Mosfet Keeps blowing

[dark_hawk]

Hi,

In this circuit, which is a PSU board for an audio amplifier, Q3 IRF740 (The one highlighted) keeps blowing up (short between the three legs)
Sometimes this short happens right away after the relay click, and sometimes it takes a few minutes. The first time it blew it took like 2 hours of operation before it blew up.
The four mosefts and the TOP243 gets hot in standby, about 50-70 C. Which is as reported by others normal for this board.

I checked everything around this mosfet and the individual components seems fine.
What can cause such failure?

AUX power:


Main PSU1:


Main PSU2:


Back of board:


Front of board:

[perieanuo]

Hi,

In this circuit, which is a PSU board for an audio amplifier, Q3 IRF740 (The one highlighted) keeps blowing up (short between the three legs)
Sometimes this short happens right away after the relay click, and sometimes it takes a few minutes. The first time it blew it took like 2 hours of operation before it blew up.
The four mosefts and the TOP243 gets hot in standby, about 50-70 C. Which is as reported by others normal for this board.

I checked everything around this mosfet and the individual components seems fine.
What can cause such failure?

hi,
i suspect something in the 'command' loop for Q3, like C57, C29, C54 maybe some diode like D23 (is she ok?if it's opened, you may have some spurious spikes there killing G-S junction), maybe bad diac (but i saw this rarely)
(you can just replace those caps tocheck)
top243 seems independent supply, so nothing connects this to your problem
voltage on C59 and C60?

[sean0118]

It's difficult to see in the photos, but aren't the mod wires for the gate and drain touching the trace connected to the source? Is there more than the soldermask providing insulation?

Another option, the replacement  IRF740s could be fakes?

[dark_hawk]

Hi,

In this circuit, which is a PSU board for an audio amplifier, Q3 IRF740 (The one highlighted) keeps blowing up (short between the three legs)
Sometimes this short happens right away after the relay click, and sometimes it takes a few minutes. The first time it blew it took like 2 hours of operation before it blew up.
The four mosefts and the TOP243 gets hot in standby, about 50-70 C. Which is as reported by others normal for this board.

I checked everything around this mosfet and the individual components seems fine.
What can cause such failure?

hi,
i suspect something in the 'command' loop for Q3, like C57, C29, C54 maybe some diode like D23 (is she ok?if it's opened, you may have some spurious spikes there killing G-S junction), maybe bad diac (but i saw this rarely)
(you can just replace those caps tocheck)
top243 seems independent supply, so nothing connects this to your problem
voltage on C59 and C60?

Thanks for the reply.

C57, C29 and C54 are all fine, desoldered and measured and they measure fine.
D23 is also fine. The two Diacs are showing no readings on a 9V multimeter, so I assume they are fine.
TOP243 is indeed independent. And the standby voltages are fine.
Voltages across the Big caps are 140-150V if I remember correctly. I didn't measure them since replacing the old CAPXCOM junk with some nichicons.

[dark_hawk]

It's difficult to see in the photos, but aren't the mod wires for the gate and drain touching the trace connected to the source? Is there more than the soldermask providing insulation?

Another option, the replacement  IRF740s could be fakes?

The solder mask seems fine, I will added an extra layer of the UV curable solder mask though, won't hurt.

The original IRF740 did blow up though, there is a problem somewhere.
The replacement may be fakes, yes. But even if they were fakes, would it blow up without any load connected?

[DavidAlfa]

Don't connect it directly to the mains! Put a 50-100W  bulb in series!
If the transistor is permanently switched on, it will bright like a normal bulb, but keep it from blowing.
Normally it will be barely lit, red-orange.
Check the reverse voltage on zener diode D23 (ensure it doesn't go over 15V), also the snubber C29/R36, it filters the huge switching spikes, the mosfet will die quickly if exposed to such voltages.

[sean0118]

The solder mask seems fine, I will added an extra layer of the UV curable solder mask though, won't hurt.

