Fake MOSFET IRFP250

[Neukyhm]

Hello all, I'm building a custom ZVS, with 136nF and 14uH to work at ~110kHz using the IRFP250. Input voltage is 48V and the gates are driven with a 12V psu.

I simulated the entire circuit in Orcad and it worked, then I build the circuit and it worked for 15 seconds until both MOSFET burnt.

I was able to measure voltage with my oscilloscope at drains before that, 138V peak, that's below IRFP250's maximum 200V so now I think that my MOSFETs (bought on eBay) are fake (?).

What do you guys think? The MOSFET should have resist that voltage unless it is not a real IRFP250.

[xani]

Did you also scoped gate voltage ? Did they started heating from the start ? What current? And did they had radiators? Not having fast enough driver at that frequency might cause mosfet to not turn on fast enough and have enough on/off losses to just burn from excessive heat

[Neukyhm]

Did you also scoped gate voltage ? Did they started heating from the start ? What current? And did they had radiators? Not having fast enough driver at that frequency might cause mosfet to not turn on fast enough and have enough on/off losses to just burn from excessive heat

The MOSFET had radiators and they were not hot. Gates voltage is below 12V as I'm driving them with another 12V PSU and I used 100ohm resistors instead of 470ohm to allow the MOSFET to turn on faster, so I don't think that's the problem.

[NiHaoMike]

That still sounds high. Typical gate resistor values are on the order of 5-15 ohms.

[duak]

I agree that the gate drive resistors should be a much lower value.  My experience is that larger values cause the transition times and subsequent power dissipation to increase.  I recall one designer's circuit actually oscillating in the MHz range with 100 R series resistors.  Note also that the gate driver should have a low output impedance and sufficiently high current drive.

Sometimes the transition times have to be lengthened to meet EMI needs.  I've found it's better to add some capacitance in parallel with the gate rather than increase the value of the series resistor.

I built a transducer test jig once based on an H bridge driver.  If memory serves, it ran at about 100 KHz and had output voltage transition times on the order of 50 ns.  The load was capacitive and I calculated that the peak currents were in the tens of amps.  The FETs were rated for a continuous current of maybe 10 - 20 A, but their peak current limit was much higher.  It worked because the gate driver was low impedance and could deliver the correct waveform.

Cheers,

[Neukyhm]

As I said, people normally use 470ohm for gate resistors in a ZVS but I used 100ohm. That combined with a 12V PSU only for the gates means 12/100=0.12A for gates.

IRFP250's gate charge is ~129nC so 129E-9/0.12=1us that the gate needs to charge, that's faster than my switching speed so I don't think that the problem is that my MOSFETs need to turn on faster.

[David Hess]

How fast is your oscilloscope?  Did you use good probing technique?  Fast peaks or oscillation might have been missed.

Is the body diode being forward biased?  Reverse recovery time can cause problems and failure.

Momentary excursions of the gate voltage cause cause failure.  Try adding a 15 volt zener diode directly between the gate and source to protect it.

Based on the capacitance, 100 ohms seems high for a gate resistor but maybe not for ZVS.

[Kleinstein]

For a switching application there should be more modern MOSFETs than the rather old IRFP250. There are modern ones that include protection against to much gate voltage.

Ebay is still a good source for fakes - but sometimes you also get real ones.

[Neukyhm]

For a switching application there should be more modern MOSFETs than the rather old IRFP250. There are modern ones that include protection against to much gate voltage.

Do you know some MOSFETs with low gate charge, high voltage drain and low Rdson? If I can find an alternative to the IRFP250 I will go for it.

Edit: I have tested the MOSFETs with my multimeter in diode mode, it seems that only one of the two MOSFETs died, the other one seems good, maybe the dead one was faulty.

[Zero999]

For a switching application there should be more modern MOSFETs than the rather old IRFP250. There are modern ones that include protection against to much gate voltage.

Do you know some MOSFETs with low gate charge, high voltage drain and low Rdson? If I can find an alternative to the IRFP250 I will go for it.

Those parameters are mutually exclusive. In other words, higher drain voltages, result in higher R_ON and lower R_ON typically requires higher gate charge.

