Well, i recapped the motherboard same thing happens. BTC one works until I try one of "LED keys", Logitech one doesn't work at all.

So that settles that, not capacitor related...

Tomorrow I should get another socket 7 and one super socket 7 motherboard so I'll see how they behave there.

A properly designed mobo should have a fuse for the keyboard port power rail. On late 486 and newer mobos it can be a polyfuse, and these can age and it results in higher resistance. Which can greatly limit the current it can pass without significant voltage drop. I'd investigate that, replace the fuse if necessary (for short-term testing with known good keyboards you can bypass the fuse as well, but keep in mind you are risking mobo damage).

Does the VCC pin stay at or close to 5.3 VDC while booting, or does it drop more than a couple of tenths of a volt below 5?

Given that two different keyboards malfunctions, dropping/fluctuating voltage would indicate a problem with something before the measuring point, i.e. the supply. Stable voltage with something after. Also, 5.3VDC is somewhat on the high side, as if the supply never was properly loaded.

Have you checked your adapter? The mini DIN connector looks like something you've salvaged from a broken mobo. Probably a good idea to clean it thoroughly with some electronics spray etc. Also, are you absolutely certain about your soldering? No cold joints etc?

Shouldn't be too hard to open up the keyboard and check the voltage in that end of the VCC line while booting. If there's a difference, the problem is likely somewhere between the measuring points.

Deunan wrote on 2022-02-02, 00:37:

A properly designed mobo should have a fuse for the keyboard port power rail. On late 486 and newer mobos it can be a polyfuse, and these can age and it results in higher resistance. Which can greatly limit the current it can pass without significant voltage drop. I'd investigate that, replace the fuse if necessary (for short-term testing with known good keyboards you can bypass the fuse as well, but keep in mind you are risking mobo damage).

Hmm, I'll have to check to see if I can find this fuse on the motherboard. Beside this only thing that I can still do is check how lines behave with oscilloscope. I have small portable oscilloscope (DSO DS211) that should be good enough to check this...

Unfortunately testing with these two other motherboards will have to wait. I just remembered that they have only AT power connectors and I have only ATX PSUs, I'm still waiting for ATX to AT power adapters to arrive 😬. This QDI board has both AT and ATX connectors so I could use ATX one directly.

kaputnik wrote on 2022-02-02, 07:25:

Does the VCC pin stay at or close to 5.3 VDC while booting, or does it drop more than a couple of tenths of a volt below 5? […]
Show full quote

Does the VCC pin stay at or close to 5.3 VDC while booting, or does it drop more than a couple of tenths of a volt below 5?

Given that two different keyboards malfunctions, dropping/fluctuating voltage would indicate a problem with something before the measuring point, i.e. the supply. Stable voltage with something after. Also, 5.3VDC is somewhat on the high side, as if the supply never was properly loaded.

Have you checked your adapter? The mini DIN connector looks like something you've salvaged from a broken mobo. Probably a good idea to clean it thoroughly with some electronics spray etc. Also, are you absolutely certain about your soldering? No cold joints etc?

Shouldn't be too hard to open up the keyboard and check the voltage in that end of the VCC line while booting. If there's a difference, the problem is likely somewhere between the measuring points.

Haven't checked yet, I forgot about that mini oscilloscope I have. I'll test today after I get back from work.

I've checked 5V points on ATX connector for continuity and it's not directly connected to 5V on AT port. Or maybe that fuse is the culprit so it interferes to much.

Connector was brand new, wasn't desoldered. It's simply PCB version, what I had at hand from few projects (mostly PS/2 to Amiga mouse adapters). There are no shorts, tested after I soldered it and after I "glued it". No shorts, every pin connected to required one. I tried pulling on the wires after I soldered them to test if they were were solid on there.
Yesterday I tested continuity on adapter while wiggling both sides of the adapter. Just to be sure that there isn't and intermittent connection but it held, no drops.

Solo761 wrote on 2022-02-02, 08:08:

I've checked 5V points on ATX connector for continuity and it's not directly connected to 5V on AT port. Or maybe that fuse is the culprit so it interferes to much.

Some motherboards might use a more advanced method of current limiting the PS/2 ports than a simple fuse. Something more like USB power monitoring. In that case there would be a way to cut the power to the port via a MOSFET switch, or something of the sort, which can be in off state by default with no power and drive applied. That would also explain why you can't find a direct connection between the port and +5V line.

Welp, it's interesting... When I don't have anything plugged in I have 5.2 - 5.3 on AT connector.

When I turn it on with BTC keyboard connected it drops to 4.5V, keyboard still works. I turned on caps lock and now it still worked, though voltage dropped further to around 4.2V.
I then turned caps lock on and it started blinking, voltage was oscilating between 3.2V and 4V. I guess it went too low, so keyboard "crashed" and stopped drawing power, voltage got up, keyboard started to boot and voltage dropped back to 3.2 so it "crashed" again and so on...

