fried VME modules

Well, you're going to have to pull something, Shirley.

Something like an HUF75652 has a gate-source voltage that's pretty steady around 4V from 5 to 50A. A zener from the load-connected drain to the gate and a 3K9 gate shunt resistor should clamp a start-up transient.

How was the low power on/off hardware configured? Hacking that 'package' makes more sense, if all it needs is a delay. Just stick the POR circuit in-line.

RL

There's a difference?

Can they access the PS connector to the backplane? Put a device between there. The module connectors are going to weld or something if this is triggered a number of times.

And what if they plug it into the backplane at the end opposite the PS? What will the other modules see?

Something in the loop compensation probably doesn't have time to discharge properly. I doubt that the schematics of the power supply are available, and the crates are bolted into test racks across three countries. A no-assembly-required plug-in fix is preferred, if it doesn't don't melt the connector pins.

It's good politics to make the effort. Interesting too.

Take one of the supplies out and find another way to hack into it.

Maybe fix the switch.

Maybe find a shutdown or feedback input.

Very likely. Expensive blunder. It could have been a power supply design error too.

So, you are hacking their hack. Shunting hundreds of amps might save your module, but moving this expensive fuse to the VME bus power rails as fuse. Can the bus rails handle hundreds of amps?

My 4-layer clamp board will have giant power and ground pours. The limit will be welding the six +5V pins on the P1/P2 backplane DIN connectors. I'll add maybe 10 milliohms of trace resistance per pin, before I hit my pours, to help equalize the currents per pin.

The spike is invoked by precise fiddling with the clumsy push-push power switch. Turning the supply off for just about 1 second makes the

"Crowbar" usually means to fire an SCR (or some mechanical device) to short a bus to ground. It doesn't recover until the power source is cycled off and on again.

"Clamp" means to limit the voltage on the bus, to about 5.5 in my case.

Adding the shunt to the backplane is one option, but plugging into a VME card slot is the preferred field upgrade. Something in series with the power supply would work but would be harder to install in the field.

Maybe some magical power strip could work too. Gotta think about that.

Backplanes ususlly use big PCB power pours or actual copper busbars for power dustribution... pretty low overall resistance.

I found an old email with the transient waveform.

In 20 ms, it's barely worth heak sinking the fets.

Heat sink won't get the pulsed heat out of the junction fast enough. A small SMT TSSOP or even SOT23 can do just as well as a TO-247. I doubt you really need hundreds of amps anyway.

Pease talks someplace about his first commercial high powr power supply design. To test the short circuit protection, his boss got out a coarse steel file and ran it back and forth across the terminals, creating a shower of sparks. Pease called it "a bastard of a test." ;)

Cheers

Phil Hobbs

(it was a bastard file)

If it's really hundreds of amps, ISTM you may run the risk of blowing up the switch in the SMPS, which could get really ugly, depending on the topology. A buck switcher with a blown switch would not be a good neighbour.

Cheers

Phil Hobbs

Yeah, those SCR + zener crowbars are really nasty--they cannibalize their own gate drive, making sure that the turn-on is really slow. That trashes the dI/dt tolerance of the SCR and leads to localized melting.

Good crowbars have fast, beefy gate drivers.

Cheers

Phil Hobbs

Hey. I got it. I got it.

Yeah, plus you don't make a pile of jet engines awaiting testing. That can get expensive.

I spent some of my formative years doing firefighting for advanced chip fabs--having 5000 wafers backed up in his sector will really get the manager's attention. ;)

Cheers

Phil Hobbs

Well, the Aussies here aren't native speakers, after all. ;)

Cheers

Phil Hobbs

We'll find out! Maybe the current limiter still works.

The other question is whether a VME module plugged into the backplane can clamp the +5 supply spike, specifically if 6 pins in the DIN connector would weld in 10 or 20 ms at the unknown peak current. I'll add some equalizer traces on my board to help balance the pin currents.

One could make an estimate, based on the delta-T required to melt copper. Heat conduction speeds up quadratically at short distances, so the amount of Cu you'd have to melt isn't as small as you might think.

Cheers

Phil Hobbs

NXP have similar demo of their "unbreakable" LDMOS, setup as a rf amplifier making 1200W disconnect the load and poke around in the connector with a screw driver