The reason why I mention it is I shorted out an audio amp once because there was a single wire strand touching a nearby trace. I couldn't see any holes in the solder mask but it was still shorted, that was only at a few volts as well. This is quite high voltage so I think you'd be better striping the wire insulation off right where it goes into the through hole.

The original IRF740 did blow up though, there is a problem somewhere.
The replacement may be fakes, yes. But even if they were fakes, would it blow up without any load connected?

It's unlikely but possible. Some of the fakes switch slower than the genuine parts so they are still conducting when they should be off etc. Just something to keep in mind, there most likely is another fault as you mention.


What are the repairs you have already done on the board?

[dark_hawk]

Don't connect it directly to the mains! Put a 50-100W  bulb in series!
If the transistor is permanently switched on, it will bright like a normal bulb, but keep it from blowing.
Normally it will be barely lit, red-orange.
Check the reverse voltage on zener diode D23 (ensure it doesn't go over 15V), also the snubber C29/R36, it filters the huge switching spikes, the mosfet will die quickly if exposed to such voltages.

Hi David, Thanks for the reply.

I'm already using a 200W incandescent lamp in series. The mosfet blows but at least it saves the fuse and the rest of the circuit.

C29 and R36 are reading fine. Also test D23 out of circuit and it's working.

[dark_hawk]

The solder mask seems fine, I will added an extra layer of the UV curable solder mask though, won't hurt.

The reason why I mention it is I shorted out an audio amp once because there was a single wire strand touching a nearby trace. I couldn't see any holes in the solder mask but it was still shorted, that was only at a few volts as well. This is quite high voltage so I think you'd be better striping the wire insulation off right where it goes into the through hole.

The original IRF740 did blow up though, there is a problem somewhere.
The replacement may be fakes, yes. But even if they were fakes, would it blow up without any load connected?

It's unlikely but possible. Some of the fakes switch slower than the genuine parts so they are still conducting when they should be off etc. Just something to keep in mind, there most likely is another fault as you mention.


What are the repairs you have already done on the board?

I changed all the electrolytic caps, replaced D9, and done some changes in the AUX part of the circuit. Nothing in the primary side.

Something new happened that I don't know if it relates to this issue. C70 bulged. It's a new Nichicon GU series cap, It just won't bulge on it's own. And it's supposed to be on Q1 and Q2!

Also something I don't understand, D22, D23, D24 and D25, the zener diodes at the gate of the mosfets, the read voltage drop in both directions while in circuit, but only in one direction out of circuit. The 4 diodes can't all have gone bad! And I can't understand why they are reading in the opposite direction in circuit.

[wraper]

It's difficult to see in the photos, but aren't the mod wires for the gate and drain touching the trace connected to the source? Is there more than the soldermask providing insulation?

Another option, the replacement  IRF740s could be fakes?

The solder mask seems fine, I will added an extra layer of the UV curable solder mask though, won't hurt.

The original IRF740 did blow up though, there is a problem somewhere.
The replacement may be fakes, yes. But even if they were fakes, would it blow up without any load connected?

Totally wrong assumption. Original MOSFET may easily fail independent from other parts. And if you use crappy counterfeit as replacements you will have further failures.

there is a problem somewhere.

As if the MOSFET as one of the most stressed components must not fail first, and there must be an issue somewhere.

[dark_hawk]

It's difficult to see in the photos, but aren't the mod wires for the gate and drain touching the trace connected to the source? Is there more than the soldermask providing insulation?

Another option, the replacement  IRF740s could be fakes?

The solder mask seems fine, I will added an extra layer of the UV curable solder mask though, won't hurt.

The original IRF740 did blow up though, there is a problem somewhere.
The replacement may be fakes, yes. But even if they were fakes, would it blow up without any load connected?

Totally wrong assumption. Original MOSFET may easily fail independent from other parts. And if you use crappy counterfeit as replacements you will have further failures.

there is a problem somewhere.

As if the MOSFET as one of the most stressed components must not fail first, and there must be an issue somewhere.

Thanks for the reply.

