Very weird computer/ electronic mystery.

[magnemoe]

Around 25 years ago I worked building and repairing computers, an customer wanted to add an scanner to his computer. Back then most used a scsi interface so you had to install an scsi card in an ISA slot and connect the scanner to the card. 
We had scsi card in storage and the boss ordered the scanner. Customer came with the computer, I added the card, then installed the driver for the card and the scanner, tested that it worked and even scanned some stuff for myself, put the lid on the case and customer collected it. 
He was back the next day as the scanner did not work. 
I started the computer and driver did not find the scsi card. Ok the card came loose or it was an bad connection, open computer put card in another slot, started it up again and it found it, I put on the cover but thought that I should also test the scanner, the card could had come loose in transport. 
Computer cases back them tend to have an U shaped cover covering both sides and the top. it was slotted into the front and secured with 6 screws at the rear as shown in the image. 


Now the weird part starts. I unscrewed the 6 screws while the computer was running and then done I found that the scsi card was detected again. Putting screws in and it disappear. 
Nailing it down to the lower right of the screws killing the card. My first thought was was the cower was twisting the case and the motherboard who could happen. however the screw had an clear pass and did not even have to be fastened for the card to die. 
No its not that it touch anything its bottom screw opposite of the one you see in this image so it would be far from anything. 

I imagined it would be something to do with electromagnetic noise. PC's back then tended to make lots of it in many computer labs radios did not work as most computers ran on 25-50 MHz and FM radio is around 100 MHz. 
But can one tiny hole have an impact. Not that the hole look to small for an 50 MHz wave to pass, its not much bigger than the screen in an microwave oven. 

Any ideas? 
I delivered the pc with just 5 screws attached customer never complained for the year I worked there. 

[StrandedonEarth]

[radonek]

My guess would be that something touched motherboard outside of mount points.  Still,  symptoms beats me. 

[magnemoe]

Just now, radonek said:

My guess would be that something touched motherboard outside of mount points.  Still,  symptoms beats me. 

No it was the screw at the bottom on the opposite side of the motherboard
The bottom part one not seen on the stock photo and also the front part of the case. 
First thing to check second was twisting. 

[wumpus]

I'd strongly believe that you couldn't see whatever was shorting out the board.

But it is also quite possible that ground return noise was the problem.  Most likely that ground current was flowing through your problem screw back into the power supply.  This current happened to flow over some overly sensitive part of the circuit board and induction forces created a current going the opposite way.  And that was enough to make the board not work (probably only sending the wrong bit once every million times, but that's easily enough to crash it in milliseconds)...

You might want to google "ground loops" as well.  Nasty little buggers.  But if there really isn't anything touching where it isn't supposed to touch, and board designers love to use mounting screws as ground, weird grounding current issues are the most probably cause.

[radonek]

The screw probably  just closed a loop that involved something else. I have absolutely no idea what could be, but most likely ended in PSU or motherboard because, well, where else? And I can't think of any PSU problem that would only affect a lone ISA card.

BTW this reminds me of magic switch.

[radonek]

10 minutes ago, wumpus said:

...and board designers love to use mounting screws as ground...

This reminds me something. There used to be combination of plastic and metal distance pillars for mounting the board and IIRC most boards used to ground on one of metal ones. So if that particular distancing pillar is removed (it can be unscrewed),  whole thing is left ungrounded. Now, back bracket of that ISA card have a screw on one side, close to our magic screw...

[steve_v]

1 hour ago, wumpus said:

possible that ground return noise was the problem.

Seconded. I've seen grounding issues cause all manner of weird and wonderful Schroedinger's-faults over the years.  Many computer components also operate at frequencies where you can throw conventional "ground is ground" and "it's just a case" wisdom right out the window.

Just a few weeks ago I had to tend to a sensor that would fail every time one (and only one) particular mounting screw was tightened against the body... Turns out it was creating a tiny ground-loop when the supposedly waterproof case was wet. Enter one plastic washer.
 

46 minutes ago, radonek said:

Now, back bracket of that ISA card have a screw on one side, close to our magic screw...

Edited March 1, 2020 by steve_v

[K^2]

You forgot to put the switch into 'more magic' position.

[magnemoe]

On 3/1/2020 at 12:41 AM, wumpus said:

I'd strongly believe that you couldn't see whatever was shorting out the board.

But it is also quite possible that ground return noise was the problem.  Most likely that ground current was flowing through your problem screw back into the power supply.  This current happened to flow over some overly sensitive part of the circuit board and induction forces created a current going the opposite way.  And that was enough to make the board not work (probably only sending the wrong bit once every million times, but that's easily enough to crash it in milliseconds)...

You might want to google "ground loops" as well.  Nasty little buggers.  But if there really isn't anything touching where it isn't supposed to touch, and board designers love to use mounting screws as ground, weird grounding current issues are the most probably cause.

It was not shorting out anything, the screw was on the front panel on an modern case with back and front panels and many cm from anything with power, now it might ground the cover as the cover as the screw hit metal rather than the paint. 
Stuff back then was noise generators after all an the card was an budget model. 
Gluing on an heat sink and selling an 486 DX25 or 50 as an 33 / 66 was common. 
We had one PC who had an 50 volt difference from other pc's in the networks so PSU was also junk. Solution was the use some napkins the connecting the coax network cable to it.

No it was not Texas because I don't want to insult people living there  

[magnemoe]

On 3/1/2020 at 2:53 PM, K^2 said:

You forgot to put the switch into 'more magic' position.

The magic switch sounded like an grounding issue, ground in the case body was not the same as ground on the motherboard.
Reminds me about the 50 volt pc above but that was on the main power line. 

One idea I thought explaining it was that the last screw might ground the cover. to the case changing  as some commented above, this changed the radio noise environment. 
Cases back then was covered in pretty thick grey paint so cover might not been grounded. 

