The Curious Case Of The Exploding Switchmode

[dexters_lab]

somewhat interesting problem here...

In my line of work i have used these particular modules, they are designed to switch 110 or 240v AC motors used in electric blinds (shades), they consist of a power supply module, Atmel uC, current transformer and switching relays. They have a considerable amount of programmable options used to operate blinds and are addressed and operated over their own module->module communication bus (through the RJ11 sockets) which is in turn accessed through a RS232/422 interface. They are manufactured in the USA.

I have had a number of these fail over the last couple of years in different installations, failure is always the same; a catastrophic failure of the power supply module. As you can see in the pictures the live supply trace to the psu module is usually vaporised with additional burn marks around the live supply connector. Some cases the psu module is cracked open.

From what i can ascertain the psu modules fail when first powered.

Today i had to visit a client and found 6 of these modules blown. Used in a domestic application from a 240v UK supply.

The manufacturer of the controller module does not admit a problem and i cant find any issues with the psu module on an internet search.

What i cant understand is how this can happen, the module (Bias Power BPS 1-14-00) according to the datasheet should be able to handle anything that is thrown at it, but i maybe missing something

I'm just curious... any ideas?

Pics show the whole module (with flash/burn marks), and the guts of one of the psu modules

[lewis]

Crap design. Looks like the PSU modules lack transient protection (can't see any varistors or transzorbs in there), or perhaps some components are underrated. (Although the FQD2N90 is rated at 900V so seems pretty roomy). Have you ascertained the failure mode? Is it the MOSFET, the bridge rectifier, the caps or the transformer primary?

I've seen some stuff designed in the US allegedly designed for 110/240V operation where the designer uses 400V rated semiconductors (bridges, fets etc), presumably because they're used to designing for the US market and 400V provides plenty of headroom. But in the EU, conventionally, when designing for 240V we'd use >600V rated components - 240V mains at +6% tolerance is 360Vpk, not much headroom for 400V rated components, and that excludes surges.

Forgetting that, the most likely failure is transient overvoltage. From memory CATII appliances need to withstand 2500V transients according to EN60950 (can't remember the exact waveshape), there's no way the bridge and fet in the photo could withstand that without external protection.

Could also be counterfeit components, happens to even the most reputable of manufacturers. Or, if they really did fail at power-up, inadequate inrush limiting or poor start-up characteristics.

[HackedFridgeMagnet]

From what i can ascertain the psu modules fail when first powered.

Just an idea, could it be some occasional start condition in the processor or the switching circuitry that initially leaves the transistor(s) on for too long.

Is it synchronous?

[Stonent]

How failure prone are those cheap ceramic oscillators I wonder?

If you can power the chip externally you could always check the oscillator was working.

[Mr Smiley]

From what i can ascertain the psu modules fail when first powered.

Just an idea, could it be some occasional start condition in the processor or the switching circuitry that initially leaves the transistor(s) on for too long.

Is it synchronous?

Boot loader 

 

[senso]

I don't see how a bootloader on the atmega would interfer with the closed module that is the power supply.

[SeanB]

I would propose a small prefilter board, with a common mode mains choke, a pair of class X caps across the mains, a pair of class Y to ground, a NTC thermistor and a VDR on the output side along with a fuse on the input side. Very likely transients breaking down the primary side components, especially if you have multiple failures in a single house or area. Basically you just need to clone the input side of any semi decent PC SMPS , and add the transient suppression they rarely have as well.

[dexters_lab]

thanks for all the great replies!

Senso, HackedFridgeMagnet, Mr Smiley:
As far as i know there is no bootloader on the uC or even an external icsp port etc. normally the uC would just blink a led a couple of times on power up. In some cases the low voltage components do survive after as i have powered some them externally, but most of them are fried and release more magic smoke if i try and power them with my bench power supply.

Lewis:
Excellent points there. I dont know the exact point of failure on the module. I didn't disassemble any further than you see in the pic. Though most of the burning occurs always around the legs of the FQD2N90 mosfet. The 8 pin device just next to the mosfet is also sometimes blown out.

[dexters_lab]

for those curious..

more disassembly of the psu module...

one board is HV the other LV, transformer sandwiched in the middle and two interconnecting wires from the HV side to a opto on the LV board. Black cap is across the bridge rectifier and is rated at 400v 3.3uf.

Unfortunately the pins of the transformer pulled out as i as trying to de-solder, but there does seem to be burning around them.

[poorchava]

I'd say voltage surges/transients. Did you have any of those devices fail in a 120V-grid country?

[dexters_lab]

I'd say voltage surges/transients. Did you have any of those devices fail in a 120V-grid country?

no, i have never installed into a 120v location, i only work within the uk

[JDubU]

Take a look at this app note from the manufacturer of the module:

http://citapower.com/wp-content/uploads/AN-125_2.pdf

[dexters_lab]

Take a look at this app note from the manufacturer of the module:

http://citapower.com/wp-content/uploads/AN-125_2.pdf

interesting and useful note, thanks for finding the link

[dexters_lab]

ok, i think on balance we have an answer, simple lack of input protection and maybe the property that blew out 6 of them is a bit prone to electrical issues.

to be fair while i was there they did mention they had other inexplicable failures

thanks for all the input 

[Evil Lurker]

Exactly how old is that thing? I say that because one of the caps is made by G-Luxon which merged with another company back in 2005.

