'The capacitors exploded, showering the lab in flaming confetti' Welcome to the very first edition of “Who, me?” a new Reg column we hope will prove as entertaining as our Friday On-Call tales of tech support gigs gone wrong. In Who, me? we’ll celebrate the times techies stuffed up, the lessons learned and the career consequences. To kick things off, meet “Alvin”, who “In the early '90s …
Like coding, where you need to plan for unexpected inputs, so you should with hardware.
After toasting a ZX Spectrum whilst building some custom breadboard hardware plugged into the expansion bus, I too learnt this lesson.
Don't plug prototyped electronics into expensive computers without building a very robust isolation circuit between the two, so only voltage limited data signals can ever get thru no matter how bad you screw up your build.
Don't give your PC to undergraduates? Undergraduates I can handle!
In days of old I ran a lab which had a recently-acquired and very expensive robot arm. It was controlled by a very large and complex PCB covered in ICs, all of them soldered in place. It also had external I/O, with optocouplers for the I and relays for the O. The software ran on an Apple II and the pair were connected via a ribbon cable.
I foolishly took a day off and returned to find that a senior lecturer had demonstrated the I/O features to his class by connecting 230V AC to one of the optocouplers.
Suffice it to say that it took me a very long time to fix the Apple and the robot itself became a decorative paperweight due to the prohibitive cost of the replacement board.
"a senior lecturer had demonstrated the I/O features to his class by connecting 230V AC to one of the optocouplers."
Hence my suggestion "without building a very robust isolation circuit between the two", where 'very' means however large you think that isolation should be to protect against it, think user stupidity, and then multiply that value by 20!
...except the extension connector on the back of the Spectrum was gold-plated PCB traces ("fingers") of the CPU's bus, as-is, and building something that would properly galvanically isolate it while still doing bidirectional I/O at 3.5MHz, while not impossible, was at the time not a practical option for 99.999999% of its owners. Anything less would only have been buffering, hopefully protecting against back-driving the bus but not against any "funny" voltages. So what you actually did instead was design stuff carefully and properly the first time*!
* let's not get into risk multipliers caused by the appropriate PCB-edge connector being unavailable and fabricating your own from two other pieces of PCB comb-slit into flexible-ish individual contacts for each finger, with a bit of wire soldered onto for better-than-PCB-on-PCB contact. Yes, I did get a Kempston joystick interface out of it and yes, I was definitely wearing my brown pants when I first powered the thing up...
Did have a twat hand over his son's car remote control unit to me.
He wanted it repaired.
Every metal transistor in it was exploded out. Every capacitor/resistor cracked in half, etc.
He thought attaching the pp3 9v battery clip to the live and neutral of the Mains would substitute for a missing battery.
"son's car remote control unit ... He thought attaching the pp3 9v battery clip to the live and neutral of the Mains would substitute for a missing battery."
I take it that the Darwin award still applies if, instead of removing oneself from the gene pool, one removes one's progeny. Darwin-by-proxy?
The RSS feeds for this & the On Call posts should be titled as much to make them easier to find among the regular news.
I know if it's from "Simon" then it may be one I'm particularly interested in reading, but if it actually described themselves as "On Call", "Who Me?", or (in the case of Mister Dabbs) "For the weekend Sir?" then we could write an email rule to flag them, play the air raid warning siren, & advise all those around us to leave us alone while we got our weekly infusion of silliness. "Don't talk to me, I'm reading Who Me? again." should be a valid defense for when the H&S folks come running to find out why you've strangled another coworker.
*Cough*
I mean, ummmm... for when you've just come up with a BOFH worthy form of retribution to enact when that nosey coworker three desks over comes trundling past to stick his honker where it doesn't belong.
*Shifty eyed nervous glances around*
I mean, ummm, that I would appreciate it if the RSS feeds could begin with the fact that they contain an On Call, Who Me?, or similar recurring article content, the better for me to know that such an article is available for me to eagerly consume.
*Pure, sweet, & innocent smile*
"I'm sorry Your Honour, I have no idea where that lake came from & I'm not responsable for the crater in which it formed."
/runs away as those nice young men in their clean white coats chase after me screaming for me to take my frog pills...
And whilst you're at it ...
Those of us who get the daily email digest (well, me, anyway) would greatly appreciate it if 'On Call' and 'Who, Me?' could be included in the digest articles titles, for much the same reasons as @Shadow Systems highlights ...
