Why a IRFP064 MOSFET gets burnt?

[xyz9915]

Please view the attachment of SMPS circuit which was modified by me to fulfill my requirements. I made three PCB's of the above SMPS (Output about 220V/1.5A DC). The problem is that one PCB is working fine and delivering my desired voltage & current (i.e. 220V/1.5A DC). But in the remaining two PCB's, power MOSFET (IRFP064) gets burnt within 2-3 minutes.
My question is that when the three PCB's are identical, so why one board is working satisfactorily & other two boards are not functioning properly? I thoroughly checked everything in all PCB's but remained unsuccessful. Hence please help in this respect.

Further another question is that can I use power transistors (e.g TIP35) despite using MOSFET's? By providing sufficient driving to the transistors?

Note: The IC is TL494
Best regards

 

[banjo]

tranzistor irfp064

Do both MOSFET get destroyed or just one?
Do the unfunctioning boards generate the proper voltage for 2-3 minutes before failure?
If you replace the MOSFETs and run the boards for 30 secs, is the output voltage OK? Are the MOSFETs running hot?
Do the MOSFETs in the good board run at a different temperature?


What is the relationship between BATT and VCC?

Generally MOSFETs fail for three reasons. The load is too high for the device and IDD exceeds the spec. Secondly, the gate drive is too low and the MOSFET never fully switches on. This causes part to overheat due to operation in the linear region. Thirdly, the device fails due to excessive transients that occur at the switching points. This usually requires snubbing circuits.

If one boards runs fine for a long time, then it really sounds like an assembly error. The easiest way to find it is to first buy more MOSFETs. Then start swapping parts from a bad board to the good board. When the good board stops working, you know which part is causing the problem.

 

[FvM]

astable tl494 duty cycle mosfet

Generally, first operation should be performed with a current limited power supply and no output load, carefully watching all critical waveforms. Most assembly errors and initial part defects should be recognizable this way before they cause catastrophic failure.

Personally, I miss a overcurrent detection for the output stage.

BJT basically can do the job, but the base drive circuit would considerably reduce the overall efficiency. To my opinion, the only suitable solution would be by a base drive transformer in this case.

 

[xyz9915]

totem pole circuit bd139 bd140

Thanks to all for the reply.
The MOSFET's are 5+5 in each side. Before destroying, the board delivers output for about 1 minutes then MOSFET's gets destroyed. I think this is due to
the gate drive is too low and the MOSFET never fully switches on. If the complementary pairs of BD139 & BD140 is not sufficient to provide proper current for gate charging/discharging, then how I drive these (5+5) MOSFET's to fulfill my requirement?
The snubbing circuitry is already used (4.7 Ohms resistor+.01uf capacitor in series)

The other question is with (FVM) That if I use BJT (e.g TIP35C) so what will be the transformer size & turns ratio? for driving the BJT's. The operating frequency is about 36KHz.

 

[FvM]

astable tl494 mosfet

I don't think, that changing the power stage to BJT is promising any benefit, unless you're intending a museum design with veteran parts.

Gate drive is basically limited by the individual 10 ohm resistors and common 3.3 ohm resistors. This results in a limited gate dV/dt and some additional switching losses, but should be sufficient for 36 kHz. I don't expect a general problem in this respect. But without seeing the actual circuit wiring, this is hard to decide.

Considering the transistor switching capacity, a five-fold parallel circuit seems to be somewhat oversized. You most likely get better overall efficiency with e. g. two transistors for each branch.

There is no alternative to visualizing the circuit opreation with an oscilloscope, I think.

 

[banjo]

burned snubber resistor

While oscilloscope waveforms would be great info, perhaps that is not available. You can also try reducing the number of MOSFETs and reducing the load current.
If you have only one MOSFET per winding and run the supply without load, are the MOSFETs destroyed? Do they run hot? If they run fine with a reduced number of MOSFETs, then it could be a gate drive issue.

Are there ten 10 ohm resistors in the gate circuits? Or do five MOSFETs share a single 10 ohm resistor? I would suggest that each MOSFET have a individual resistor. Ten ohms may be too low a value. One job of these resistors is to damping ringing in the gate circuit that can cause MOSFETs to oscillate at the transition points.

 

[xyz9915]

resistor gets burnt transistor bad

under minimum load or no load condition, circuit works fine but when the load is applied, mosfet's get destroyed within 1-2 minutes.
The all (5+5) MOSFET's are individually connected with 10 ohms resistor. they are not sharing a common resistor. I agree with you that this is a gate drive issue. so what will be the solution? I tried to reduce the number of MOSFET's from 5 to 2 or say 1, but the MOSFET's gets too much hot even it is mounted on a large heat sink (exhaust fan is also used). That is why I am insisting to use minimum 5 MOSFET's in each side.
The other question which is still unanswered is that why the BD134/BD140 are not sufficient to provide proper current for gate charging/discharging?