I already have 5 failed mosfets in 5 tries. One of which is an original part since it was disassembled from a faulty HP board ( of a printer I think). I do have a problem somewhere.

[sean0118]

Something new happened that I don't know if it relates to this issue. C70 bulged. It's a new Nichicon GU series cap, It just won't bulge on it's own. And it's supposed to be on Q1 and Q2!

That's interesting that it can fail that quick, from the photos it looks like they are installed the right way. I can't see how their voltage rating could be exceeded with the 230V jumper set? Maybe ripple current is too high? Otherwise counterfeit capacitors are very common too unfortunately.

Also something I don't understand, D22, D23, D24 and D25, the zener diodes at the gate of the mosfets, the read voltage drop in both directions while in circuit, but only in one direction out of circuit. The 4 diodes can't all have gone bad! And I can't understand why they are reading in the opposite direction in circuit.

I'm not sure either, you might be measuring the volt drop over the resistors when in circuit? If they measure okay out of circuit they should be fine.

[sean0118]

This is a tough one, it doesn't look that complicated, but I still can't see a clear reason for the MOSFET to blow or C70 to bulge. Is R23 the correct value? Looks like it might have got hot at some point?

[dark_hawk]

Yes. Both R23 and R2 are reading fine out of circuit.
Turns out the mosfets I have are authentic. I measured Drain-source on-state resistance using this method:

Those vishay branded mosfets that I have are reading 0.43 ohm which is even better than the 0.55 ohm on the datasheet, and the original which reads around 0.5. Compared to some other mosfets I have from another source which reads 1.1 ohm.

This board had a lot of that conductive glue which sticks very bad to the board, hence the lifted pads and missing connections and the jumper wires. My guess now is that the problem might be there is a short somewhere on the circuit or a trace that's not conducting, or reads fine when measured in continuity and fails when high voltage is applied . Hunting it will be tedious, but I think I don't have a choice.

[dark_hawk]

A question: Q2 and Q3 are on the same transformer. Should both mosfets be identical? Same manufacturer and properties.
Or is it acceptable that they have slightly different RDs-ON and different manufacturer since both have the same model number IRF740?

[timenutgoblin]

Just some observations...

R36 (47R) is part of the snubber network and looks like it has a dry joint. Reflow it.

On the schematic, component designator C56 is duplicated. C28 (220pF) on the PCB is C56 (220pF) on the schematic. The other C56 is 3.3nF (greencap) capacitor. This isn't part of Q3 circuit, though.

Since Q3 MOSFET has failed numerous times, what is the likelihood that the primary HVDC winding (pins 5 and 6) has sustained some damage when Q3 drain-source channel short circuits? As a last resort, try swapping transformers T1 and T2 and see if the fault follows the transformer or stays with Q3 MOSFET. I suggest this idea because PSU1 and PSU2 are seemingly identical from what I can see in the schematic and are common to the startup circuit of Q6 etc. Maybe exercise some caution/discretion with this suggestion.

[dl6lr]

This is a tough one, it doesn't look that complicated, but I still can't see a clear reason for the MOSFET to blow or C70 to bulge. Is R23 the correct value? Looks like it might have got hot at some point?

For 230V version (which is configured on the board), the upper (A-X) and lower (Y-B) parts are in series to mains voltage. Both sheets of the main PSU schematics are in parallel, so C59||C69 and C60||C70. If the upper part (with the Mosfet Q3) is shorted, the whole voltage is applied to Y-B, where the cap C70 is only rated to 200V. R2 and R23 are the bleeders/balancers (where during normal operation the balancing should be done by the current the switches are drawing). Performance of caps C59/C60/C69/C70 will also affect the balancing, but they are replaced. With one bulging, I doubt these are still in spec. Even short times of overvoltage will degrade the caps, maybe they resume from insulation defects with some time reforming them. Running it on 110V with changed configuration is more safe to the board during testing.