Or better yet, how about a specific feed and/or email with just BoFH, On Call, SFTW and now Who Me? So it can be specifically marked as high importance, must read immediately and a possible solution to get me through until beer o'clock on Friday?
DainB, I don't know where you live but power sockets around here are limited to 10, 13 amps. If it's a 400V socket it may deliver up to 16 amps. (For higher currents we are talking about serious hardware.)
The power supply in my PC could deliver 70 amps @ 12V - enough to roast a few things...
"....n my garage too, I have a 32A supply (@400V), which I could even upgrade to about 70A if needed. But it's rare to find those in office spaces, powering some PCs. ..."
WHAT THE BLARNEY are you powering?
That's 12 800 Watts (32 amp) and 28 000 Watts (70 amps) !!! That's the type of energy we use to power our plasma-arc welding systems or our mainframe busses! What exactly is the sort of computer system are you using to need over 12 THOUSAND WATTS+ of power in a small office?
An old VAX 780 or IBM Zxx series mainframe?
"those 2 mm wide and few microns thick power lines on PCB won't be able to pass them without melting"
Having seen many prototype PCBs go up in flames (despite being made of flame-retardent material) I would say that is not necessarily the end of the problem. As has been pointed out, old switch mode power supplies didn't always have reliable current limiting.
As for 5V being too low to cause a current surge, have you ever used an arc welder?
You forget two things:
1) 30 years ago, the science of Switched-Mode power supply design was less advanced than now. Plus, really cheap "offshore" PSU's were often a bit marginal even when not abused.
2) Cascading failure. Imagine you connect +12V (or -5V) to +5V, you would fry a lot of components powered from the +5V rail on the motherboard . If they then fail short, the PSU can then go into overload. Or even the PSU itself might not tolerate +12V on the +5V rail. Once caps start failing in the PSU, the high fault currents will start killing semiconductors, which could overvolt or reverse polarity the remaining caps. While there may not necessarily be actual flames, I could well imagine much smoke, and an acrid smell.
This post has been deleted by its author
Y'mean theyr'e not now?
I always buy a cheap psu for my computers. One from a reliable manufacturer with decent specs. One that's unlikely to give me weird data errors due to spikes and noise and one with proper protection so that its unlikely to blow up random bits of the machine.
Unfortunately, they tend to be somewhat more expensive. Cheap at the price.
I can still get a new case complete with psu delivered to my address from halfway round the world by UK distie for around 25 squid. Apart from the lacerated arteries assembling the machine, what else do you think I'd get?
>> Thanks - it's the one with the unused first aid kit in the pocket.
""As for 5V being too low to cause a current surge, have you ever used an arc welder?"
On a PCI bus ?
No."
DainB, no matter the rated current of a PSU, if there are large filter capacitors on the output it can supply many times the rated current while those capacitors discharge, even after any upstream short circuit protection has triggered. That is essentially how a spot welder works, so I'll stick to my welder analogy.
You have clearly been very lucky with all of your circuit prototyping. My experience has been quite different.
This post has been deleted by its author
Or he might as well have been testing his 8-segment indicator driver with 220V AC, or as many idiotic scenarios as you can imagine trying to prove your point, but none of them change the fact that it is improbable that short circuit on PCI bus can turn PC into "either glowing, or smoking, or both."
That would seem to be a plausible scenario. If 12V ended up on a lower voltage rail, I could imagine the capacitors on the lower voltage rail going. Alternatively, if the computer was old enough, there would have been a negative supply that one could short a positive rail to.
This reminds me of an experience I had in the early 1980s. My group had acquired a memory expansion board for a Floating Point Systems array processor. Unknown to us, there were slightly different versions of this exact model that differed in the backplane power pin assignments. On being told by the sales guy that "you just plug it in - what's hard about that?" we did just that with seemingly catastrophic results - smoke and, on further investigation, melted PCB traces. Miraculously, we were able to revive it after improvising repairs for the burned traces - it seems that the large current flow that resulted when the negative bias supply on the memory chips became forward biased was distributed over a sufficiently large number of chips to avoid destroying them.
Yeah, had to do that to prevent a meaty arc welder consuming 13 A fuses. (Bit 'o fuse wire soldered across the 13 A fuse to give it a bit more "delay".)
