Predator99 wrote:

Did you switch the jumpers to CGA as described by Deunan? Any output?

Ah i see, same without BIOS. Therfore the Roms will not help at the moment.

Yeah, I switched them (I've tried both modes). No output.

I replaced the 4,7uF caps with 10uF caps, but nothing have changed.

Just FYI the AT28C256 is pretty cheap 5V EEPROM from Atmel, comes in DIP28 package with 150ns latency. Should be enough even for 486 systems (since there's usually plenty of wait states during BIOS ROM reads). Supported by TL886.

Though if there is a problem with ROM signals then swapping the chip in the socket might not help all that much. It would be best to try and figure out how it's connected to the CPU - directly or maybe through these '244 and '245 transciver chips. If you're up to some light soldering you can also make a simple LED probe that'll let you see if the pins are toggling or stuck with one logic level. Nothing like a proper oscilloscope but it does help considerably.

There was recently a thread where you can read what you could do now to find out how to proceed.
Deunan gave good instructions there, which apply to your case as well.

Yeah, I followed this thread for a while… Which pins should I use with this tester? I think I also can source some 8088-2 I've seen some cheap on local auctions...

A known-good processor is always great when testing. Not a mistake to have one.

Regarding the simple LED tester: You said, the boards' behavior is the same no matter whether there is a ROM inside or not.
So I would look whether there are ROM accesses, i.e. CS, OE, data, address signals etc. Basically the same approach like in that thread.
After one knows what states are at each pin, one could possibly draw conclusions where to look next.

HanJammer wrote:

Which pins should I use with this tester?

All of the BIOS ROM ones, only the populated socket (for now). Test the power delivery pins as well, this is helpful because it validates the result, makes sure the chip is properly powered, and also lets me know which LED color is what signal level.

Will do. I will a couple of day though…
I also found a local source of dirt chip AMD 8088-2s (like a 3 USD per piece).

Deunan wrote:

HanJammer wrote:

Which pins should I use with this tester?

All of the BIOS ROM ones, only the populated socket (for now). Test the power delivery pins as well, this is helpful because it validates the result, makes sure the chip is properly powered, and also lets me know which LED color is what signal level.

OK, I tested it… both without and with reset depressed.

EPROM VCC - 4,7V

1 VPP: bright red
2 A12: bright both / bright red / bright green / sometimes changes state / or quickly alternates after reset/ bright red with reset
3 A7: bright both / bright red with reset
4 A6: bright both / bright red / bright green / sometimes changes state / or quickly alternates after reset/ bright red with reset
5 A5: bright both / bright red with reset
6 A4: bright both / bright red with reset
7 A3: bright both / bright red with reset
8 A2: bright both / bright red with reset
9 A1: bright both / bright red with reset
10 A0: bright both / bright red with reset
11 D0: bright both / bright both with reset
12 D1: bright both / bright both with reset
13 D2: bright both / bright both with reset
14 GND: bright green

15 D3: bright both / bright both with reset
16 D4: bright both / bright both with reset
17 D5: bright both / bright both with reset
18 D6: bright both / bright both with reset
19 D7: bright both / bright both with reset
20 CE: bright green / bright green with reset
21 A10: bright red / bright red with reset / sometimes changes state / sometimes both
22 OE: bright red / bright red with reset
23 A11: bright both / bright both with reset
24 A9: bright both / bright both with reset
25 A8: bright both / bright both with reset
26 NC: bright red / bright red with reset
27 PGM: bright red / bright red with reset
28 VCC: bright red / bright red with reset

Also I'm pretty sure that on D-pins I observe very faint blinking of the green LED when motherboard does the thing with codes - but it's probably too fast so it's almost like the LED is lit still…
It's likely that similar behaviour like on A12, A10 and A6 pins could be observed on other A-pins but it's erratic/intermittent… also sometimes on A12 pin there is bright red, but for a short while the green will blink after depressing reset...

I also noticed that GD74S138 chips are getting really warm after a while…

Also I found at minuszerodegrees.net a motherboard which is almost exactly like the one I have (just minor differences): PIM-TB10 (highres photo: minuszerodegrees.net/xt_clone_bios/DTK% ... v%209).jpg, manual: minuszerodegrees.net/manuals/DTK/DTK%20 ... Manual.pdf)
There are no schematics, but it explains 2 ROM sockets: U38 is for 27C64 ROM BIOS and U49 is for 27C256 ROM BASIC.

Alternative, different manual: http://www.retroprograms.com/IBMPC/PIM-TB10-Z … oard_Manual.pdf

Vogons thread with similar motherboard: Help identifying an XT 8088 motherboard?