[duak]

Is the circuit based on something like this? http://www.instructables.com/id/ZVS-Driver/  If so, I hadn't realized that it was a self oscillating design.  I was thinking it used a separate driver that had defined rise & fall times.  I see that this article mentions the BVds of the FETs should be at least four times the supply voltage.

My experience with power FETs is that they really don't like being in the linear region with high Vds & high Id because their active cells do not share the drain current equally.  In a ZVS driver this shouldn't be a problem because Vds should be quite low, however, it'd be a good idea to check to see if the FET is ever on in a high dissipation mode for long.  Note that FETs can also oscillate at high frequencies in the linear region because of parasitic inductance in the circuit.  The designed operating frequency might be 100 KHz but these are fast devices and something else could be happening.

Cheers,

[Kleinstein]

The gate charge an On resistance also depends on the technology - modern MOSFETs got quite a lot better, though less robust when it comes to FBSOA. The IRFP250 is more like a very old type with rather good FBSOA performance but not so good at switching.

A rather random chosen more modern one would be an IPP320N20N3:
less than half the R_On and still only about 1/3 the gate charge.

[Neukyhm]

Is the circuit based on something like this? http://www.instructables.com/id/ZVS-Driver/  If so, I hadn't realized that it was a self oscillating design.  I was thinking it used a separate driver that had defined rise & fall times.  I see that this article mentions the BVds of the FETs should be at least four times the supply voltage.

My experience with power FETs is that they really don't like being in the linear region with high Vds & high Id because their active cells do not share the drain current equally.  In a ZVS driver this shouldn't be a problem because Vds should be quite low, however, it'd be a good idea to check to see if the FET is ever on in a high dissipation mode for long.  Note that FETs can also oscillate at high frequencies in the linear region because of parasitic inductance in the circuit.  The designed operating frequency might be 100 KHz but these are fast devices and something else could be happening.

Cheers,

Yes, it's exactly that circuit (different capacitor/inductance) but without zeners because I have two PSU, one to power the ZVS and a second 12V PSU only for gates, so zeners are not needed.

Also if the problem here were that my MOSFETs operated in the linear region for too long I think that they would have lasted longer and become hot, but it died in 15 seconds and the heatsink was cold.

[T3sl4co1l]

Not having fast enough driver at that frequency

That still sounds high. Typical gate resistor values are on the order of 5-15 ohms.

I agree that the gate drive resistors should be a much lower value.

For a switching application there should be more modern MOSFETs than the rather old IRFP250.

Sorry gentlemen but this is NOT a switching application, it's a linear circuit!

(Indeed, one can attempt to switch-ify the circuit, but it quickly becomes a cluster, until you give up and embrace the full digital equivalent, a resonance-tracking PLL.  At that point, you might as well drop the push-pull configuration as well, and use a more traditional bridge circuit.)

FWIW, using newer transistors may well have the downside that they can oscillate at 100-400MHz.  Usually, a ferrite bead on the gate and source is enough to deal with this, but that's typical of a low impedance drive situation, and YMMV here.  If you have a 100MHz scope, you might miss such behavior!  But with a scope much slower still, you'll definitely be able to verify if the IRFP250 is getting adequate gate drive (i.e., it's not slew rate limited).

Note also that the circuit requires shorting-mode commutation.  It's a current-source inverter topology, hence the series inductance in the supply.  Ideally, the high impedance supply should extend all the way to DC, using a constant current bench supply, say -- unfortunately, almost no bench supply has a true CC output, they inevitably screw it up with a big stupid capacitor across the output terminals, for no good reason.  

Many of the quirky properties of the circuit, only seem quirky because we are not used to seeing the current-mode transformation of traditionally voltage-mode circuits.  Hence, instead of dead time between switching, there must be shorting time; instead of CV supply, CC supply; bypass inductor instead of bypass capacitor; push-pull instead of half-bridge; etc.  (If you're curious, it's a fun exercise to enumerate the series-parallel, voltage-current transformations of conventional switching topologies.  Some are in use under a different name; many are not!)

