shorting mosfets

[russlk]

I am at my wit's end. I am building a switching power supply (push-pull with 3524 IC) and the power mosfets are shorting. The mosfets are STW26NM60 (30 amps, 600 volts). I have 300 ohms in series with 560 pF across the primary for damping and 150 volt MOV across each half of the primary. I don't see a lot of spiking at the drain and the driver is NPN/PNP emitter follower, so I don't think the gate voltage is being exceeded (the mosfets have built in zeners also). As I increase the input voltage, the mosfet shorts at about 100 volts DC input. What else can cause the mosfets to short?

[Ron H]

Russ, I'm no expert on switchers, but I've heard of MOSFETs self-destructing due to high frequency oscillation. Do you have really short wires from your drivers to the gates? Have you added a small series resistor (47 ohms or so)? Also, emitter followers will oscillate with capacitive loads (MOSFET gates), which could also kill the MOSFETs. The same resistors will also help this problem.<p>Cheers-
Ron<p>[ August 05, 2003: Message edited by: RonH ]</p>

[russlk]

Thanks for the reply. The NPN/PNP drivers are surface mounted on small PC boards on top of the MOSFETs, the PC board is held by the same screw that holds the TO-247 Mosfet. The lead to the gate is about 1 inch, no resistor, but there is 100 ohms in series with the driver bases.

[russlk]

I tried putting 10 ohms in series with the gate, which has the happy effect of not smoking the driver when the mosfet shorts. The waveforms on my 50 mHz scope are clean, so I am still mistified, and I am running out of mosfets.

[rshayes]

First, look at your shorted parts. Is the short from gate to source or from drain to source? Gates are often protected with an internal zener diode, but it is probably possible to short that with sufficient energy. Drain to source or drain to substrate shorts would result from excessive drain voltage. Since you have 600 volt devices operating on 100 volts, this may not be very likely.<p>One unadvertized characteristic of power MOSFETs is the gate to drain capacitance. This can easily be several thousand picofarads. This capacity must be charged during turn on, and, with a high impedance driver, may slow the switching transitions considerably. Slow transitions increases the power dissipation, and may cause failure from overheating. This capacitance can also couple signals from the drain to the gate when the device is of. For example, when the opposite device turns on, a 200 volt pulse will appear on the drain of the off transistor. This will be coupled to the gate and will have to be absorbed by the driver circuit.<p>One clue here will be if you have to replace driver parts as well as switching devices. Sometimes the driver needs protection from the output circuit. Diodes from the gate to ground and to the driver power source can help insure that the driver output can't be driven outside its normal range by external signals.

[russlk]

As an experiment, I put 5 ohms in series with the drain. The mosfets now survive up to 125 VDC (the max my supply will provide). Do you suppose the turn-on transient exceeds the 120 amp pulse rating? I don't want a series resistor, but what other option is there?

[Edd]

Hey Russ:
When U refer to the drivers , I’m assuming that U are using the onboard drivers of that 3524 PWM chip. What I was wondering was if U had an RC snubber network across your xformer. For better lack of a schematic…refer to this Korean one and you can at least use the graphics unless you also want to process thru a translator site for the text. Its using the brother IC to yours….site is:……the fifth graphic down:
http://www.cpe.ku.ac.th/~yuen/204471/po ... ample.html <p>This, being in case that failure analysis revealed a punch through of the Drain/Source considering that the gate was OK.<p>Addenda: Also with an operational ??? circuit..at least as long as it lasts ....scope the center tap power input ...if that happens to be your config...and see how contaminated that point is .
I like to use a low ESR cap and a .1 ceramic bypass cap at that point with very short lead dress to gnd. Its also helpful to run a thermal temp check on the FETs after mods to check effectiveness. <p>73's de Edd
[email protected].........(Interstellar~~~~Warp~~~Speed)
[email protected]........(Firewalled-Spam*Cookies*Crumbs)

<p>[ August 05, 2003: Message edited by: Edd Whatley ]</p>

[Chris Smith]

What’s your frequency? <p>Are you saturating the Fet transistor and it stays on?<p>Are you driving a transformer from the Fets out put? <p>Does your switching between the two fets have one on, while the other is off? <p>Is Your transformer winding the right way, if they are center tapped? <p>Have you tried the push or pull, with out the other half running or connected?

[russlk]

Chris:
The frequency is 50 kHz.
The gate drive is 12 volts and the drain appears to stay saturated.
The transformer is push-pull, connected to the drains.
There is 1uS dead time, even at max pulse width.
The transformer is custom designed, by a professional designer.
I have not tried running one side only as that might saturate the core.
Thanks for the reply.