This a self resonating circuit, it starts with the pulse to Q3/Q4. If there is a short or break in the transformer, current flow may not stop before the FET is broken, it will happily convert to a self destructing circuit. If possible, disconnect everything from the transformer and try to inject small pulses and see the ringing with a scope. Compare both transformers.

[dark_hawk]

Just some observations...

R36 (47R) is part of the snubber network and looks like it has a dry joint. Reflow it.

On the schematic, component designator C56 is duplicated. C28 (220pF) on the PCB is C56 (220pF) on the schematic. The other C56 is 3.3nF (greencap) capacitor. This isn't part of Q3 circuit, though.

Since Q3 MOSFET has failed numerous times, what is the likelihood that the primary HVDC winding (pins 5 and 6) has sustained some damage when Q3 drain-source channel short circuits? As a last resort, try swapping transformers T1 and T2 and see if the fault follows the transformer or stays with Q3 MOSFET. I suggest this idea because PSU1 and PSU2 are seemingly identical from what I can see in the schematic and are common to the startup circuit of Q6 etc. Maybe exercise some caution/discretion with this suggestion.

R36 is making a good connection, I reflowed it just in case. Your observations on the schematic are correct. Thanks for examining the schematic thoroughly to find such mistakes.

I have measured all the windings on the both the transformers and compered the results, using an LCR meter at 1 KHz, inductance and resistance values are identical. I don't know if that is enough.
When testing the circuit I always used an incandescent lamp in series with the 220v input along with a high quality fast blow fuse. When the short happens the lamp just lights up preventing - hopefully - any damage to other components.
I don't think at this point that the transformer is damaged, or at least that's what I hope.

[dark_hawk]

This is a tough one, it doesn't look that complicated, but I still can't see a clear reason for the MOSFET to blow or C70 to bulge. Is R23 the correct value? Looks like it might have got hot at some point?

For 230V version (which is configured on the board), the upper (A-X) and lower (Y-B) parts are in series to mains voltage. Both sheets of the main PSU schematics are in parallel, so C59||C69 and C60||C70. If the upper part (with the Mosfet Q3) is shorted, the whole voltage is applied to Y-B, where the cap C70 is only rated to 200V. R2 and R23 are the bleeders/balancers (where during normal operation the balancing should be done by the current the switches are drawing). Performance of caps C59/C60/C69/C70 will also affect the balancing, but they are replaced. With one bulging, I doubt these are still in spec. Even short times of overvoltage will degrade the caps, maybe they resume from insulation defects with some time reforming them. Running it on 110V with changed configuration is more safe to the board during testing.

This a self resonating circuit, it starts with the pulse to Q3/Q4. If there is a short or break in the transformer, current flow may not stop before the FET is broken, it will happily convert to a self destructing circuit. If possible, disconnect everything from the transformer and try to inject small pulses and see the ringing with a scope. Compare both transformers.

The strange thing is that the bulged cap still reads the same values as a new unused one. Capacitance and ESR using at ESR meter at 1 KHz. A credit to quality caps I think. I changed it of course though.
Unfortunately I don't have a scope.

Since this a self resonating circuit, and the mosfets are in pull push configuration. Should both mosfets be identical? Same manufacturer and properties.
Or is it acceptable that they have slightly different RDs-ON and different manufacturer since both have the same model number IRF740?
And if the mosfets are slightly different, will that cause failure.
Another question, When I'm testing the circuit, I have disconnected the sound amplifier boards which are the load to that output, will the lack of a load on such a circuit cause a failure?

[sean0118]

If the upper part (with the Mosfet Q3) is shorted, the whole voltage is applied to Y-B, where the cap C70 is only rated to 200V.

Ah, that makes sense! I wasn't thinking about the voltage over C70 during the fault, that's one mystery solved.

[perieanuo]

This is a tough one, it doesn't look that complicated, but I still can't see a clear reason for the MOSFET to blow or C70 to bulge. Is R23 the correct value? Looks like it might have got hot at some point?

