1200 volt clearance

If it was responsible, it would be first in my knowledge.

It's not easy to do a proper forensic failure analysis on an off-line psu, but exploded PFC semiconductors are most likely the result of the failure mechanism, than the cause or location of it.

It's a lot harder if end-use isn't accurately documented, and the dead units aren't carefully bagged, untouched, for examination. An accurate and demonstrable root-cause report is usually a convincing document, but it is only proven in downstream performance.

RL

Correction: I reviewed my emails from a couple of years ago (when the faults happened) and it had 0.7mm clearance (and creepage) at the corner of the square pad where many of the failed units had burn marks. That clearance sounds a little low.

I have no record that the manufacturer sent analysis reports for any of the RMAs. Their best guess (in an email) was conductive debris (e.g. metal whiskers from our PCB or chassis) being sucked into the PSU by the cooling airflow.

We found that most of the failures occurred when the system was powered on after a period of storage, which perhaps does indicate a whisker problem (or possibly several other things). Ones we looked at under magnification seemed clean.

Regards, Allan

Working voltage would be sustained. But as usual it depends. On the applicable standard, altitude (that's a big one), environmental conditions, et cetera.

Ouch. Should have been flagged in a thorough design review but such small stuff easily slips through. Then there are companies which ... don't do design reviews.

Do you remember what the peak voltage across that 0.7mm clearance was?

Was the clearance covered with solder mask or bare?

Isolation voltage is 3kV, 1minute, no mention of working voltage

A lot of the isolator products, opto couplers and digital isolators has datasheets that are hard to compare (isolation and working voltages are not tested under same conditions)

Cheers

Klaus

Yes, exactly. And you also need to check the flame ignition rating of the material of the device

Working voltages are the actual voltages, not the rated voltage. So if you have a flyback, in which the peak voltage at the drain is normally higher than the rails, the working voltage is higher

Cheers

Klaus

The problem is that these tests can take very long to perform and UL will perform witness tests to confirm your own tests and that test is expensive

Cheers

Klaus

How does this look?

It's easy in PADS to click on a part, select properties/pad stack, and radius the pads on all parts of that type. My production people say it's not a concern at all, to pick-and-place and solder with rounded pads. Maybe all parts should always have rounded pads.

You seem to be talking only about turns and corners on pads. How do you get a rounded edge along the trace where it has been etched?

We use parylene coating for HV stuff very expensive :(

Beautiful!

My layouter is currently going through the same exercise. I told him to just use the same rounded footprints elsewhere in non-HV areas. Saves CAD modeling time.

Hire a bunch of ants and give the micro sand-blasters :-)

That effect is unavoidable but the chance of arcing is greatly reduced if you reduce the pointiness in at least one dimension, the horizontal one. You don't have to believe me, just read what Johanson Dielectrics engineers say. They know a thing or two about high voltage handling:

the traces will also be usually be covered with solder mask so probably not as critical as the corners of exposed pads

-Lasse

PCB layout can be beautiful. It has a nice visual appeal, and a lot of aesthetics. I like to do a board myself now and then, even though I have a layouter person, or maybe two.

Yeah, I changed all the parts on this board; I kind of like the look. It takes seconds in PADS to tweak the radii on all parts of a similar type.

I suppose I should think about HV issues on vias.

Good stuff. But 3KV 1206 caps? Awesome. I'm using 2220s at 1200 volts.

Maybe you can sneak this onto there somewhere but round the nose a little more:

Best to avoid them in the HV area where possible. If you have a ground plane it might make sense to move that a few layers down from where HV traces are and maybe mesh it (to avoid board warping in production).

Yes. But one should always consider that solder mask can flake off when boards age, depending on environment.

Same here, I am using 2225 at 1300V (rated working voltage is 2kV).

About cutting it close, one of our cars just passed smog. HC max at

on the occasion I have to remove solder mask to hack a pcb, it seems to be indestructible substance and peeling it of ought to make you king of England but I guess if it happened to flake off it would in the most inconvenient place :)

-Lasse