[russlk]

Edd:
I looked at the schematic, I don't understand the diode bridge between the two transformers. Very strange.

[Joseph]

Try putting a current limiting resistor in series with each MOV, maybe about 10 ohms.<p>I have wondered if built-in zeners could cause an inductive spike on the gates. Try mosfets without them.<p>Like Stephen said, the Miller effect can cause the mosfet that is supposed to be off to be turned on if the dV/dT on its drain goes too high.
That is one reason for gate resistors, to slow the turn-on times enough to protect the other mosfet.<p>Push-pull can easily saturate the core if even a little imbalance builds. I have found that a capacitor can be placed in series with the drain of one of the mosfets and almost full power can still be output from the circuit! Try that and see what happens.<p>Try finding a large supply of lesser mosfets for a cheap price to experiment with. I once bought 200 IRF840's for $50. <p>I am glad you placed the 560pF 300 ohm snubber across the transformer primary. But, you may also still need a series drain resistor on each mosfet to limit inrush current. You may not need this resistor if you rectify full wave, and pass the rectified voltage though a choke about 5 times larger or more than the secondary winding's inductance. Thos way, you can shift the current draw from the mosfets away the switching times.

[Joseph]

Maybe you can get some ideas from my yahoo thing also
http://groups.yahoo.com/group/switchmode/<p>If all else fails, try converting the configuration to half-bridge like I always use. You won't get imbalance problems as the capacitor keeps the output in balance. You would only be able to use one primary winding on your transformer and tie it to the totem pole with the mylar coupling capacitor.<p>I hope you solve the problem, as I experimented with the push-pull design because the lack of level shifting makes for simpler mosfet drive. But since I use ferrite toroids, the core got hot because there is no gap in the magnetic material. I like winding my own because it allows me to try different winding configurations to get just the perfect inductance on the primary. I found that if it was not in a sweer zone, reliability became a problem, especially if the primary inductance was jusr a little too low.

[Chris Smith]

50khz might be saturating the the core of the transformer? Run it slower? <p>Try running one leg, bare, no additions, no caps, no movs ,no nothing but the transformer. If you can slow it down also, you will be able to find out many things? <p>Keep your Digital volt meter away from the out put, I fried one that way!

[Joseph]

What Chris may mean is to slow the frequency down to very slow frequency and keep the pulse width the same as it was at the 50khz to observe what happens when the duty cycle is reduced to less than 1%. Once I reduced it so low that even though the core was saturating, it did not overheat the mosfets. I have also blown a meter,as Chris warns, but that can be a sign that you have huge RF waves being generated and very high dV/dT spikes; I actually use the warning beep on my meter nearby as a sign that I better kill power to the circuit or the mosfets are about to blow due to excessive dV/dT.<p>Do you have mosfet drivers being switched on by the output of the SG3524? It is also possible that your rise times and fall times are not getting fast enough. I always use some kind of CMOS gated driver even if I also buffer the output of something like a MAX626 OR International Rectifier IR2113. It is very hard to build up fast switching times for high power switching without one of hese standard drivers to build the fast, high peak current pulse. Then you should bufffer the outputs with BJT emitter source followers, or as I mention at my Yahoo site, the IRF7343 set as a source follower. You can use the IRF7343 in this way because its gate threshold voltage is only 1 volt, and the power supply of 15 volts or less is not enough to blow the gate if the source does not keep up with the gate during switching.

[Joseph]

The output of the mosfet driver should be buffered so that your peak drive capability is at least 2 amps per each 1000pF of mosfet gate capacitance. The IRF7343 has an on resistance of .1 ohm or less per device in the package, so theroretically has extremely high instantaneous peak current capability, even though it is listed as only about 20-30 amps per device in the package. It also has very fast rise and fall times, so it does not degrade that figure compared to the mosfet driver switching its gate. You do not need a gate damping resistor for the IRF7343 in this application because the low power supply voltage and the source follower configuration. <p>Your gate damping resistor for the output mosfets should be about 10 Ohms for 1000pF of gate capacitance. If your gate capacitance is doubled, you approximately cut the gate resistance in half. One thing I like about International Rectifier mosfets is that their data sheets specify the minimum gate resistor to use in high voltage switching applications, eg., 9.1 ohms for the IRF 840 which has a gate capacitance of something like 800 pf.