sucker 😀

74F should normally work fine. F means "fast", being near to S (Schottky) in speed, but nearer to LS (Low power Schottky) in consumption.

Did you test the chips in U23 and U66 too?

For now I didn't. Perhaps I will desolder them too and install sockets later, but first I will install this F245 and check if anything changed, but I will do this tomorrow.

F and S are very close in terms of speed but I think F uses less power. In most cases F will work fine as S replacement.
If replacing U58 doesn't cure the mobo, try with the resistors again and check the voltages - I'm curious if the bit 4 will still show different readings. In fact if you get odd voltages still, remove U58 from socket and measure again. There's a '373 connected there that could have a faulty input.