Since this a self resonating circuit, and the mosfets are in pull push configuration. Should both mosfets be identical? Same manufacturer and properties.
Or is it acceptable that they have slightly different RDs-ON and different manufacturer since both have the same model number IRF740?
And if the mosfets are slightly different, will that cause failure.
Another question, When I'm testing the circuit, I have disconnected the sound amplifier boards which are the load to that output, will the lack of a load on such a circuit cause a failure?

i'm not really mosfet expert, but i think rdson is not very very important, mosfets could be different (in some limits of course), for me frequency specs could move the boat (commute time).
i doubt the transformer is damaged, you can swap and test, fortunately you can in this case.
btw, if you want real protection, put the bulb in series with mosfet drain (the one who blows), then you save everybody, mosfet included. calculate the 'biggest' bulb accepted and use one close to that wattage (regarding max current accepted by your mosfet). this was always my technique, putting the bulb on mains lets place to such thing like trace popcorning, transformer overcurrent and so on. of course it doesn't protect 100%, with shorted mosfet you have to disconnect quick the mains, you'll put some big dc current into the smps transformer, but keeps it under control better
so i think you missed something in that mosfet control loop or maybe we don't really understand how this ps is working
you may also change c32 ( i strongly advice you to CHANGE the caps, not measure them, until this is resolved).
why? look at the bulged cap example...with your lcr or multimeter powered from 9V you can measure 2-3 parameters, but never ever you will put that cap to 1KV voltage, so the (potentially damaged) isolation will never show up. usually your method is ok, but nor those non-baptized damaged pcb, the caps is better to be replaced, we will analyze what's happened after successful repair

[dark_hawk]

i'm not really mosfet expert, but i think rdson is not very very important, mosfets could be different (in some limits of course), for me frequency specs could move the boat (commute time).
i doubt the transformer is damaged, you can swap and test, fortunately you can in this case.
btw, if you want real protection, put the bulb in series with mosfet drain (the one who blows), then you save everybody, mosfet included. calculate the 'biggest' bulb accepted and use one close to that wattage (regarding max current accepted by your mosfet). this was always my technique, putting the bulb on mains lets place to such thing like trace popcorning, transformer overcurrent and so on. of course it doesn't protect 100%, with shorted mosfet you have to disconnect quick the mains, you'll put some big dc current into the smps transformer, but keeps it under control better
so i think you missed something in that mosfet control loop or maybe we don't really understand how this ps is working
you may also change c32 ( i strongly advice you to CHANGE the caps, not measure them, until this is resolved).
why? look at the bulged cap example...with your lcr or multimeter powered from 9V you can measure 2-3 parameters, but never ever you will put that cap to 1KV voltage, so the (potentially damaged) isolation will never show up. usually your method is ok, but nor those non-baptized damaged pcb, the caps is better to be replaced, we will analyze what's happened after successful repair

Thanks for the reply.
My understanding of how this PS works is that Q6 drives the gates of Q3 and Q4. Q3 works with Q1 to drive transformer 1 and Q4 with Q2 to drive transformer 2. So you are correct, I don't really understand in depth how it works, some rudimentary info.
If you mean by the mosfet control loop Q6 and the components around it, all is fine. I even desoldered the diacs and tested them and they are working fine. Also Q6 was desoldered and tested out of circuit using a transistor tester.
Unfortunately I don't have a stock of those caps. C32 was desoldered and tested with an LCR meter is reading fine.

I switched the mosfets around so that Q3 and Q1 are the identical and Q4 and Q2 are identical, to prevent further damage I followed dl6lr advice and hooked the board to 110V power source and it worked fine. Gave it 5 minutes and it kept working fine.
Switched the jumpers to 230V configuration, and started the system, the lamp in series started to glow after I heard something like a hiss or small sparks somewhere on the board, immediately disconnected the power and nothing seemed to blow. Tried again but still can't figure out where that sparky sound is coming from. Switched the jumpers back to 110v and thought of measuring some voltages, it worked for like 10 seconds and the the relay started clicking on and off rapidly. Now Q1 is blown.