Worked fine until the welder developed a shorted turn. Took out the 30 A fuse for the ring in the main fuseboard rather than the one in the plug.
Remember, there's (almost) always something meater further upstream ... £30 A in the consumer unit, 80A or 100 A main service, 400A in the substation ...
As my first governor used to say "So it steams a bit."
Reminds me of back in those days in a test gear department ... One of the Test Room boys brought in a bit of kit that "didn't work". I whipped off the plug cover to discover that one of his coworkers had replaced the fuse with a 1 inch length of pot spindle - nylon pot spindle.
> I don't know where you live but power sockets around here are limited to 10, 13 amps.
No, your sockets are fused at 32A. They are limited by the total resistance of the wiring and impedance of the transformer that feeds your house, so it's safe to say they're limited to a few thousand amps.
Hmm, electricity suppliers aren't keen on you taking thousands of amps without being warned. That is why they stick a fuse on the feed to the house. Back in the day when I lived in a bedsit we regularly blew the 60A electricity board fuse every winter as we tried to get warm with bar fires. Annoying as it is a call out job to get it replaced :(
completely possible. i had a machine in not so long ago. it had a powercool 550 watt psu in it.
the sata power cables were melted, it absolutely stank, the inside of the case had black smoke / scorch marks on it, the sata connectors for the dvd-rw and SSD were burnt to a crisp, they fell to bits in an avalanche of disgusting, smelly, acrid burnt plastic.
i think what had happened was a layer of dust and crud had built up on one of the sata power connectors over time. its one of the ones that isnt fully moulded, it had little blanks removed on the back of the connector. (for testing the voltage i suppose?). the sata connector was really tight up against the chassis, a mm ish away.
i think the whole thing shorted through the crud buildup against the case.
I didn't believe the customer when they told me what had happened, but after seeing it, i was converted. quite spectacular by all accounts when it happened. I'm not sure about massive flames and flying debris across a whole room, but enough to scare the shit out of you. and 30 years ago, PSU's were indeed very different. (for different read "absolute dog shit"), so yeah, why not.
i now make sure all the psu's i buy dont have those little blanks taken out own the back of the power connectors..... corsair cx 550m's these days.
EA noted, "...power sockets around here are limited to 10, 13 amps."
"Limited" isn't quite the right word in this context. A current "limiter" typically means a circuit with effectively instantaneous current limiting action. If you have a current limiter that limits current to 10 amps, then you'll probably be unable to extract 10.5 amps. This doesn't apply to household wiring.
Fuses are not bad, speed-wise, but can still take SECONDS to open. Magnetically-assisted circuit breakers (Square-D QO series for example) are pretty quick with huge overloads, but can still allow significant overload to remain connected for many seconds. The older thermal breakers are hopelessly inept, taking a Very Very Long Time™ to open. They protect the wiring from catching fire, but will sit there stupidly warming up slowly while you draw 50% over current for perhaps a full minute (YMMV).
May I suggest a Varistor (aka Voltage Dependent Resistor) in-between your vital circuits.
for the bigger cicuits, an entire regulated power supply system with BIG VARISTORS need to be installed.
I think the SurgeX/Ametek company now sells such beasts. Get the biggest one you can afford. The surge gets shunted into a heat-producing/taking resistor that will survive MULTIPLE big hits!
There are SMALLER or LARGER lightning fast varistor setups that have microsecond and less switching/shunting speeds but you will pay a FORTUNE for them!
I once designed and built a rack of TTL logic powered by a 5v 100 amp (or was it 500 amp - I forget) switching power supply wired with welding cable. I once hot swapped one of the cards without turning off the power and there was a noise like a firecracker - POW! followed by a tinkling sound. Investigation showed that one of the chips had totally exploded down to the pins and sent shrapnel cascading down the rack. Replacing the chip fixed it but I never did figure out why the chip exploded.
Another time I pulled a large prototype card plugged into an extension card plugged into a NOVA computer without powering down the computer. This succeeded in making the NOVA unusable. I noticed that the -15v power run on the back plane was next to a system bus line so I spend a day replacing all chips in the NOVA that touched that line.
I was awarded an "ataboy award" for that one: "you collect three ataboy awards and it makes you a leader of man with super powers, the ability to walk on water etc." One "aw shit" wipes out all of the ataboy awards though.
This post has been deleted by its author
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