[dexters_lab]

the controller is maybe 4 years old from when i bought it, the power supply module has a datecode C10943, so maybe late 2009?

[megajocke]

I'd suggest showing those pictures to the supposed manufacturer of that power supply module asking whether they actually made it or not. That thing looks absolutely horrible! No fuse for starters, and I see no mention in the datasheet that an external one would be required. There is no wonder that the trace on the board got vaporized when the power supply module failed.

And further, can you really rely on that grey goop to insulate everything properly? The live optocoupler wires (going from one board to the other) look like they are very close to the output capacitor, and the metallic capacitor ends seem to be very close to wiring on the opposite PCB. The transformer core also seems to be too close to both secondary and primary side parts.

It looks like either the power supply module was misapplied by the maker of the blinds controller (they should have included a fuse, for example) or that the power supply module should have included the required protection. Judging by the datsheet, I'd say the power supply maker (the actual one, maybe not Bias Power) is to blame here...

The manufacturer of the blinds controller might also be shocked at what's inside that PSU module if you showed them. :O

[dexters_lab]

The interesting thing to me is the maker of the controller module has been making devices like this for many years (15+), their previous generation used a transformer, bridge rectifier and a linear reg.

This particular module is their newer type, released just before i bought the ones that like to pop, where they started use to this smps module, now the interesting point is they switched back to a transformer on their latest version. So the smps module design was only used for a couple of years.

Personally i think they had a lot of failures when it operated on 240v and they just switched back to a transformer rather than fix whatever problem it was, maybe a combination of bad design from bias power and poor implementation my the controller maker.

[Richard Head]

There is another possible failure mode other than simply a crap design.
I assume the motor is mains driven.
If the power feed to the PC board is interupted while the motor is running under load the energy (1/2LI^2) stored in the airgap of the motor magnetic circuit can cause a sizable transient on the PSU input. This transient can only occur when the power feed to the board is interupted since the mains impedance is far too low to allow a transient while connected.
I have seen this failure mode on several occasions with small induction motors but assume it can happen with small series wound motors also.
The solution would be either an RC snubber across the motor or a varistor or both.

Dick

[lewis]

There is another possible failure mode other than simply a crap design.

If the power feed to the PC board is interupted while the motor is running under load the energy (1/2LI^2) stored in the airgap of the motor magnetic circuit can cause a sizable transient on the PSU input. This transient can only occur when the power feed to the board is interupted since the mains impedance is far too low to allow a transient while connected.
I have seen this failure mode on several occasions with small induction motors but assume it can happen with small series wound motors also.
The solution would be either an RC snubber across the motor or a varistor or both.

Leaving out a snubber or TVS sounds like crap design to me!

[lewis]

And further, can you really rely on that grey goop to insulate everything properly? The live optocoupler wires (going from one board to the other) look like they are very close to the output capacitor, and the metallic capacitor ends seem to be very close to wiring on the opposite PCB. The transformer core also seems to be too close to both secondary and primary side parts.

I was thinking the same thing. There's not a lot of room for error there.

The various EU standards do allow a significant reduction in creepage and clearance distances if a unit is potted. 0.6mm or 0.9mm springs to mind, would need to look it up. (No air gaps are allowable so the potting has to be done under vacuum.) The gap is ~0.2-0.4mm inside optocouplers, so the distances through insulation can be very small.

[dexters_lab]

There is another possible failure mode other than simply a crap design.
I assume the motor is mains driven.
If the power feed to the PC board is interupted while the motor is running under load the energy (1/2LI^2) stored in the airgap of the motor magnetic circuit can cause a sizable transient on the PSU input. This transient can only occur when the power feed to the board is interupted since the mains impedance is far too low to allow a transient while connected.
I have seen this failure mode on several occasions with small induction motors but assume it can happen with small series wound motors also.
The solution would be either an RC snubber across the motor or a varistor or both.

Dick

certainly a possibility yes, they are 240v AC motors

it's hard to tell as i have never witnessed one failing so have no idea of the state of the device or the supply

[Richard Head]

Dexters_lab

You can try to initiate a failure if you have a board to spare. Reproduce the suggested failure mode conditions repeatedly and see if it can cope.
All you have to do is interupt the mains supply while it's operating.

[dexters_lab]

Dexters_lab

You can try to initiate a failure if you have a board to spare. Reproduce the suggested failure mode conditions repeatedly and see if it can cope.
All you have to do is interupt the mains supply while it's operating.

i have one spare left, i might look at doing this, just because i am curious...

[dexters_lab]

ok, just sacrificed one of my spare modules to open up one of the power supplies that hasn't been blown to bits... it looks to be a slightly different design to the other one i disassembled ( typical!), looking at the date codes on the caps it seems to be an older one.

Have taken pics around the area which always shows flash damage when they blow. which is around the legs of the mosfet