@BitWrangler
That's by definition - minimal or blank autoexec.bat / config.sys

@Deunan

This was done with pins 2 and 3 on the serial port connected?

ugh, i forgot about it. hah
did it quickly without thinking in-between other things going on in my day.
disregard, will update later today.

Actually the result with pins 2-3 shortened is exactly the same.

This could happen if the chipset or TTL chip has one buried bad transistor that only shows up when specific item is used for example serial like in your case.

I'm not surprised, I had a TV with one transistor cell is bad in EEPROM IC and microcontroller had one bad spot somewhere. Shows up on one volume showed one artifact (volume pointer changes in shapes at only one location) in one location and wouldn't keep settings correctly every time tv is off then on. Replaced both fixed the TV.

I had a video card do this same thing. Only one location is bad in memory chip. I found it only in one specific situation only on firefox when I scroll long bookmarks. There is one bookmark folder about 4/5 way down produced 1" wide horizontal band of scrambled pixels instantly appear on the screen.

Diode readings is faster this way using the multimeter as I described to check all the TTL chips to see if junction voltage drop is either good or bad. Ohm checks does not work this way and not clear.

Cheers,

pshipkov wrote on 2023-06-27, 22:19:

Actually the result with pins 2-3 shortened is exactly the same.

I made some changes to the code last minute and I didn't test it, I will do it tomorrow to make sure I didn't give you a broken program. I want to be sure because the result is unexpected.

It looks to be some sort of electrical problem with RS232 drivers, not with the serial port logic. Usually I'd blame the card but since several were tested, and they work in other systems, it has to be the mobo somehow. Perhaps the 12V (both + and -) delivery to the ISA slots it somehow at fault, but the measurements of the serial port pins were OK...

Just one more question, the photo is a bit small so I can't see it very well, is the last result (before FAIL) LSR=0x60 or 0x68? I'm hoping for 0x68...

Deunan wrote on 2023-06-27, 23:42:

Just one more question, the photo is a bit small so I can't see it very well, is the last result (before FAIL) LSR=0x60 or 0x68? I'm hoping for 0x68...

That definitely looks like a 0x60 to me.

actually, I only see one 8 on the photo and that's in the serial port address.

Yes, it is 0.

@pentiumspeed
Yeah, it is either a cold joint somewhere, or specific element being damaged.

pshipkov wrote on 2023-06-28, 06:50:

Yes, it is 0.

I see. And this is exactly the same when pins 2-3 are connected (since this particular photo was done with no connection)? I tested the program, it works as it should, although I have also discovered that shorting the pins is not always very reliable, so make sure your connection is good - otherwise it'll obviously fail the test.

Since I was testing I also connected the scope, here's the results of me using Terminal in NC4, I set COM1 to 1200 7N1 because this is what the mouse uses. Then I've set a single trigger capture on the scope and pressed 'a' key. If you have some time try that and compare what you get . There should be a clean -12V (that's logic 1) and then there is a single pulse to +12V (start bit = 0), then it goes 1-0-0-0-0-1-1 (and another 1 for stop bit, but since all trailing bits are 1 it just goes -12V and stays there).

This corresponds to a = 0x61 (only 7 bits are transmitted) with LSb sent first. Note the voltage levels, low is -11V, high is only about 10V but good enough, clean sharp edges, no noise. If there are any power delivery/quality issues I would expect your results to be much worse.

It looks like the connection was loose indeed.

Oscilloscope result is similar to yours. My reading was slightly more noisy, but i didn't adjust anything before taking the measurement.

pshipkov wrote on 2023-06-28, 22:52:

Oscilloscope result is similar to yours.

what are the hi/low voltages?

pshipkov wrote on 2023-06-28, 22:52:

Oscilloscope result is similar to yours. My reading was slightly more noisy, but i didn't adjust anything before taking the measurement.

Ah, now we are getting somewhere. 0x79 is much more reasonable status, it means byte received but there was a framing error. I'm going to skip the details and just state the obvious conclusion: start bit was detected, stop bit was not, and all received bits are zero. So that would indicate there is a problem with -12V supply.

Please check the resistance between the power connector and ISA slots. Focus on -12V but check all other rails as well, each one should measure no more than 1 ohm (and preferably a small fraction of it). If the resistance is low, then perhaps there is some corrosion or the power connector pins are bent - I had such cases.

If both the connector and resistances look OK then it could be one of the capacitors on the -12V line being open, or shorted but not completly but more like a zener diode (tantalum caps can do that). But I don't see tantalum caps on your mobo, in fact I don't see any big caps there at all. This also might be a problem if the PSU caps are shot and the -12V line is very noisy. So the last thing to try would be to solder 10uF (anything up to 100 will do in a pinch) cap, temporarily, on the bottom of the mobo between -12V and GND. Preferably near power connector or the ISA slot in which you put the serial card. And test again with that capacitor. Make sure you get the cap polarity correct.

Thanks for the insight.
Will check into that soon.

Success !
It was the -12V line.
A rail somewhere flaking on and off frequently from even gentle touches of the components.
So, apparently when i measure the -12V is mostly ok. So even for a moment it shows 0V i assume it is me not holding the probes right and pushing/tweaking them which brings the voltage back.
Pulled a wire on the back to pin B7 and that was it.

Thank you very much guys and especially to you @Deunan for prompting me to inspect this more carefully.
Command ports and how they function was never of big interest to me, but the information in this thread was a bonus to the fix itself.

Glad there was a resolution, kinda suspected that it was a -12 issue all along

Time for retro hobby was restricted to just the essentials in the last several months, so this was a side effect of it - sporadic shallow efforts that resulted in missing the clues. Anyhow. I can see how others probably had a hunch about this from the beginning.

pshipkov wrote on 2023-07-01, 02:34:

Time for retro hobby was restricted to just the essentials in the last several months, so this was a side effect of it - sporadic shallow efforts that resulted in missing the clues. Anyhow. I can see how others probably had a hunch about this from the beginning.

Just that it's basically the only thing on a motherboard that's "serial port specific"

maxtherabbit wrote on 2023-07-01, 02:45:

pshipkov wrote on 2023-07-01, 02:34:

Time for retro hobby was restricted to just the essentials in the last several months, so this was a side effect of it - sporadic shallow efforts that resulted in missing the clues. Anyhow. I can see how others probably had a hunch about this from the beginning.

Just that it's basically the only thing on a motherboard that's "serial port specific"

Indeed, yet everytime it's mentioned it gets scrubbed off as having been measured already, so deeper digging is required to finally get to the problem & solution.
This was fun.

one picture of scope screen would solve it immediately 😀

This was one of the more confusing factors.
It shows on the oscilloscope screen.
When push/press the board during measurement that cold joint makes contact quite consistently.

We all hit cold joints at some point.
Some can be forced to make contact by bending/twisting the board.
Others need reflowing.
This one was a gentle snowflake.