Whoa, you did your homework. Too bad there's no schematics but these XT mobos are "easy" to work with in the sense that there aren't usually any higly integrated ASICs that are mobo-specific. Everything can be replaced.

First question, when you assembled the probe, did you use 1 or 2 LEDs in series for each color? With just 1 LED and 5V it's possible both colors will shine even if the probe tip is not connected to anything. Depends on the LEDs used (their forward voltage drop to be precise). If you have this issue and it makes probing difficult then try with 2 LEDs. Originally this circuit was meant for simple testing of 3V3 systems, hence the single LEDs.

So your tests look OK-ish, there doesn't seem to be any address or data lines stuck, except maybe A10 but you did see some state changes here as well.
What is not OK: /CE being green and not toggling, also when mobo is reset (although this can be address decoding glitch, not necessarily an actual problem), and /OE not toggling except apparently being stuck high.

Now, it depends on the actual address latching and decoding method, and control signal processing, but usually you'd expect the /OE to toggle a lot (during any memory read cycle) and /CE when the chip is addressed. And it's not the case here so I'd like to investigate that further.

First check if /CE (pin 20) is not actually tied to ground permanently. Possibly only the /OE is used on this mobo since it's a ROM anyway. If it's not tied to GND then try to find where it goes to, and also where /OE (pin 22) goes to. It should be one of the 74 chips (in both cases) except the '244, '245 and '373 ones (test these last). If you can't find it by optical means then try sweeping the meter probe slowly over the pins of the other chips using the beep function.

Any 74S chip will get warm or even hot with time. 74F even more so. It's normal, you get more Speed but at the cost of way higher power draw and thus heat dissipation. 74LS can get warm but usually not hot, if you get a hot LS chip it's worth investigating if it's outputs are not shorted.

EDIT: It would seem most PC XT boards are closely enough related to the original IBM 5150 to use this: http://classiccomputers.info/down/IBM/IBM_PC_ … 25005_AUG81.pdf
Schematics start near the end, D section. /MEMR is decoded from 8088 state signals by 8288 chip, it's the socketed one next to CPU. See if the probe shows pin 7 toggling. It's possible the /MEMR is directly tied to /OE.
The chip select/enable on the other hand is generated based on 4 highest address bits by a simple 4-input NAND gate - 74(LS)20. So look for it and if you find one, test pins 6 and 8, one would be connected to /CE.

Deunan wrote:

First question, when you assembled the probe, did you use 1 or 2 LEDs in series for each color? With just 1 LED and 5V it's possible both colors will shine even if the probe tip is not connected to anything. Depends on the LEDs used (their forward voltage drop to be precise). If you have this issue and it makes probing difficult then try with 2 LEDs. Originally this circuit was meant for simple testing of 3V3 systems, hence the single LEDs.

1 each color. Both shine constantly if the probe is not touch anything. I can add non-LEDs in series though...

Deunan wrote:

Now, it depends on the actual address latching and decoding method, and control signal processing, but usually you'd expect the /OE to toggle a lot (during any memory read cycle) and /CE when the chip is addressed. And it's not the case here so I'd like to investigate that further.

I will modify blinky-thingy and repeat the tests more carefully.

Deunan wrote:

First check if /CE (pin 20) is not actually tied to ground permanently. Possibly only the /OE is used on this mobo since it's a ROM anyway. If it's not tied to GND then try to find where it goes to, and also where /OE (pin 22) goes to. It should be one of the 74 chips (in both cases) except the '244, '245 and '373 ones (test these last). If you can't find it by optical means then try sweeping the meter probe slowly over the pins of the other chips using the beep function. (…)

Will do.

Deunan wrote:

First check if /CE (pin 20) is not actually tied to ground permanently.

Well, it's definitely shorted to the ground (no matter if I test it on EPROM's socket or one of the ground on the PSU for example). I removed EPROM from socket and this pin is still shorted to ground.

Deunan wrote:

Possibly only the /OE is used on this mobo since it's a ROM anyway. If it's not tied to GND then try to find where it goes to, and also where /OE (pin 22) goes to. It should be one of the 74 chips (in both cases) except the '244, '245 and '373 ones (test these last). If you can't find it by optical means then try sweeping the meter probe slowly over the pins of the other chips using the beep function.

I used meter, It wouldn't possible to find it by eye…
It's only connected to one 74 chip:

Deunan wrote:

Any 74S chip will get warm or even hot with time. 74F even more so. It's normal, you get more Speed but at the cost of way higher power draw and thus heat dissipation. 74LS can get warm but usually not hot, if you get a hot LS chip it's worth investigating if it's outputs are not shorted.

