Fixing an industrial Smps with a blown mosfet.

[Fflint]

I bought this 1900W 48V switch mode power supply made by ASCOM for a Cnc project few years ago. When it arrived I tested it and it used to work fine. Fast forward few years of storage in dry, but sometimes cold conditions (up to - 20, in a hermetic box) when I plugged it into my 6a fused test mains socket it blew up two out of 8 n channel mosfets it has. Unfortunately the 6a B rated fuse wasn't fast enough. When I opened the chassis I also noticed a ferrite core of what looks like an EMI suppression choke cracked in half and fell onto the pcb.

The core is E shape and uses two plastic gaps so I assumed extra tenth of a mm of glue is acceptable. I glued it together. I also replaced two blown mosfets.

I tested all mosfets and two bi-diode packs right next to them. I measured the esr of electrolytes on the high voltage side in circuit, is that useful or do I have to pull them out? The measurement came as 1100uF and 0.03ohm esr (3large caps in parallel).

This time I tried powering up the SMPS with an old style incandescent halogen light bulb in series (400W). The psu clicks to power on, the halogen light flashes, the psu switches off and the cycle repeats.

This is a 1900W psu, do you think I should try powering it up with a larger load in series? I have a 2000W heater I could use in lieu of the light bulb. However, I don't want to blow up more of those mosfets. Although not that expensive, they are quite hard to find. Or should I continue looking for more faulty components before using the bigger load in series?

The SMPS has a large control board that detects the mains voltage and switches it on for the ac mains rectifier. Then after the rectifier we have the previously mentioned choke. Unfortunately tracing signals in the back is difficult due to large capacitors obscuring the view from the top so I only found out what I could probe.

I'm having difficulty understanding how 8 mosfets and 2 bi-diode packs are all connected. I understand simple single and double mosfet SMPS schematics, but with more mosfets I'm a bit lost. Could someone point me in a direction of some example schematics how such multi mosfet PSUs are designed?

For example two mosfets are in parallel. They seem to have their sources directly connected to the minus HV SC rail. They are controlled one driver IC. The drains "go somewhere non obvious". Then I found the plus HV DC goes through those 2 bi-diode packs and then there are high voltage electrolytes in parallel. Probing around I did find the HV DC plus past the capacitors on one of the remaining mosfets, but as I mentioned the tracks on top are obscured so tracing is very difficult seeing only one side of the board.

There are some driver ICs there. Could someone suggest some method of checking them, please?

Discrete diodes and resistors look fine. I tested diodes I could get to and they are OK. I might have missed two underneath a transformer radiator.

Most SMPS repair guides I find deal with small, simple SMPSes that have a single mosfet or a pair. Is there some guide that deals with large ones, perhaps?

If there is no guide, and no one has any idea how to troubleshoot other than start desoldering components and testing them one by one, could someone with experience of using a light bulb in series comment if it is possible the SMPS could be OK and still flash a 400W halogen bulb while powering on?

Quick back of the hand calculation. It has 1100uF of caps. Assuming 340V it can store 0.374C of charge (~64J). Such charge could be charged at 340V with 1.1amps at 1ms. That's  374W for 1ms. Is it possible to see a 1ms flash? I have no idea. I'm kind of tempted to try the 2kw heater in series instead.

What do you think?

[TheMG]

The light bulb trick (or heater resistor or whatever) in series I find really doesn't work well for SMPS testing, since most have an under-voltage lockout. Even if it does it probably wouldn't save the transistors anyways as the charge in the bulk capacitors is usually enough to fry them (although perhaps less catastrophically).

My approach usually with large SMPS that have blown primary switching transistors, is to take them all out, that way I can fire up the unit and check the gate drive waveforms for proper operation without risking blowing up anything. Usually if the driver IC is sending proper gate switching waveforms, the feedback network is working, and nothing is shorted on the secondary side, it should be safe to populate new switching transistors and fire it up.

The large "choke" after the rectifier is likely the inductor for the PFC boost converter. It's job is to step up the rectified mains voltage to somewhere around 380V typically, while also approximating a sinusoidal current waveform to achieve close to unity power factor, especially important with high power SMPS.

[jonpaul]

Bonjour, been designing SMPS since 1970s, debugging a blown 2 KW supply is not trivial. Also you need the proper  safety and lab equipment.

BEWARE: 2 KW off line SMPS has severe DANGER OF SHOCK, DEATH OR DAMAGE TO EQUIPMENT

Just a few quick tips I am  jammed here today.......

1/ 240V /120V 1 phase or 3 ph input mains

2/ You should use an isolation transformer (2-5 KVA) and a variac NOT a series lamp.

3/ Assume any blown FET has also blown the drivers.

4/ Try to get or trace a schematic.

5/ 2KW is most likely a full wave bridge or possibly double forward.

6/ Run driver on separate LV lab supply with NO HV bus to check drivers.

Bon Chance and BE SAFE!

Jon

[Fflint]

Thank you both for really good tips.