[perieanuo]

of course q6 was good, without him the ps blows at first start or does not start at all
the whissle sound is the frequency going rogue and i think we should start to investigate if 'lower' branch is working (i mean Q2 and his friends, maybe the negative alternance isn't right causing the positive to die)
if you don't want to replace caps, at least start to exchange them between supplies, but C54 should be changed being unique.
the relay is powered by the other ps, he sould remain steady, except if D1 bridge is on excessive load (shorted on output), making top243 ps to fail.
i think you covered the basis like verify if D1 is ok and you have the correct DC voltage on C6 and C69-C70, a shorted diode in that bridge will make dust from our suppositions 
you're right about mosfet control loop Q6+D6-D7 putting Q2-Q3 in conduction, but the control loop extends to those secondary windings 7-8 and 1-2 for example, those are shutting off the mosfets via C57...
for me, resistors and caps are easy to diagnose so the heavy task is to exchange T1 and T2 eliminating this (i don't think T1 or T2 are bad, those models are pretty sturdy, but who knows, attacking them with repetitive blows may damage 5-6 winding.
keep digging

[dl6lr]

I switched the mosfets around so that Q3 and Q1 are the identical and Q4 and Q2 are identical, to prevent further damage I followed dl6lr advice and hooked the board to 110V power source and it worked fine. Gave it 5 minutes and it kept working fine.
Switched the jumpers to 230V configuration, and started the system, the lamp in series started to glow after I heard something like a hiss or small sparks somewhere on the board, immediately disconnected the power and nothing seemed to blow. Tried again but still can't figure out where that sparky sound is coming from. Switched the jumpers back to 110v and thought of measuring some voltages, it worked for like 10 seconds and the the relay started clicking on and off rapidly. Now Q1 is blown.

This is somehow strange. The FETs to be identical should be Q3/Q2 and Q1/Q4, as those build a PP drive to the transformer each. I would just disable the second PSU completely to isolate the problem (or use it first to get some measurements on the working circuit).
So the unit worked for 5 minutes in 110V configuration with 110V supplied? With or withoud load?

Some thoughts, maybe wrong: TH1 is the inrush current limiter. The relay is driven from electronics supplied by the AUX PSU so if the voltage is low because the bulb starts to glow, the relay may switch off. Just bridge the relay for tests. When the MAIN PSU gets its DC input voltage, limited by TH1, C59,C60,C69,C70 will be charged, either in parallel (110V) or series (230V). With Q6 and R14 the two DIACs will be pre-charged, at about 60V the DIAC will drive the Gate of Q3 positive against the source, so Q3 will switch on and current from will flow thru the transformer. The source will "jump" to high voltage but the gate floats with it. The current induced into the second winding and as long as voltage on pin 7 of T1 rises, current flows thru C54 and keeps the gate high. When the core saturates, current thru Q3 increases but current/voltage at pin 7 will decrease, so gate will be pulled low thru R63/R66. Current thru Q3 stops and decreases, inducing voltage in opposite direction which will turn on Q2 thru C56.

Ratings of the FETs are 10A, the bulb should draw a current that is lower. So even one FET stuck should not blow it up. 200W on 230V will be roughly 1A with higher inrush. I would try a lower value like 60W to start with, but a IRF740 should not bother the current from the lamp. So there must be another error, like failing snubber or excessive startup drive.

Check R14, R15, R16, Q6, D14 and C54. With D15 and D16 removed, the PSU should not start up (but beware, any kind of spike to one of the gates will start it). Check D23 as a failing zener may exceed the Ugs (20V) of Q3. It should have 0.7V forward and 15V in reverse. In circuit you will not get the right values as R63-R66 are parallel to it.