No hot chips - I would describe them as warm - ~40-45*C

Deunan wrote:

EDIT: It would seem most PC XT boards are closely enough related to the original IBM 5150 to use this: http://classiccomputers.info/down/IBM/IBM_PC_ … 25005_AUG81.pdf
Schematics start near the end, D section. /MEMR is decoded from 8088 state signals by 8288 chip, it's the socketed one next to CPU. See if the probe shows pin 7 toggling.

Pin 7 on 82C88 is solid red when board is turned on, and when it does the crazy thing with codes, green light is faintly lit/slightly blinking and when it finally stops doing it - green will be faintly lit constantly… I didn't detected any direct connection of pin 7 on this chip to CPU.
Both according to the manual you provided and on my motherboard pin 7 on 82C88 is directly connected to pin 12 of ISA slot (/MEMR).

Deunan wrote:

It's possible the /MEMR is directly tied to /OE.

Nope, there's no direct connection between /MEMR or 82C88 pin 7 to /OE.

Deunan wrote:

The chip select/enable on the other hand is generated based on 4 highest address bits by a simple 4-input NAND gate - 74(LS)20. So look for it and if you find one, test pins 6 and 8, one would be connected to /CE.

74LS20 pin 7 is connected to /CE (but it's ground) and pin 14 (VCC) is connected-ish (ohm-meter goes crazy). Pins 6 and 8 are not connected to /CE.

PS: I added additional diodes (not LEDs, just regular diodes) in series with LEDs but both LEDs are still lit when probe is not touching anything, but a bit dimmer...

pin 7 of newer (post ~1970) 14-pin TTLs is always GND 😉

You need to find out why the LS138 doesn't get enabled. This in turn leads to the ROM never being enabled. There is some connection between the LS20 and LS138, I am too lazy to look into the schematics now. Follow the signal path and find out.

Actually back then I had replaced a number of 138, they seem to fail more often than average. But as back then many boards were socketed, it was very easy.

retardware wrote:

pin 7 of newer (post ~1970) 14-pin TTLs is always GND 😉

You need to find out why the LS138 doesn't get enabled. This in turn leads to the ROM never being enabled. There is some connection between the LS20 and LS138, I am too lazy to look into the schematics now. Follow the signal path and find out.

Actually back then I had replaced a number of 138, they seem to fail more often than average. But as back then many boards were socketed, it was very easy.

I tested all of the four 138 chips vs 20 with my UNI-T UT61E multimeter (set to continuity buzzer), and then did it again reversing the polarity of the probes…

I marked connected pins with colours - circles are 'beeping' immediately, and lines measure 80Ohm... it seems that red circles behave oddly - sometimes they are Over Limit, sometimes 80 Ohm and sometimes are shorted - it's seems like it's triggered by measuring some other pins against ground 😮

I have no prior experience with TTL… in fact my experience with electronics is limited to analogue circuitry… sorry for my n00bish behaviour here…

Edit: I corrected the picture, also 138 marked with orange lines behves odd… Something is going on between pins 6, 8, 16 (once they shorted, then they are not, or show 80Ohm, also probe polarity matters here).

Well then, it's even more closely related to 5150 than I thought. I figured they'd replace the '138 with something else since this mobo has only 2 ROM sockets, but I guess it's easier to copy a working design and just leave pins unconnected.

Anyway - pin 20 (/CE) of the ROM socket being tied to ground is actually also part of the IBM schematics. What they call ROS (Read Only Socket?) has only a /CS connection but but not /OE. In this case /CE is always active and /OE serves as /CS then. And the '138 is sort of auxiliary address decoder that also takes /XMEMR and /RESET into account.

So, more test:
1) Probe pin 7 of that '138 to see if you get a green LED activity at all. If possible, try to observe if that happens just as you release the reset button. Chances are the CPU starts executing BIOS ROM but is somehow redirected to RAM (some sort of glitch in data for example) and then it'll stop addressing the ROM chip.
2) Probe pin 6, it should follow the reset button state and be low/green when it's pressed, and high/red otherwise.
3) 74(LS)20 pins 6 or 8 should be connected to this '138 pin 5. It should be low/green shortly after reset since the CPU should be executing ROM code.
4) The '138 pin 4 should be connected to a '245 chip somewhere near the CPU. Can't tell you which one, the chip numbers are completly different vs the schematic. There should also be a 4k7 pull-up resistor to +5V so measure that.
5) And finally pins 1,2 and 3 of the '138 should be connected to a '373 chip near the CPU.

