buck converter repair

[Pinus94]

Hi, everyone. Ten days ago, the board of an old metal bender failed, it refused to turn on. I extracted the pcb and started checking components. I founded a p-mosfter and a tvs diode shorted, replace it and the machine started again. The repair lasted 10 days, (each day the machine has been on for 8 hours. Everything seemed to be going well); until Today, the machine failed to start, again the same components shorted.

Let’s me try to explain the schematic and the problem.

The section to be repaired is the power supply section, it is a buck converter, it converts 24vdc to 5vdc and 12 vdc. The busk converter is based on the LTC1267CG-ADJ5, the shorted components are

1) the p-mosfet (originally it was 10P03L, replaced by me with AOD417)
2) a TVS diode marked “DX” from ST Component (placed between the 12vdc output and ground, it is a tvs diode with 14,5v threshold)

Output voltages seem to be perfect; output capacitor seems to be ok (checked only the capacity). No other components seem to be shorted or bad.

What can I check? What can cause these two components to fail?

For now, I re-replaced these two components, I put a 24vdc power supply exclusively to power this board, added a proper fuse (before, a wrong “size” fuse was used)

I hope I explained it well and sorry for my bad English. About my electronic knowledge, I’m an automation engineer, for hobby I like to repair electronic boards.
Hope I can receive some useful tips

[madires]

Is the shorted p-ch MOSFET also for the 12V output? Could you please provide some details about the 12V load?

[Pinus94]

yes, the p-mosfet is for the 12v output, looking at the typical application schematic (first page of the datasheet) it is in place of P-CH Si9435DY

the load is a mix of memories, microcontroller and other stuff. The board has a serial communication with a "cn main board" and a video output.

[PA4TIM]

Did you buy the parts direct from China (ebay, amazon, Aliexpress, temu) If so, there is a huge chance they are not what they are labelled.

If they are 100% sure original then there is something other wrong that kills the parts. It can be very difficult to find things if they are not constant wrong.

Are fans working like they should ? 
Is the PCB clean ? I had to repair a scope that was used by service technicians who serviced all kinds of metal transforming machines (I do not know the correct English term, things like drilling, turning, cutting, milling)
When I opened the scope there was a lot of greasy/sticky dirt inside and in the there where several flakes of metal stuck to the dirt. One of those had caused a short.
Also had a bow thruster controller that had a conformal coating that was so dirty it was conducting causing all kinds of (destructive) problems.

[Pinus94]

I bought replacement parts from "tme" (Poland online shop) they are genuine.
The fan is working and the pcb is very clean, no dust, no metal shavings. Neither the mosfet nor the diode are warm to the touch.

[PA4TIM]

The original MOSFET is a logic level version. The one you used is not. That could be the problem but that depends of the amount of gate voltage and it's function. But that has no relation to the tVS.

A TVS shorts if the voltage over it is to high. You could log the voltage over it with a scope using one shot triggering at the hight of the max voltage of the TVS. But all that time you can not use the scope for other repairs. But first I would monitor the voltage over the TVS with the scope in normal upgoing flank trigger to see if there are many spikes on the rail it protects.

How good is your electronic knowledge and how sure you are the P-ch mosfet was really dead ? How do you test the mosfet ?
You can not test that in situ and if you charge the gate by accident during your tests the channel will test as a short.

If it is only the TVS that is bad, then there is something that is causing high voltage spikes/levels over the TVS.  Did you use exactly the same TVS or like the mosfet, one you think is the same ?
 

[Pinus94]

the originale mosfet has a Vgs of - 1.0 - 2.5 (min max), mine -1 -3, the one used in the application notes of the ic (i.e  Si9435DY) –1  –3
so from my "electronic noobness" i'm pretty confident the aod417 is fine for this application.

Mosfets, both the original one and the replaced one, were completely shorted between all tree pin.

I replaced the TVS with the exact original model.

I have a 20 mhz oscillsocope + a battery one (about 1Mhz), but for now is difficult to take measurement on the board, it is installed and the machine is working right now.

I asked here and i'm studying to understand what can cause this problem, in case it will recur in the next weeks.

for sure "there is something that is causing high voltage spikes/levels over the TVS"

[PA4TIM]

This is not how it works. -Vgs(th) is the voltage to start the conduction. Not max/full conduction. Often -Vgs(th) is specified for only 100uA or 1mA.
A logic level mosfet often needs less then -5Vgs to conduct fully. A normal mosfet needs a lot more than that. But again, it depends on things only you know, the -Vgs it really sees. Measure that, then look in the graph of the datasheet of your mosfet and the original and you know the max current it can transfer at the measured -Vgs. The datasheet from the controller also uses a logic level mosfet. Your mosfet replacement is not logic level.

Is the TVS protecting the input or at the output protecting the load? If input then the incident voltage comes from the external power source. If on the output the smps itself has a problem (maybe a dead snubber or bad filtering) Based on your limited info I can not tell more.

But again, I do not see a relation between the TVS and mosfet. Or it is the mosfet that dies and then the output goes high and kills the TVS (If it is at the output) In that case it is simple, you used the wrong mosfet.

20 MHz sounds like an analog scope or a very low budget toy-scope. 1MHz for sure will be a toy scope. (or an old scope meter) both will probably not support  a usable single shot triggering and can be to slow for transient detection.

[Pinus94]

thanks for the usefully information,
Comparing the 3 mosfet datasheet (application note, original, replaced) the difference between logic level and not logic level mosfet is clear

the VTS is protection the output
"it is the mosfet that dies and then the output goes high and kills the TVS (If it is at the output) In that case it is simple, you used the wrong mosfet."

yes, the oscilloscope, is an old hitachi analog model

the circuit is drawing 200mA from the 24vdc in input

[Pinus94]

the machine worked properly for 10 days, today it didn't turn on, the fuse blew, replaced it and start working again