[dark_hawk]

of course q6 was good, without him the ps blows at first start or does not start at all
the whissle sound is the frequency going rogue and i think we should start to investigate if 'lower' branch is working (i mean Q2 and his friends, maybe the negative alternance isn't right causing the positive to die)
if you don't want to replace caps, at least start to exchange them between supplies, but C54 should be changed being unique.
the relay is powered by the other ps, he sould remain steady, except if D1 bridge is on excessive load (shorted on output), making top243 ps to fail.
i think you covered the basis like verify if D1 is ok and you have the correct DC voltage on C6 and C69-C70, a shorted diode in that bridge will make dust from our suppositions 
you're right about mosfet control loop Q6+D6-D7 putting Q2-Q3 in conduction, but the control loop extends to those secondary windings 7-8 and 1-2 for example, those are shutting off the mosfets via C57...
for me, resistors and caps are easy to diagnose so the heavy task is to exchange T1 and T2 eliminating this (i don't think T1 or T2 are bad, those models are pretty sturdy, but who knows, attacking them with repetitive blows may damage 5-6 winding.
keep digging

When the relay went to clicky town the last time it was powered, there was indeed a short on the bridge due to Q1 being blown. Once Q1 was removed the short disappeared. All the bridge's diodes are fine.
C64 is fine, reads 10.5N and has a better ESR that any other cap I have with the same value.
All resistors read the intended value in circuit, expect for the bleeder resistors which I have to desolder to measure, and both check out fine.

This is somehow strange. The FETs to be identical should be Q3/Q2 and Q1/Q4, as those build a PP drive to the transformer each. I would just disable the second PSU completely to isolate the problem (or use it first to get some measurements on the working circuit).
So the unit worked for 5 minutes in 110V configuration with 110V supplied? With or withoud load?

Some thoughts, maybe wrong: TH1 is the inrush current limiter. The relay is driven from electronics supplied by the AUX PSU so if the voltage is low because the bulb starts to glow, the relay may switch off. Just bridge the relay for tests. When the MAIN PSU gets its DC input voltage, limited by TH1, C59,C60,C69,C70 will be charged, either in parallel (110V) or series (230V). With Q6 and R14 the two DIACs will be pre-charged, at about 60V the DIAC will drive the Gate of Q3 positive against the source, so Q3 will switch on and current from will flow thru the transformer. The source will "jump" to high voltage but the gate floats with it. The current induced into the second winding and as long as voltage on pin 7 of T1 rises, current flows thru C54 and keeps the gate high. When the core saturates, current thru Q3 increases but current/voltage at pin 7 will decrease, so gate will be pulled low thru R63/R66. Current thru Q3 stops and decreases, inducing voltage in opposite direction which will turn on Q2 thru C56.

Ratings of the FETs are 10A, the bulb should draw a current that is lower. So even one FET stuck should not blow it up. 200W on 230V will be roughly 1A with higher inrush. I would try a lower value like 60W to start with, but a IRF740 should not bother the current from the lamp. So there must be another error, like failing snubber or excessive startup drive.

Check R14, R15, R16, Q6, D14 and C54. With D15 and D16 removed, the PSU should not start up (but beware, any kind of spike to one of the gates will start it). Check D23 as a failing zener may exceed the Ugs (20V) of Q3. It should have 0.7V forward and 15V in reverse. In circuit you will not get the right values as R63-R66 are parallel to it.

Thanks for explaining how it works.

It was tested without a load.
All of R14, R15, R16, Q6, D14 and C54 were tested and checks out. I desoldered D23 and checked out of circuit and it's working fine, 14.8V. In circuit it gives voltage drop in both directions, all the zeners connected to the 100 ohm gate resistors read voltage drop in both directions while in circuit.

As a last resort here is a photo of the current sorry state of my board, maybe there is something wrong glaring me in the eye and due to the lack of experience I can't see it.

This board went through a lot of thermal stress due to bad thermal design. When the mosfets were working it would reach 70-80 C while in standby doing nothing, and that's in open air not inside the subwoofer box where there is no ventilation.  It also had that horrible glue which turns brown and conductive when heated. The solder pads and pours are in bad shape, they get lifted and torn with the slightest move. Even at low soldering temperature of 300 C it tears through the pads. And as you can see, it turned into a wire jungle.