 

[wumpus]

15 hours ago, magnemoe said:

It was not shorting out anything, the screw was on the front panel on an modern case with back and front panels and many cm from anything with power, now it might ground the cover as the cover as the screw hit metal rather than the paint. 
Stuff back then was noise generators after all an the card was an budget model. 
Gluing on an heat sink and selling an 486 DX25 or 50 as an 33 / 66 was common. 
We had one PC who had an 50 volt difference from other pc's in the networks so PSU was also junk. Solution was the use some napkins the connecting the coax network cable to it.

No it was not Texas because I don't want to insult people living there  

 

29 minutes ago, magnemoe said:

The magic switch sounded like an grounding issue, ground in the case body was not the same as ground on the motherboard.
Reminds me about the 50 volt pc above but that was on the main power line. 

One idea I thought explaining it was that the last screw might ground the cover. to the case changing  as some commented above, this changed the radio noise environment. 
Cases back then was covered in pretty thick grey paint so cover might not been grounded. 

 

If it was an issue of case ground vs. logic ground then the screw shouldn't have made a difference (unless part of the flow from case ground to logic ground or vice/versa) was going through said screw.  It seems more likely the screw was used to ground to logic, but grounding to case is normally how to layout those boards.

I normally tell those new to electrical troubleshooting that if the problem moves around and won't let you isolate it, check power and ground.  Ground is great for a lot of intermittent errors, especially *different* ones.  Isolating the problem to a single screw either means a physical issue with said screw (probably shorting out the wrong thing) or letting ground currents flow where they shouldn't flow (which is a "ground error", but not one you are likely to solve by replacing the power supply and improving your connection to ground).

[K^2]

20 hours ago, magnemoe said:

One idea I thought explaining it was that the last screw might ground the cover. to the case changing  as some commented above, this changed the radio noise environment. 
Cases back then was covered in pretty thick grey paint so cover might not been grounded.

I think grounding the cover is the key, yes, which is why I remembered More Magic. But I don't think it has anything to do with RF. For starters, completing the circuit could only reduce RF interference inside, which should have made things better. No, I suspect something somewhere else was shorting against the case. That's usually your first suspect. Change in capacitance could also do it, but that's a scary thought if something in the circuit is that sensitive.

The most annoying case of ground doing weird things I've had to debug involved an unintended short circuit caused by the fact that oscilloscope used to diagnose the circuit and the power supply the circuit was fed from shared ground by virtue of being plugged into outlets on the same grid. Swapping the leads on the oscilloscope fixed the problem. The circuit was actually fine and operating as designed. Until that day, I never would have even thought to consider the two outlets and the wiring behind the wall as part of the circuit I'm debugging.

[Nuke]

when i was doing computer builds we delivered a bunch of computers to a charter school. the school was for the most part modern, except they had this old timey school house out front. this thing had to be built in the late 1800s and wired up sometime in the 1910s supposedly using the bare wire on ceramic insulators approach that was common back then. all the recepticals were installed external from the wall with drop down conduit, not flush installed like normal outlets and was somewhat more modern, like the 1950s. anyway, for some reason they decided to use this building for the computer lab rather than the significantly larger modern building with good up-to-code wiring. after 3 months computers were failing left and right. i was fixing them and sending them back. this went on for several months. almost all had power supply issues. finally i said what everyone was thinking but didnt want to fess up to. "check your power quality", which of course turned out to be bad. i convinced them to invest in line conditioners, of course at our usual markup. solved the problems. though i dont think the higher ups liked this, they were probably trying to milk them for support fees. they fired me a week later.

Edited March 4, 2020 by Nuke

[magnemoe]

8 hours ago, Nuke said:

when i was doing computer builds we delivered a bunch of computers to a charter school. the school was for the most part modern, except they had this old timey school house out front. this thing had to be built in the late 1800s and wired up sometime in the 1910s supposedly using the bare wire on ceramic insulators approach that was common back then. all the recepticals were installed external from the wall with drop down conduit, not flush installed like normal outlets and was somewhat more modern, like the 1950s. anyway, for some reason they decided to use this building for the computer lab rather than the significantly larger modern building with good up-to-code wiring. after 3 months computers were failing left and right. i was fixing them and sending them back. this went on for several months. almost all had power supply issues. finally i said what everyone was thinking but didnt want to fess up to. "check your power quality", which of course turned out to be bad. i convinced them to invest in line conditioners, of course at our usual markup. solved the problems. though i dont think the higher ups liked this, they were probably trying to milk them for support fees. they fired me a week later.

Yes you can get some weird stuff. At the same workplace we had an issue that after an power out all the computers booted at once and as computers PSU has an huge spike in power demand on boot this triggered the fuse, server was on the same line. 

Having an PSU blowing up on me. This was an a bit especial high capacity PSU and was set to 110 V from factory. We use 220 V, plugged it in and it did not work well. 
Opened it up and replaced an fuse, the fuse had not protected the PSU but probably saved the rest of the computer. 
Heard another sound and after a few seconds an capacitor became an rocket and hit the roof

Got an pc from an house struck by lighting. That was pretty impressive. This was back of the AT standard where computer had an physical power button, this had almost saved the pc but the sound card was connected to an stereo system who had an sleep mode. The lighting strike has charred the stereo unit and continued into the soundcard there it popped the  output transistors, however at this stage the pulse had used almost all of its energy and also burned bridges behind it so it wrecked the motherboard but all the other components still worked. 
 

[Nuke]

ive had random components explode on me. like one i just got it im my bench and took the lid off. then i booted it up, worked fine. next thing i know i hear a loud pop and see a flaming power mosfet flying across the room like a tracer round. 

Edited March 4, 2020 by Nuke