 

[banjo]

bd139 and bd140 based totem pole circuit

As FvM stated oscilloscope traces would quickly answer your question about gate drive. In looking at the datasheet for TL494 and your circuit again, I question whether the problem is not in the stage preceding the BD134/BD140 combo. The TL494 outputs are configured as emitter follower. For turning on the MOSFETs, current goes through either D2 or D3 to turn on the BD139 as an emitter follower. The gates should be driven at voltage of V2 -(2.5 +2*(0.7)). This suggests that V2 must be around 14V to get a 10V Vgs at the MOSFETs. What is V2?
What concerns me more than the turn-on is the turn-off. From my crude analysis, it appears that turn-off is controlled by the passive turn-on f Q3 or Q4. With the TL494 outputs "low", the transistors passively turn-on due to the 1K bias resistors. This may not produce a very crisp turn-off waveform and could leave the MOSFETs to wander through the linear region for some time.
Perhaps you could state what design you modified and what changes you did?

 

[FvM]

ırfp064n smps

I must admit, that I simply assumed a sufficient gate drive voltage, but as banjo pointed out, this is rather unlikely. Also Vgs fall-time is questionable. But why guessing about parameters, that could be seen from a single Vgs waveform? Generally, the gate drive should be amplified to full level and sufficient rise and fall time.

It's not clear however, if the problem arises mainly from bad gate drive. Another interesting point is Vds at switch-off. With a push-pull transformer, I would expect some amount of stored energy absorbed by MOSFET avalanche at full operating current. Depending on the actual leakage inductance, it may be too high.

 

[xyz9915]

bd139 bd140 replacement

After receiving suggestions, I tested all three boards with frequency counter and found that even the one board is working satisfactorily but the frequency at the gates of MOSFET's is displaying randomly (ranging from 200KHz to 500KHz), when I disconnected the gates driving resistors (10 Ohms), the frequency counter displayed the actual frequency of the circuit (which is about 36KHz). I then removed the driving circuit (consisting of BD139/BD140) & also removed the diodes D2/D3 (1N4148) with 220 Ohms resistor and then again connected the gates. The frequency counter displaying 36Khz in one side & other side the frequency is showing as 64KHz (just double) but the frequency remained stable. I also tried with different values of resistor but both channels are showing different frequencies (36KHZ & its multiple) but the frequency remains constant in all cases, also tried the circuit function with different frequencies (from 20KHz to 60KHz) but same situation. The transformer (2+2turns/50 turns) I again rewound but the situation still remained unchanged. I also connected two capacitors of 2200uF/25V for decoupling, & snubber circuitry (0.01uf with 4.7 Ohms resistor in series) is also attached.
So, I think that this is the problem within gate driving section, so please help me for resolving the trouble.

 

[FvM]

bd139 alternative

Generally, individual gate resistors are recommended when paralleling transistors, they can be expected to reduce parasitic oscillations rather than enforcing them. But as I don't know the wiring of your circuit in detail, I can't know, if this general experience applies here.

Unfortunately, a frequency measurement by a counter doesn't say much about the monitored signal, at least for this type of circuit. So I don't want to jump to conclusions based on this observations. It is much more important, to know about stable gate voltage levels in on and off state. Particularly, if Vgs,on is too low, this may cause a too high resisdual voltage and parasitic oscillations as well. If so, increasing the drive voltage would be the only fundamental solution, adjusting gate resistors can cause only minor changes then.

 

[xyz9915]

bd139 bd140 base drive resistor

I used only one MOSFET/side in this modified schematic. File is already attached as 3-D. So the paralleling is not an issue.

Furthermore, someone informed me that since the circuit is running in voltage mode... a push-pull running that way can be subject to "flux walking".
Therefore, my question is that how I prevent the core from reaching saturate in this push pull mode?
Regards

 

[banjo]

driving mosfet by tl494

The MOSFET gate is a very high impedance node. It always has some ringing on it. Attaching a frequency counter to the gate will also count the ringing and thus produces the 200K to 500K readings you were seeing. If you want to use a frequency counter as a test, then it is best to either use a coil to sample the flux in the transformer, or add a small resistor between source and ground and sample across that. By sampling in that manner, you are measuring the frequency of the current waveform, which is what you are really interested in.
As FvM stated, the frequency count is not a very good test of circuit operation. Even the board that works for you gives weird frequency counts. My scope at work can measure frequency, however when looking at waveforms of properly working switching power supplies, it often complains that it cannot determine the base frequency. This is because the waveforms are too complex. (Frequency counters really only like 50% duty cycle square waves with clean edges.)

