Lightly Toasted MOV in SMPS

[SmokeDetector]

Fixing up a Carvin EM900 in-ear monitor for a friend.

The only visible issue is an MOV that has burned up a little, which seems odd, as I thought they failed open-circuit? This one is showing about 28K ohms across it.

The MOV looks to be part of a flyback/snubber - it's in series with a fast diode across the primary coil of the transformer, which I assume is to increase efficiency by not clamping until the voltage goes above the 170V DC rating of this MOV, but it also makes me think the MOV is going to have a limited life if it's sitting there clamping all those spikes? Clearly I'm missing something.

I'm going to replace the MOV as a first step, but am wondering if there's some other issue that could cause this component to fail this way?

[BradC]

No they generally gradually fail short. They then open up by burning out or a massive surge.

[SmokeDetector]

Ahh,that makes more sense.

I checked the fast recovery diode in series with the MOV and that looks to have failed, so it's conducting all the time, exposing the MOV to the full DC rail voltage (300V+) whenever the MOSFET turns on, leading to its failure.

Will replace diode and MOV and see where that gets me.

Russell.

[Armadillo]

If only the flyback snubber components failed, then you would have not sensed a problem with the power supply failure unless the mosfet went with it because of larger voltage peak impressed onto it.

Hence better turn on using light bulb in series as caution.

[T3sl4co1l]

I would think the MOSFET has failed as well, and controller, and...

It may've saved itself by switching very short pulses, but it needs to be checked.  That sort of thing can be a cascading failure.

MOVs aren't a good design idea in that application.  Normally a P6KE200 or so is used (avalanche diodes / TVS are good for ~infinite pulses, as long as ratings are observed).  Alternately, an R || C network, which charges up to some peak voltage and burns the excess power in the resistor (which can be much larger than a diode or MOV: 5W or more).

Maybe they needed more power dissipation?  Also a result of bad design...

Tim

[SmokeDetector]

My half-baked theory (a bit like the MOV) is that once the diode started conducting all the time, the MOV pulled down the voltage across the primary of the transformer, dropping its output enough to stop the unit working, my friend noticed this, then smelled something burning, then turned the unit off.

I've attached an attempt at a schematic showing what I think are the relevant bits, including the component values. Also pics of the board in question. Toasted MOV is the charred black thing tucked up against the transformer, D1 is the diode.

I pulled out the MOSFET and did a basic test with the diode setting on my meter, and it seems to be behaving properly. The 1A fuse on the mains input also hadn't blown, so I'm feeling hopeful, but there's also the control bits, which are SMDs under the board. I will take the advice above and sort out a light bulb before I fire it up (hopefully not literally)...

Russell.

[Armadillo]

Is the labelling a "C7" ?

[SmokeDetector]

Is the labelling a "C7" ?

Yep, that's the one.

[Armadillo]

Is the labelling a "C7" ?

Yep, that's the one.

I meant could that be a capacitor instead of "MOV".

[T3sl4co1l]

Oh, and there is, in fact, a resistor R2 right next to it (under shrink wrap, as is often done for whatever reason).  So it IS actually an RCD clamp, but they put an MOV in there? !?

Tim

[Armadillo]

Oh, and there is, in fact, a resistor R2 right next to it (under shrink wrap, as is often done for whatever reason).  So it IS actually an RCD clamp, but they put an MOV in there? !?

Tim

There should instead put a Zener diode, all this is for code compliance. Put a capacitor to ring back the energy for saving instead of compliance? LOL.   

[T3sl4co1l]

There should instead put a Zener diode, all this is for code compliance. Put a capacitor to ring back the energy for saving instead of compliance? LOL.   

Code doesn't care about peak voltage snubbers..?

No ringing, the R||C is behind the clamp diode.

Tim

[Armadillo]

There should instead put a Zener diode, all this is for code compliance. Put a capacitor to ring back the energy for saving instead of compliance? LOL.   

Code doesn't care about peak voltage snubbers..?

No ringing, the R||C is behind the clamp diode.

Tim

How about after design "EMI"?
That mod, I think is a leakage induction transformer.

[carl_lab]

Is the labelling a "C7" ?

Most likely it's a ceramic disk capacitor not a MOV.

[SmokeDetector]

Is the labelling a "C7" ?

Yep, that's the one.

I meant could that be a capacitor instead of "MOV".

Well, it's blue where it isn't discoloured from the heat, it has the distinctive shape of a MOV, not a ceramic cap, and the part number on the package (as per my schematic) is for a MOV with a 170V DC rating.

[Armadillo]

Well, it's blue where it isn't discoloured from the heat, it has the distinctive shape of a MOV, not a ceramic cap, and the part number on the package (as per my schematic) is for a MOV with a 170V DC rating.

Yes, they could have installed a MOV instead, and your identification is not wrong.
Like I said, likely for code compliance because the transformer is quite leaky.
I would not challenge the manufacturer, the have all the tests done.

[SmokeDetector]

Oh, and there is, in fact, a resistor R2 right next to it (under shrink wrap, as is often done for whatever reason).  So it IS actually an RCD clamp, but they put an MOV in there? !?

That shrouded resistor is connected to the positive DC bus, but doesn't connect to anything on the high side of the MOSFET, rather the trace runs down under the transformer to a diode and electrolytic cap, and looks to be feeding VCC on the controller chip (a UC3843).  However, I'll have a closer look tomorrow with the datasheet for that chip and see  if I can confirm what it's doing.

[T3sl4co1l]

Oh, so that's the startup resistor?  Guh...

Tim

[schmitt trigger]

The MOV looks to be part of a flyback/snubber - it's in series with a fast diode across the primary coil of the transformer, which I assume is to increase efficiency by not clamping until the voltage goes above the 170V DC rating of this MOV, but it also makes me think the MOV is going to have a limited life if it's sitting there clamping all those spikes? Clearly I'm missing something.

I fully agree with your assessment; I also believe this is a snubber.

I remember that Power Integrations had app notes where they showed P6KE-series Tranzorbs for the exact same purpose. The company I worked for in the 1990s copied that idea, which went into a high volume product.
The Tranzorb became the highest field failure for this product.

An MOV would be an ever worse idea.

[SmokeDetector]

Just to round this out, finally ordered parts for this - as this was a free repair for a friend, I was waiting until I had enough stuff to order to get free shipping 

Replaced the fast diode and MOV and she works, 15 or so volts on each rail. One clue as to why this thing has a limited life is that the power supply is always on, the "power"button just varies the consumption between 6W and 1W, so I'm going to suggest to my friend that he unplugs the mains when he's not using it. That said, if and when it fails again, it's only $3 worth of parts...

[schmitt trigger]

I hope he invited you to have some cold beers for your effort....

[SmokeDetector]

I hope he invited you to have some cold beers for your effort....

That's the tricky thing - I think once people know you can fix stuff, they just assume it's some effortless thing that you just do, and consumes no time at all.

I probably should specify a beer-based schedule of fees 

[BradC]

That's the tricky thing - I think once people know you can fix stuff, they just assume it's some effortless thing that you just do, and consumes no time at all.

I had a couple of friends like that. I solved the problem by insisting they sit and "help". After several hours they got the message. Complex devices for repair are often accompanied by a good bottle of Scotch or some nice craft beer. Problem solved. Most of my other mates already knew so it was never an issue.

Relatives are the worst for that though.