Pin 7 of the 8288 is not supposed to be connected to CPU. Other ones are, but this signal is generated by the 8288 itself based on what CPU wants. Which gives me an idea. This mobo is TTL logic and so any floating inputs will default to high. So after you're done with the test above, do this:
- remove the CPU and the 8288 from their sockets
- apply power to the mobo
- probe the '138 pins 1-6, you should get this state: high, high, high, high, low, high
Well, hopefully the DMA engine will not interfere with this. If you do get this state then short the 8288 socket pin 7 to ground with a wire (you can just stick a short wire into pins 7 and 10). This should toggle '138 pin 4 to low state. If so, check the '138 pin 7 now, it should be low as well (and high before the wire).

PS. For these 74 chips (it's not always the case but mostly) the last pin is always VCC (so +5V) and the opposite corner of that is ground. So what you measured is just the power planes.

EDIT: I just realized that you've found a connection between the '20 pin 8 and '138 pin 4. Chances are pins 4 and 5 on the '138 are swapped vs what I wrote above. This would also work.

Deunan wrote:

1) Probe pin 7 of that '138 to see if you get a green LED activity at all. If possible, try to observe if that happens just as you release the reset button. Chances are the CPU starts executing BIOS ROM but is somehow redirected to RAM (some sort of glitch in data for example) and then it'll stop addressing the ROM chip.

All but U39 show no reaction to Reset (on all of them red LED is constantly, brightly lit, on U39 though both LEDs are lit, and when reset is depressed - only red is lit, then after releasing reset - both are lit again).

Deunan wrote:

2) Probe pin 6, it should follow the reset button state and be low/green when it's pressed, and high/red otherwise.

Pin 6 on:
U39, U50, U52 - red LED constanly lit, no reaction to reset
U46 - green LED constantly lit, no reaction to reset

Deunan wrote:

3) 74(LS)20 pins 6 or 8 should be connected to this '138 pin 5. It should be low/green shortly after reset since the CPU should be executing ROM code.

No connection found between pin 5 on any of the '138s to pin 6 or 8 on '20.

Deunan wrote:

4) The '138 pin 4 should be connected to a '245 chip somewhere near the CPU. Can't tell you which one, the chip numbers are completly different vs the schematic. There should also be a 4k7 pull-up resistor to +5V so measure that.

No connection found between pin 4 on any of the '138s to any pin on '245. I found 3 4k7 resistors and each of them measures closely to 4,7kOhm...

Deunan wrote:

5) And finally pins 1,2 and 3 of the '138 should be connected to a '373 chip near the CPU.

No connection between any of these pins on four of the '138 to any of the pins on two '373 next to CPU...

Deunan wrote:

Pin 7 of the 8288 is not supposed to be connected to CPU.

And it's not.

Deunan wrote:

Other ones are, but this signal is generated by the 8288 itself based on what CPU wants. Which gives me an idea. This mobo is TT […]
Show full quote

Other ones are, but this signal is generated by the 8288 itself based on what CPU wants. Which gives me an idea. This mobo is TTL logic and so any floating inputs will default to high. So after you're done with the test above, do this:
- remove the CPU and the 8288 from their sockets
- apply power to the mobo
- probe the '138 pins 1-6, you should get this state: high, high, high, high, low, high

As follows:
U39 '138: red, red, red, red, red, red / and pin 7 is red
U50 '138: red, red, red, green, red, red / and pin 7 is red
U46 '138: red, red, green, green, red, green / and pin 7 is red
U52 '138: red, red, green, red, red, red / and pin 7 is red

Deunan wrote:

Well, hopefully the DMA engine will not interfere with this. If you do get this state then short the 8288 socket pin 7 to ground with a wire (you can just stick a short wire into pins 7 and 10). This should toggle '138 pin 4 to low state. If so, check the '138 pin 7 now, it should be low as well (and high before the wire).

When pin 7 and 10 on 8288 is shorted - IRDY lits up on POST Card and pin readout on '138 is as follows:

U39 '138: red, red, red, red, red, red / and pin 7 is red
U50 '138: red, red, red, green, green, red / and pin 7 is green
U46 '138: red, red, green, green, red, red / and pin 7 is red
U52 '138: red, red, green, green, red, red / and pin 7 is red

Please notice, that there are four '138s total on this board.
U39 is DM74LS138N
U50 is DM74LS138N
U46 is GD74S138
U52 is GD74S138

Oops. When I wrote "that" '138 I meant the one you've traced the connection to from the ROM socket. But it's actually helpful that you did all of them, since this mobo has 4 and the schematic has 5 and the U numbers are different. Helps me map which might be which.

