SMPS eats mosfets

[nashkato]

I´m trying to repair this powersupply .
2 sections , one for digital supply the other for analog .
it is supplied at the input with app. 60V DC .

i´m at it for quiet some time now with different symptoms .
right now one of the pair of mosfets shorts out after 2 seconds .

am i doing something stupid when i assume that it is possible to test only one side (digital or analog) and remove the mosfets from the other ?

here´s the schematic

 

[Bo Deadly]

What tests have you performed?

So the two mosfets of one side were burned out?

And then you took them from the other side and they burned out too?

Explain the exact sequence of events of what you did and what the result was.

Note that if you replaced mosfets, did you mount them properly? Where they mounted on pads or with compound or what?

Also note that if two mosfets are burned out, it seems to me unlikely that both mosfets simultaneously died. It is much more likely that something is sorted like It could be that one mosfet failed short and cause the other then fail but it could also be a diode burned out or the controller is sending the wrong signal like it's not switching right or the transformer is shorted.

IMO debugging SMPS is almost certainly not worth the time because you can almost always find a replacement that is smaller, better and more efficient for very low $$. What are the voltage and current of each output?

 

[JohnRoberts]

Without digging too deeply into the schematic. MOSFETs failing may be from both being turned on at the same time. There is probably digital logic in the controller board to prevent that.

With one or both MOSFETs removed can you look at the gate drive signals with a scope to confirm there is no overlap or simultaneous conduction.

JR

 

[nashkato]

With one or both MOSFETs removed can you look at the gate drive signals with a scope to confirm there is no overlap or simultaneous conduction.

thanks ! i will do so and report back.
But ... aren´t those mosfets parallel anyway ?

What tests have you performed?

So the two mosfets of one side were burned out?

non other than poking around with a DMM
no ! T1 respectively T4 burned out .

And then you took them from the other side and they burned out too?

both sides share the same 60VDC supply and there´s no possibilty to separate it from either side.
the controllers are on separate sub-pcb´s and removeable
(if i remove one of them, of course the supply current is only limited through the transformers DCR , please correct me if i´m way off here )
that´s why i removed one side of mosfet-pairs to trying to narrow down the problem to any of the 2 sides.

Note that if you replaced mosfets, did you mount them properly? Where they mounted on pads or with compound or what?

they´re insulated have been and don´t have no connection to any part of the heatsink or pcb other than their supposed connections .

Also note that if two mosfets are burned out, it seems to me unlikely that both mosfets simultaneously died. It is much more likely that something is sorted like a diode burned out or the controller is sending the wrong signal like it's not switching right or the transformer is shorted.

yes , that´s what i was thinking , too
i can more or less eliminate any shorted parts (except the controller pcb)
i changed the controller
if the transformer(s) really are the problem it would be A) quiet a time consuming thing to unsolder them with the doublesided pcb .
( can i scope them with low signal ( don´t know 1kHz) inside the pcb ? )
and B) most likely a total loss .

IMO debugging SMPS is almost certainly not worth the time because you can almost always find a replacement that is smaller, better and more efficient for very low $$. What are the voltage and current of each output?

well , the voltages are on the schematics ( 15, -15 , 48... 15 and 5 for digital ) i´d need at least 5A on each the 15/-15V analog supply , any suggestions ? very much appreciated , i had a look around , but couldn´t find a proper fit .

 

[gswan]

Could be a voltage spike G-S (Vgs) which are very sensitive to overvoltage.
Usually you would find zener clamps on the GS pins right next to the FET to prevent this from happening.

 

[nashkato]

at the gates of the mosfets ( unsoldered) i see a pulsating (app. 1sec) chainsaw with 50kHz which corresponds to controller datasheet of UC3844 that says 52kHz .
the 1sec pulse is the startup i guess ?
oh yes ... the cainsaw is 15Vpp .

 

[gswan]

It's the sub-microsecond switch-on and switch-off transients you need to watch for. You need to make sure your probe is calibrated and then look carefully at the leading and trailing edges to see how high those transients are.

 

[JohnRoberts]

at the gates of the mosfets ( unsoldered) i see a pulsating (app. 1sec) chainsaw with 50kHz which corresponds to controller datasheet of UC3844 that says 52kHz .
the 1sec pulse is the startup i guess ?
oh yes ... the cainsaw is 15Vpp .

can you look at both gate drives at same time to confirm no gate overlap, that would cause mutual conduction.

JR

 

[nashkato]

can you look at both gate drives at same time to confirm no gate overlap, that would cause mutual conduction.

the mosfets are wired parallel and there‘s only 1 gate drive from the controller . the analog and digital section each are very similar and do have each the same controller sub-board .
i dont really understand which to gate drives i shall compare ?

however , yesterday i exchanged the 2 irf640 for one irfp250 mosfet .
( there‘s another version / revision of this PS with irfp250 , that‘s why i tried)
the ps now is pulsating in startup i guess and the secondary 15V that goes back to the controller isnt stable .

do i need to look closer at the secondaries ?

 

[nashkato]

It's the sub-microsecond switch-on and switch-off transients you need to watch for. You need to make sure your probe is calibrated and then look carefully at the leading and trailing edges to see how high those transients are.

there‘s a small spike at turn on and the opposite at turn off , but as long as my oscilloscope is able to show these short peaks in it‘s full capacity it‘s something in the area of 2-3 volts

 

[JohnRoberts]

the mosfets are wired parallel and there‘s only 1 gate drive from the controller . the analog and digital section each are very similar and do have each the same controller sub-board .
i dont really understand which to gate drives i shall compare ?

however , yesterday i exchanged the 2 irf640 for one irfp250 mosfet .
( there‘s another version / revision of this PS with irfp250 , that‘s why i tried)
the ps now is pulsating in startup i guess and the secondary 15V that goes back to the controller isnt stable .

do i need to look closer at the secondaries ?

My mistake, I ASSumed there was push/pull so harm could be caused by both turned on at same time.

If the multiple supplies are fed by same master switching wave form, but behaving differently the problem is local to the individual supplies.

Sorry about the misdirection.

JR

 

[Bo Deadly]

When transistors are subject to a lot of power over a long time, they can start to leak current and generally not turn on / off so fast. That could be catastrophic for an SMPS.

I'm curious be what is the device that contains this supply?

 

[nashkato]

fed by same master switching wave form

that they´re not either . Except the common DC source , both share same almost the same design but are more or less independent

 

[JohnRoberts]

I am clearly not paying enough attention... feel free to disregard.

JR

 

[gswan]

there‘s a small spike at turn on and the opposite at turn off , but as long as my oscilloscope is able to show these short peaks in it‘s full capacity it‘s something in the area of 2-3 volts

Is your scope probe calibrated? Can your scope show very narrow spikes?
If the mosfet Vgs goes outside +/-20V then the mosfet usually fails. This is why zener diodes are usually attached to GS to clamp those spikes.