Potting horror...

Ultrasonics *might* show something, but i would guess that would not be dependable.

Even uncured potting is not going to be conductive. Trapped water will.

If he were to un-pot the thing (lots of x-acto knifing) and rip parts off the board, he may be able to spot voids and green electrolytic corrosion underneath.

Potting makes all this detective work much more difficult.

John

On Feb 13, 7:25=A0pm, John Larkin wrote: [....]

It often has ionic stuff in it that goes away when it hardens. I have had potted assemblies made go bad by applying voltage before the reaction is done.

er hammer!"

back in germanium days, i potted some circuits that got cooked from the heat of the resin setting.......

klipstein wrong.

falk willberg wrong.

ThompsonTard wrong, as usual.

I'd bet such a story is as old as the hills, and that such potting materials are as well, and likely not even in use any longer.

I have never heard of a potting media that turns 'bad' simply because the circuit was energized 'before the reaction is done'. To me that means some late point in the process. If you were lighting up circuits on obviously, blatantly 'wet' media, it *should* have exploded on you and splattered all over you, killing you so we wouldn't have to see your stupid claim here.

That was a bad choice for potting material. Obviously a highly exothermic media. If it cured hard, however, the circuit failure may well have been a stress related fracture of a part. Epoxy can snap resistors. Through-hole and SMD both.

AlwaysWrong can't even read.

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

I'll have to ask whether the potters cleaned the PCBs before potting.

I've milled/cracked one module open but the compound seems to be fully cured.

So far 4 modules have been potted. The customer has a different product of which hundreds have been potted by the same firm without any problems.

cs;

id

re

Cleaning the "no-clean" stuff is double trouble. You know that stuff that kills all the birds in a 100 mile radius and rots the "O- zone"[1]. Thats the stuff you have to use to get rid of the "no- clean". If any hydrocarbon touches it, you are forced to use that stuff.

You can often tell by the smell.

We can tell that you are a potting compound total retard. No smell required for that proof.

Most of the potted circuits I've seen outside of automotive were blatantly done to prevent reverse engineering but also drove customers away because they were so horribly unrepairable. Finding a circuit board with a potted section generally irks me off about like finding one where chips have all had their makers part numbers all ground off or dissolved off.

The only potting I ever used was with silicone rubber RTV compound.

Have you determined yet whether the compound is altering the time base wave form?

If it's not, then the suggestion to use some kind of dielectric oil and selectively dipping portions to find the failure was a good one I think.

I would find some way to hang or mount the board so various parts are in the dielectric oil and leave it in that position for a few hours worth of tests before dipping further in and testing a few more hours.

I don't have an engineering degree.

Could you confirm that the four failed potted boards all failed in exactly the same way?

Nope. I have to be at the customer's and they're closed during the weekend.

Well, a lot of people that have a degree can't engineer :-) The answers to the questions 'do you have a soldering iron at home?' and 'do you build electronic devices?' tell a lot more.

It appears its always the microcontroller that behaves erratic.

I'll work on this tomorow.

- I'll check the waveform from the crystal and compare it with the waveform on a non-potted module

- See if there are any open pins on the MCU (the ESD suggestion someone made)

- Check the power supplies, reset, etc

- Have two more modules potted, one with all open pins connected to ground.

- See where the software goes wrong (at one point or random). As someone suggested it might be a race condition which surfaces due to changes in timing.

Please let us know what you find so we can learn!

Waveform and powersupplies look okay (at least, I get the same readings from a non-potted working unit). Tried to alter the software slightly to add more debug but this resulted in a change of behaviour. That sounds like the internal flash may be 'broken' at some point. Depending on which codes sits there, the behaviour will change of ofcourse. Two more potted modules are on their way. Meanwhile I'll try to check the internal flash.

If the two new potted unit perform without flaw, the problem was not your supply, it was the pre-pot preparation that did not get performed on the assembly.

If they also fail, then the circuit has some components (likely one or more caps) that the potting influenced enough to cause the jitters.

Those two new units will help you to determine what the actual condition faced by the first supply is.

I loaded several test patterns into the flash without any errors (the flash has ECC so actual errors may have hidden). After that I tried to load the firmware again. The module has been running fine since then. No more weird behaviour.

Weird. Must have been hit with a few Gamma particles. :-)

"John Larkin" skrev i meddelelsen news: snipped-for-privacy@4ax.com...

Ackkhhh - pfuii! Caching is Evil; roll-your-own complex memory structures are worse still ;-)

I very much doubt that anything that needs just the most basic search functionality will beat SQlite3 in performance and ease-of-programming - SQLite databases reside in in memory too but gives you SQL queries, indexes, e.t.c. on top of that.

Anyway, it's just my opinion based on very little information on what you really need to do so pardon for busting in like that.

Most/Many commercial packages use an XML file as database, then cache everything in memory via a DOM parser; then have to traverse the tree and iterate over *every* sub node to do a parameter search, then it's programmed in Java too to get that Glacial Speed Feel.

I often get the feeling that Java developers think that data structures shall look like the SUN jdk class library ....