With Logitech keyboard it went straight to 3V on power on and it kept oscillating between 3V and 4.7V.

That no name was the most interesting. I guess it's the newest from all of them and most modern and has widest range of acceptable voltage input.
On power on voltage was 4.2-4.3V, with num lock it went down to 3.5V, with num and caps lock it went further to 3.1V and with all three leds it dropped to 2.8V. But it remained functional all the time.

I've checked where 5V trace from AT connector goes and actually it comes via 2N3903 NPN transistor.
It has 5V from PSU connected directly to collector pin, base pin is connected to 12V line from PSU via 1k resistor and it's emitter pin goes to AT 5V connector.

I might change this transistor with "fresh" one and see if anything improves, luckily 2N3903 is pretty much a standard transistor and I have them on hand...

*Edit*

Almost forgot, got one super socket 7 motherboard today which has ATX connector so I these keyboards via this DIY adapter on it. All three keyboards work fine with it. So I guess something is definitely up with this motherboard.

Solo761 wrote on 2022-02-02, 22:28:

I've checked where 5V trace from AT connector goes and actually it comes via 2N3903 NPN transistor.
It has 5V from PSU connected directly to collector pin, base pin is connected to 12V line from PSU via 1k resistor and it's emitter pin goes to AT 5V connector.

What a silly setup, I'm having hard time figuring what was the point, because it's not really much of a current limiter except if you short it, then it'll just blow. What's more, if during power off 5V line drops faster than 12V one, and the keyboard has any electrolytic capacitors, it'll reverse the transistor. Modern bipolars really don't like that, will degrade even after a single event and many will surely bring the beta down a lot.
Replace it, and if it's still misbehaving then just remove it and maybe put a mini fuse for some 500mA between collector and emiter.

It does sound weird. 7V ish from input of 1k resistor to Emitter, so something like 6mA in to the Base. 3903 datasheet says current gain min 50ish at that current for Vce=1V. So at least 300mA before the voltage would drop below 4V. But the maximum continuous Ic is only 200mA. So it would actually act as a rubbish fuse. On the upside, all the reverse biases look ok, nearest is Veb which allows up to 6V. Although that's the absolute max rating and it doesn't sound good to run so close to that.

Maybe it's an aggressive but silly keylock that can be bypassed by supplying the keyboard with external power? Keylock switch could connect in between the 1k resistor and the base, and pull the base to Ground, which would turn off power to the keyboard. Or part of some way to allow 5vsb to power the keyboard if using the ATX power connector by disconnecting it from main 5V is 12V isn't present?

Board is QDI Titanium IB+, from cca 1997.

I know that older AT boards used one time fuses, so if you plugged your peripherals hot you could fry it and then it would be "no more worky", later ones used "smart" resettable fuses.

But this is one of the later ones and yet it uses this setup. Maybe 3903s were cheaper than fuses so they went this route. And it is better to have some limit on AT 5V line, if it was connected directly to PSU 5V line you could short everything out much easier (let's just say that back in the day when AT boards were norm one guy in my high school managed to plug AT keyboard into BNC connector of network card 😅).

@Deunan

Maybe something like that happened and that's why it has problems now. We'll see what happens when I put new one in. And what will this one say when I put it in tester 😁.

@snufkin

That also crossed my mind, to make another AT-PS/2 adapter that will take 5V from somewhere else on the motherboard, maybe even USB port, I have to check where that one goes. But longterm it's better to fix (or better said restore functionality, it's not necessarily fixed by restoring this 😅) this than to work around it.

This is how it looks (kinda, it's cramped so it's hard to take a picture). These pins above AT connector are for PS/2 header for mouse. I wouldn't be surprised if this transistor also powers this. I forgot to check if it's connected to it. And I'll have to check with magnifier where this 5V leads to exactly. There's continuity between 5V pin(s) on PSU connector and collector of this transistor, but in this image it looks like it turns towards AT connector, though maybe it's just routed that way...

Replaced transistor and now all keyboards work.

Tested voltages and now it's solid at 5V regardless if nothing is plugged in, or most power hungry keyboard with all LEDs turned on is plugged in.

Motherboard had 5 of these 2N3904 transistors, to be on the safe side I replaced them all. They were easy to desolder, no huge ground planes to soak the heat 😀.

And one from keyboard power rail (also connected to PS/2 port power header) was definitely bad. Tester detects it as two resistors, while other 4 detect correctly as NPN transistors, same as new ones. One of them on some test runs detects as two diodes so maybe it is a bit flaky. But now all are replaced so it should be good 😀.

Thanks for the tips, hopefully this thread will also be of use to someone else with similar problem 😀.