Buck switcher fails when potted

RTV162 is recomended.

Cheers

They fail like that when you exceed Vin Max. I would clamp the input. And make sure you clamp negative Vin spikes. And like Spero pointed out use a soft potting compound or a electronics grade RTV (RTV162) around the device.

Cheers

Good thought. If it cracked the core, say, life could get ugly pretty fast.

Cheers

Phil Hobbs

Some HV manufacturers dip their magnetics in RTV before potting.

Cheers

Drill baby drill ;)

Cheers

Do you have a diode or resistor or something across the relay coil to get rid of the spike when it switches off?

Can you stop the switcher chip from being potted and see if that makes a difference? Something like a plastic pen barrel or piece of tubing over the chip would do it... give it a little "chimney" to free air.

The inductor has already been mentioned. There was a post here several years ago that mentioned a possibly similar problem, but with ferrite beads:

Matt Roberds

thermal conductive properties of the potting will be less.

That part does not rely on package conduction, the heat is conducted throug h a copper lead frame out the pins and into the copper pour. You could pack it in thermal insulator and the junction temperature will be negligibly di fferent from using air or epoxy encapsulant.

n to any other suggestions.

If you develop any kind of intermittent connection in just about any of the solder joints, that will kill the chip real fast regardless of its self-pr otection features. You need to change to a potting compound specifically fo rmulated for SMT construction.

On a sunny day (Fri, 7 Mar 2014 10:15:13 -0800 (PST)) it happened snipped-for-privacy@yahoo.com wrote in :

If the potting gets _under_ the part, could it be it is lifted of the PCB tracks, maybe at one point, and you get some failure due to bad connection (pin) or bad thermal contact to PCB traces?

The other thing that occured to me is that some potting materials may get hot when settling, how high is the temperature of that substance? And finally is that an hermetic package, something may get pressed into it?

re

Does it fail immediately on power up or after some time with power dissipation?

Does it fail with light loads vs heavy loads?

I've had potting compounds that turn conductive at high temperatures. If the potting is black in color, that may be clue.

I agree with the potential for oscillation causing overheating but the temp limit should be handling that. it must be a transient on the output.

Could also the capacitance of the epoxy making a tuned circuit with the inductor. This could cause spikes of the inductor goes discontinuous. That would reverse bias the transistor. Is there a diode from pin 7 (a) to pin 5 (cath) to input to catch a spike like this and shunt it around the internal NPN?

You're running close the the switch rating so this is a possibility.

Yup, if it is the inductor, that's the most likely failure mechanism. The only other might be shorted turns in the inductor, but you'd think an electronic potting matirial would be compatible with magnet wire.

If it is cracked cores, you could probably detect it without even opening the epoxy, just with a scope or frequency counter and a pickup coil. If the switcher frequency is wrong, you know that unit is bound to fail sooner or later.

Jon

OK, well, that blows one really NICE theory! Have you looked at the soldering before removing the chip? Just wondering if the hardening epoxy was somehow breaking solder joints, maybe by getting under the chip and prying it up.

You can almost certainly pick up the switching frequency with a small pickup coil from outside the unit. Might be instructive to read the frequency accurately on 5 units before and after potting. If the frequency changes, you KNOW something strange is going on inside. One crazy story I have heard is some guy was potting tuned circuits with a bunch of toroids in them, and the frequency response changed several hours after potting, then slowly went back to near correct over the next couple days. it was determined the epoxy was cracking the toroids, and then as the entire mass of epoxy fully set, it was returning everything to the original shape. If you are testing these modules shortly after potting, there may be internal heat generation or uneven stresses that screw things up, and this has all settled down by the time you open up the modules.

(Just a lot of random brainstorming here as to what might be going on...)

Jon

How will that help? The LM2574HV is fixed-frequency, not current mode.

Best regards, Spehro Pefhany

Glue a thermocouple onto the top of a switcher, then pot it, then run it through whatever test it takes to make it fail.

Better yet, scatter several thermocouples around, pot them all, then collect lots of data for your effort.

Are you sure that the IC is the initial failure, and not a victim of some other problem?

Is the potting compound pressing on the inductor windings, causing a short?

Is the potting compound causing an _open_ circuit in the inductor?

Is the potting compound (I'm reaching, here) getting into the inductor windings and causing a change in inter-winding capacitance? I can't imagine that being the issue unless you're on the edge without potting compound, but still...

Does the circuit fail immediately, or after some time?

Can you probe the circuit operation while its potted, to see if anything is different right off the bat, rather than after heat has worked its evil magic?

Re : this,

If you check the bad IC with a meter, does it seem like the output is shorted ? These things when designed to run an FET really aren't up to alot of current. That means anything loading the gate drive to the FET could affect it adversely.

I think if the unductor was the problem, the MOSFET would have failed, not just the IC.

That was similar to my line of thinking. Operating close to Vinmax, and does that parameter have a tempco? Notwithstanding the bloggs view on thermal considerations, the chip temp will be higher in the potted situation.

I'd blame thermal stress. I've encountered some surface mount chips that are crazy sensitive to strain. The board has to be the right thickness and the solder needs to be at just the right level on the pins. Anything else and the chips crack internally. A big chunk of epoxy would certainly do it.

I'd recommend a soft enamel to keep dust and moisture out.

Maybe a dollup of heat sink compound would do the trick ?

Are there any ferrite beads involved?

The beads MUST magnetostrict slightly to operate properly; otherwise their characteristics can dramatically change.

Ridigly potting them is asking for trouble.