fried VME modules

søndag den 6. juni 2021 kl. 17.59.40 UTC+2 skrev snipped-for-privacy@highlandsniptechnology.com:

I could get out my X-chapters resistor nuker and try it on some mated connector pins.

I don't think I've ever seen anything about peak current loading of connector pins, sort of a SOAR curve. H+H did some great stuff on resistors.

It would be harder to get the smart intern!

Maybe I can get a crate and characterize the loading behavior of the power glitch.

Borad?

Pure cruelty. ;)

You probably don't much care if the contacts weld, provided the crate and supply limiter board are otherwise undamaged--so you have to leave it plugged in, oh well, it'll keep doing its job.

Cheers

Phil Hobbs

You know, that Sacha Baron Cohen character.

Cheers

Phil Hobbs

I do NOT want my gadget taking down a test stand. It must be understood that my plugin can only go into a slot in the crate that is truly unused, since I might blow out some pins.

Once one of our modules shorted out and burned a hole in the backplane board. That was unfortunate.

This is a likelier scenario than it might seem at first sight. With the feedback of the stepdown messed up at the moment - as John's posted shot suggests - the stepdown switch might fight until death, . it won't take that long. An SCR/zener plus a fuse on the boards would have been a good idea on design time but it is sort of late for that. With some luck it might be possible to just load the thing enough to not exceed 5.5-6 volts while it tries, shorting it completely might not be so good.

Dimiter

In any case, it would be an extremely good idea to test it in a spare card cage _yourself_ before shipping a unit anywhere. I hope one of those sites has a spare.

Make it a bastard of a test. Making things worse would undo all that good will you're trying to achieve with this.

Dumb response. Do your FPGAs use hundreds of amps? No? So it should be easier for a single MOSFET to quickly cut the supply than to sink enough current to overload the external supply.

I can't see why this would be a more diffcult retrofit than a device that'll just move the failure mode into the external supply.

CH

Welded contacts probably have lower resistance. It's a feature, not a bug. I wonder if there's some surface treatment that will promote spot-welding.

Making a kelvin connection to the power bus (that is, using different pins for shunting and monitoring) is probably a good idea.

afaict that would involve modifying every module rather than just plugging in an extra module

Why bother with a big load for a few milliseconds? Can you use a resistor + capacitor as a snubber to make the overshoot go away?

A mosfet circuit added to the many VME waveform generators? The +5 power input pins connect to an inner layer power pour.

Plugging another module into a crate is pretty easy.

If I can clamp the rare startup overshoot, there won't be any more failures.

Something around 20 to 50 farads with milliohm ESR might work.

If the 5V rectifier has circa 100 milliohms resistance, and isn't the sensed item for a switchmode feedback loop, and the overshoot lasts a couple of milliseconds, my pencil suggests that capacitance is three orders of magnitude high.

Four UPW0J682MHD1TN in parallel, for instance. There should be some 5V drain, so that the second or so of switch interval does a big part of discharging them, of course.

Of course it's a switchmode feedback loop. That's why it overshoots.

Blowing up the modules shuts down the system - permanently.

Better yet, put in a reliable power switch.

Unless these crates are on a satellites, change the out the f****ng power supplies. Bandaids on a bogus power supply is just dumb.