So, I take it you didn't see my last minute edit? I wrote that pins 4 and 5 might be swapped around based on your photo and connections you discovered. So just retest this, and only one chip now:

1) Pin 4 of U50 should be connected to either pin 8 or pin 6 of the '20.
2) Pin 5 of U50 should be connected to one of '245 near the CPU. You said you found the pull-ups, check these '245 first.

U50 seems to the be the chip I was interested in, and the no-CPU test went well. Pin 7 changed state so it is working and should be selecting the ROM socket. That wire you put in 8288 socket tested the /MEMR connection to a '245, and from there (as /XMEMR) to the U50. So both the connection and the '245 (at least for this signal) are OK.
So now we need to check the high address lines A13 - A19.

Since you didn't find that '373 near the CPU it's possible this part of the circuit is a bit different as well. Unfortunately that means more work for you. Now you have to find which one of the '373 and also possibly '244 / '245 on the mobo is connected to CPU socket pins 2 and 3.

Deunan wrote:

1) Pin 4 of U50 should be connected to either pin 8 or pin 6 of the '20.

Yup, It's connected to pin 8 of the '20 like on my picture above.

Deunan wrote:

2) Pin 5 of U50 should be connected to one of '245 near the CPU. You said you found the pull-ups, check these '245 first.

It's connected to pin 18 on the U2 '245. I'm not quite sure what do you mean by pull-up resistors - I'm not familiar with this term... 🙁

Deunan wrote:

Since you didn't find that '373 near the CPU it's possible this part of the circuit is a bit different as well. Unfortunately that means more work for you. Now you have to find which one of the '373 and also possibly '244 / '245 on the mobo is connected to CPU socket pins 2 and 3.

CPU pin 2 to U24 '373 pin 7
CPU pin 3 to U24 '373 pin 13

I didn't found any connections to other '373 or '244s or '245s.

PS: I also found another '20 I havn't noticed before - right next to the lowest ISA slot (marked U86). There are some connections between it and U50 as well.

Deunan wrote:

Oops. When I wrote "that" '138 I meant the one you've traced the connection to from the ROM socket. But it's actually helpful that you did all of them, since this mobo has 4 and the schematic has 5 and the U numbers are different. Helps me map which might be which.

It is possible that this mobo does not have a demultiplexer for the ROS chip selection, as the 4-input NAND goes low when the high address bits are all high. Could well be possible that this directly goes to !OE on all ROM sockets. This can be verified by checking if both (small and large) ROS sockets share the same !OE line.

@HanJammer: just one suggestion: look up "74LSxxx pinout" etc in the web. This will help you understanding what you are measuring (and let you learn a lot) and save us from translating pin numbers to signal names.

HanJammer wrote:

It's connected to pin 18 on the U2 '245. I'm not quite sure what do you mean by pull-up resistors - I'm not familiar with this term... 🙁

Sorry about that. Here's a quick explanation:

In most digital systems (you can assume this to be true for anything made in '80 or later) the 0 and 1 values are represented by 2 voltage ranges. Low level, or 0, is 0V - ideally it's perfectly at GND level but that's not possible, especially with Transistor-Transistor Logic, since it uses silicon bipolar transistors which have collector-emitter saturation voltage of some 0.2V. Therefore TTL logic accepts up to 0.8V as logical 0. Now, the high level, or 1, should be close to the positive power supply VCC. So in 5V system it would be 5V - except just like with the logical 0 it's not possible to get exactly that. In fact because of how TTL gates are constructed internally, the logic high in 5V system is usually at 3.5-4V at best. The standard accepts anything above 2.4V as high level.

To drive the logic input in absence of other signals, or to improve switching speed or noise immunity, you can add a resistor between the logic line and GND or VCC. If the resistor is connected between the line and GND, it is therefore pulling it to low level. So it's a pull-down. In case the resistor is connected between the line and VCC (in this case, +5V) is it pulling it to high level. So it's a pull-up.

Pull-ups are quite common even in TTL systems which default to logic level high, due to the relatively low voltage on output which gets worse the more inputs of other gates are connected.
When you look for one, just check the resistance (on unpowered system, obviously) between the signal you are measuring and +5V. This system would have pull-ups in range of 1k to 10k. In general for a TTL a value less than 100 ohms or greater than 47k is definitelly not a pull-up.

Anyway, either I'm going blind or the original IBM 5150 did not have a reset switch. And it bothers me that you didn't see any reaction to it on the '138. I need to think a bit about your results so I'll make another post later but for now can you check what the reset switch connector pins are actually connected to. I would expect one pin to be ground, and the other to go to pin 11 of 8248 (on the left, beside ISA slots). Possibly it's tied along with the PWR GOOD signal.