Jon, I do have a variac and an isolation transformer, but I wasn't sure how to use them to protect it from frying more mosfets. I assumed if there was a short I needed current limiting. That's why the light bulb is used. Is there a better way?

You mentioned running the drivers off a separate low voltage psu to test. I'll definitely do it.

I have the usual test equipment such as an oscilloscope(probes allegedly good up to 600V), bench psus up to 60V etc.

Further answers and comments online below.

Bonjour, been designing SMPS since 1970s, debugging a blown 2 KW supply is not trivial. Also you need the proper  safety and lab equipment.

BEWARE: 2 KW off line SMPS has severe DANGER OF SHOCK, DEATH OR DAMAGE TO EQUIPMENT

Just a few quick tips I am  jammed here today.......

They are very useful, thank you. Regarding safety I have a proper bench I'm testing it on. So it sits there in a stable way (no risk I'll grab it instinctively if it trips etc). I have a switchable power extension before the isolation transformer. I use it to power on/power off the AC. No live wires lying around being connected together by hand... Other than that, my floor is made of wood and I do have an RCD in the power panel so I'm hoping that would help. Also I always discharge the HV caps before working on it after powering it on. Unfortunately that relies on remembering to do it, which is not great, but I don't know of any better solution.

1/ 240V /120V 1 phase or 3 ph input mains

1 phase 240v. The manufacturer claims the smps will work from 180v to 240v. Here is a manufacturer page if you're interested to see the exact model. http://web.tiscali.it/gnnsites/rdz_SMPS48V1900.htm (only in Italian)

2/ You should use an isolation transformer (2-5 KVA) and a variac NOT a series lamp.

I mentioned it above. I wasn't sure low voltage would be enough to protect mosfets in case of a short so I wanted some way of current limiting. If possible, please let me know if there is a better way while still having current limit capability

3/ Assume any blown FET has also blown the drivers.

OK. I'll do that. Two blown mosfets used the same driver. I'll get a replacement.
Edit: not so fast. I thought the mosfet were hard to find.. This driver IC is very hard to find. Only one distributor located half a world away has it. I guess I'll have make sure 100% it is faulty before I spend half of the value of the Smps on shipping a part for it.

4/ Try to get or trace a schematic.

There is no schematic available online. I searched in the usual way. So I'm trying to trace it myself. If I have to I might remove those big caps to get to the top of the board.

5/ 2KW is most likely a full wave bridge or possibly double forward.

6/ Run driver on separate LV lab supply with NO HV bus to check drivers.

Bon Chance and BE SAFE!

Jon

Will do :-) Thanks

[Fflint]

I've pulled the driver chip out and I tested it on its own by applying power and pulling inputs up or down. It is fine. I also pulled out a bunch of other components surrounding the driver chip. All tested fine (3 caps and a 2 diodes) . I've reassembled everything and I tried using the heater for current limiting. The heater is selectable 1kw,2kw or 3.5kw. With 1kw the smps clicks repeatedly same as with the 400W light bulb. With 2kw my 6amp circuit breaker triggers. I haven't tried on a bigger breaker circuit. Unfortunately I don't remember if when this smps was originally tested it was on a 6 or 16amp circuit. Is it possible it can be fine and still trigger the 6amp fuse on startup?

I think that's unlikely.

Before I start pulling all mosfets out, is there something else to try? I attach few photos.

Connection wise there are 6 mosfets, not 8. 3 are on the HV minus and 3 are on HV plus. After more testing I found out only one out of 3 in the minus circuit was blow (2 were replaced, but one tests fine on the bench). The minus mosfets are on the right, plus on the left.

The transformer and low voltage rectifier are fine.

The first two photos show how logic boards are connected. The smps is missing a number of clamps holding mosfets to the radiator and some screws. This is just for testing.

[jonpaul]

Rebonjour: The PFC is a boost and will give zero OR 360VDC at bus. NO current limit will do any good. Variac is just a habit. NO current limiting.

Isolation Tr is ONLY to allow safe probing  with the scope ref to bus LO side! NO RELATION TO CUR LIM

MOSFETS will die if the drive or gate is exceed for microseconds. NO "current limit" or fuse will ever protect.

We debug with LVPS on the drivers, DISCONNECT mains and 360 Bus is cut from boot PFC and we run a separate isolated variable PS to give 0-360V at the bridge bus.

Suggest to contact mfg, say you are a customer and badly need ser vice manual and schema.

VERY hard to tshoot w/no doc.

Bon chance,

Jon

[Tomorokoshi]

I notice some dust on the circuit boards. At those voltages watch out for conduction through parasitic paths that could interfere with the control and feedback circuits. For instance, high-voltage bus leakage into a feedback circuit. Perhaps clean off the circuit boards.

[Fflint]

It seems I'm quite lucky, because the Smps is working :-)

I pulled all caps (both high and low voltage). I tested them for esr and capacity. They were all OK. I run the smps same as I did before planning to probe with a scope while it is clicking starting from the output side. I noticed a small ramp of voltage. 2v for 20ms, then 4v for 10ms and then it shuts down. Waits half a second and repeats. So I decided to try it with a heater set to 3kw on a 16 amp circuit. I initially set the variac to 180v with no joy. However, once I gave it full 240v with the heater it clicked twice and on the third one started fine.