Assuming that you used to proper phasing when winding the transformer primary, then I would not put core saturation at the top of the suspect list. If you want to eliminate the core saturation issue, take the transformer from the working circuit and substitutie it into a non-working one. If the non-working start operating, then its something about the transformer.

You may want to look at:
https://www.electronics-lab.com/projects/automotive/012/index.html

Here they used the same PWM IC, but employ a special driver IC to insure adequate gate drive to the MOSFETs.

Finally, I keep coming back to your statement that 1 board out of three works fine. Can you really run one of the assemblies at full load with the MOSFETs running cool? If so, then either this one assembly has a "happy mistake" that makes it work or its components are not equivalent to the other assemblies.

 

[FvM]

bd139 bd140 1n4148 schematic

Core saturation isn't an issue to be handled by guesses rather than precise measurements, also other (more) likely candidates as unsufficient gate drive voltage, parasitic oscillations or exessive losses from MOSFET avalanche. It's not absolutely necessary to have state-of-the-art equipment like a current probe, although it's really helpful. Performing the necessary measurements with minimal equipment may require some intelligent workarounds and be more time consuming, however.

 

[xyz9915]

irfp064 resistor gate

Finally, I keep coming back to your statement that 1 board out of three works fine. Can you really run one of the assemblies at full load with the MOSFETs running cool? If so, then either this one assembly has a "happy mistake" that makes it work or its components are not equivalent to the other assemblies.

The board which works fine at full load gets much hot even the five MOSFET's were used in each side.[/quote]

 

[banjo]

IRFP064 gets burnt

If you swap the transformer from the non-working board, I think you can get a clue about core saturation. If the working unit continues to work with the other transformer, it is unlikely that its a core saturation issue.

Since an oscilloscope does not seem to be available to you, I think your next best choice is to do some SPICE simulations of your driver circuit. I would not concern myself the the TL494 SPICE model. Just replace it with a voltage source driven by a 36KHZ square wave. Set the amplitude of the voltage source to (V2-2.5). Use a series resistor to limit the current, probably about 30 ohms is a good starting point. Then use the SPICE simulator to track this square wave through a simulation of your driver circuit. This will give you an approximation of what the oscilloscope would show. If the simulation indicates insufficient drive, then you can tweak the simulation to improve the drive before transferring the changes to the actual circuit.

 

[xyz9915]

Re: IRFP064 gets burnt

Please see the attached waveforms taken from oscilloscope. The pattern (A) shows the waveform of MOSFET's gates when the positive supply (12V) not connected to the push-pull transformer. When it is connected, the waveform (B) produces (too much noise/spikes) even the snubber is connected. After changing frequency from 36KHz to around 50KHz, waveform (C) appears which seems better but the MOSFET's gets very hot within one minute at 30% load. I changed the circuit with SG3525 (totem pole output) but the same result. So, the most suspected thing is transformer. changing the turns ratio or frequency slightly change the waveform. So, what is the solution?

 

[FvM]

IRFP064 gets burnt

Waveform B and C look like Vbat is breaking down peridiodically, so you are unable to get a stable waveform display (at least with an anaog scope). Generally, you ahould be a stable waveform with spikes as well. I think that's something different than just spikes.

 

[banjo]

IRFP064 gets burnt

What is the volts/div of the scope shots? Where is the ground reference with respect to the scope shots?

The waveform in "B" seems to be noisy due to a triggering problem. Try using an external trigger back at the PWM IC output. The other thing that can make the scope display look like this is that the PWM IC is hunting for an operating point.

The waveform in "C" looks worse. Assuming the volts/div are the same, then I think "C" is showing the effect on the limited bandwidth of the MOSFET driver circuit.

Transformer issues can often be detected by looking at the coil current waveform. The easiest way to do this is to place a very small resistor, like 0.5 ohms, in series with the source lead of the MOSFET. One end of the resistor will be connected to gnd the other connected to the MOSFET. How you have a ground referenced voltage which is proportional the the transformer current. The current waveform should have a sawtooth sort of ramp to it. This indicates that the transformer is building up magnetic flux and is not entering saturation.

 

[xyz9915]

Re: IRFP064 gets burnt

The attached file contain waveforms taken from oscilloscope. Circuit work fine under 30% load. The problem is that about 70% of the load, the MOSFET's get suddenly hot and if the power supply is not disconnected, they burn within 3-5 minutes. I increased MOSFET's from 1 to 5 each side but the result is same, MOSFET's gets burnt. So, please help me is this respect.

Furthermore, although TL494 has a 45% maximum duty cycle, fix the problem?is possible by increasing deadtime?