Great stuff, so it was just a cracked ferrite core and 1 blown mosfet. Now I hope this wasn't a fluke and once I reassemble it it continues to work.

I found it quite difficult to solder huge caps back in while they had to reliability solder on top and the bottom of the pcb. To do it I mixed some smd soldering paste with paste flux and I smeared it on top of the pads hoping the heat will melt it making a bond on top as well. I hope this connection will prove reliable enough.

EditThe joy was short lived. I washed the pcb in IPA and assembled it in its chassis. On startup it blew the little 1.25amp fuse for the logic... That's unfortunate. I'll have to investigate further.

[Fflint]

So, it turns out another mosfet broke. I'm not sure if it was because my IPA I used to clean it was contaminated with water (it measured 200k between two electrodes submerged in it half inch away). Or did I have a short from one of mosfet backs to the radiator, or elsewhere.

Long story short, I replaced the faulty mosfet. I desoldered all the caps again and I cleaned any solder paste residue very carefully. I dried the board with a heat gun. After resoldering the caps the smps started fine with a 3kw heater in series.

I was worried about plugging it into the wall directly suspecting that perhaps my super glue repair of the ferrite core could have something to do with the second blown mosfet when it was plugged into a high current AC circuit. So I lowered the series load resistance by adding another 3kW heater in parallel with the first. It started fine. So plugged it directly into my 6amp ac circuit. It works fine without the chassis.

I definitely don't want to assemble it all together, plug it in and see white smoke again. So tomorrow I'll try it on a 16amp circuit after it rested all night. I hope nothing will blow up.

Only then I'll install it in the chassis. I may get some new silicone heat transfer pads for between the mosfet and the radiator just in case.

[Fflint]

This is really puzzling. I tested the smps on the bench (out of the chassis) few times. It worked perfectly. After I put it in the chassis it blew a fuse again. I was very careful. I measured that no mosfet has a short to the radiator. No screw holes are shifted etc... I have no idea why it works fine without the chassis and blows up when in it.

Additionally when I ramped up the voltage slowly it seemed fine. Only when it was plugged into ac directly the fuse blew. Any suggestions?

[TheMG]

Maybe there's a cracked solder joint somewhere, and when the PCB is flexed a certain way it works, but when you screw it back in to the enclosure that solder connection goes open and it blows up.

As for the blowing fuse are you using the type and rating of fuse recommended by the manufacturer? If yes, does the power supply have an inrush current limiting circuit? Often this will be one or more NTC thermistors in the AC path, or large resistors, which get bypassed by a relay after a short time delay. If so equipped, is the inrush limiting working correctly? Relay contacts could have welded closed if it was switched into a short circuit fault. I didnt run the numbers myself but the math doesn't sound right on the inrush number. For a power supply like this the inrush would be quite considerable actually, especially with no inrush limiting circuit or one that is not functioning.

[Fflint]

Maybe there's a cracked solder joint somewhere, and when the PCB is flexed a certain way it works, but when you screw it back in to the enclosure that solder connection goes open and it blows up.

As for the blowing fuse are you using the type and rating of fuse recommended by the manufacturer? If yes, does the power supply have an inrush current limiting circuit? Often this will be one or more NTC thermistors in the AC path, or large resistors, which get bypassed by a relay after a short time delay. If so equipped, is the inrush limiting working correctly? Relay contacts could have welded closed if it was switched into a short circuit fault. I didnt run the numbers myself but the math doesn't sound right on the inrush number. For a power supply like this the inrush would be quite considerable actually, especially with no inrush limiting circuit or one that is not functioning.

The inrush numbers are certainly wrong, I "assumed" a time constant of 1ms and another person already said I used wrong voltage (340v used, 380v is correct) . You are correct, there is a small glass fuse rated for 1.5A, then this gets bypassed once the control board starts. I measured up to ~250-300mA of current with the control board detached. I didn't power it long enough to observe it going down.

I did suspect the relay at one point (after the first time the fuse blew following assembly). It sure did look like it was stuck shut, but it turned out one of the mosfets had a short.

However, I now believe I found the true first cause of the failure. It is deterioration of the thermal silicone pad between mosfets and the radiator. I show the arrangement below in a photo. Mosfets are well clamped so it must be squeezed pretty thin. I had some smaller silicone pads for to-220 transistors so I put two behind each mosfet as an experiment. I then clamped them the same and I reassembled everything. It works now.

I already ordered a large sheet of silicone to cut an entire new pad. However, if the first one failed, what's stopping the next one from following in its steps? Would more knowledgable people suggest mica instead?



The grey pads (too small I know) are just temporary until I get something better. I'm not planning to actually use the smps until they are replaced.

[TheMG]

There was likely a burr or metal shaving which punctured the silpad, it only takes the tiniest little piece of metal.