IRFP250 is a pretty damn solid MOSFET, so if it's cooled adequately, there must be something else going on.  Current surges?  Perhaps due to the above reason (excessive current available from the supply)?  Is the oscillation being quenched (low Q load) resulting in shorting out the supply through the FETs?  Is the gate drive accidentally being turned off before the supply inductance has discharged to zero current?  (Again, shorting mode operation: the gate drive must be available before and after drain power is applied!)

As for counterfeits, if you bought them off eBay or Ali Express or whatever, you've no one to blame but yourself.  Suspect 100% of cheap crap bought there.  If you don't have the time and money to test and verify suspect components, buy from a reputable distributor.

Tim

[Neukyhm]

Sorry gentlemen but this is NOT a switching application, it's a linear circuit!

(Indeed, one can attempt to switch-ify the circuit, but it quickly becomes a cluster, until you give up and embrace the full digital equivalent, a resonance-tracking PLL.  At that point, you might as well drop the push-pull configuration as well, and use a more traditional bridge circuit.)

FWIW, using newer transistors may well have the downside that they can oscillate at 100-400MHz.  Usually, a ferrite bead on the gate and source is enough to deal with this, but that's typical of a low impedance drive situation, and YMMV here.  If you have a 100MHz scope, you might miss such behavior!  But with a scope much slower still, you'll definitely be able to verify if the IRFP250 is getting adequate gate drive (i.e., it's not slew rate limited).

Note also that the circuit requires shorting-mode commutation.  It's a current-source inverter topology, hence the series inductance in the supply.  Ideally, the high impedance supply should extend all the way to DC, using a constant current bench supply, say -- unfortunately, almost no bench supply has a true CC output, they inevitably screw it up with a big stupid capacitor across the output terminals, for no good reason.  

Many of the quirky properties of the circuit, only seem quirky because we are not used to seeing the current-mode transformation of traditionally voltage-mode circuits.  Hence, instead of dead time between switching, there must be shorting time; instead of CV supply, CC supply; bypass inductor instead of bypass capacitor; push-pull instead of half-bridge; etc.  (If you're curious, it's a fun exercise to enumerate the series-parallel, voltage-current transformations of conventional switching topologies.  Some are in use under a different name; many are not!)

IRFP250 is a pretty damn solid MOSFET, so if it's cooled adequately, there must be something else going on.  Current surges?  Perhaps due to the above reason (excessive current available from the supply)?  Is the oscillation being quenched (low Q load) resulting in shorting out the supply through the FETs?  Is the gate drive accidentally being turned off before the supply inductance has discharged to zero current?  (Again, shorting mode operation: the gate drive must be available before and after drain power is applied!)

As for counterfeits, if you bought them off eBay or Ali Express or whatever, you've no one to blame but yourself.  Suspect 100% of cheap crap bought there.  If you don't have the time and money to test and verify suspect components, buy from a reputable distributor.

Tim

So much information, thank you 

Now that you mention it, I remember to turn off the gates PSU BEFORE the other 48V PSU feeding the circuit, that may has killed the IRFP250.

If that's the problem then it's my fault, I know that I have to turn off the 48V PSU before turning off the gates PSU in order to allow the resonant circuit to discharge the stored energy through a fully on MOSFET.

[Audioguru]

An entire large city in China makes nothing but fake parts for ebay.

[duak]

This is indeed an analog circuit with a number of analog sub-circuits.

"Yes, it's exactly that circuit (different capacitor/inductance) but without zeners because I have two PSU, one to power the ZVS and a second 12V PSU only for gates, so zeners are not needed."

There's a difference between a non-clamped RC network driven from 12 V and the same network driven from 48 V but clamped by a 12 V zener.  In the latter, the gate voltage will linger in the linear mode for a much shorter time.  I'd guesstimate 1/4 to 1/8 as long.

"Also if the problem here were that my MOSFETs operated in the linear region for too long I think that they would have lasted longer and become hot, but it died in 15 seconds and the heatsink was cold."

These devices are arrays of small cellular FETs in parallel with slightly different gate thresholds & transconductance (Id vs Vgs) and don't share current equally in linear mode.  Some cells will get hot enough to fail even though the overall device is well under the thermal limit.  If these devices were in the old TO-3 package it'd be easy to decap it to look at the die.  I'd bet that only a few cells would be damaged.  It's a bit harder with the TO3P package but not impossible.  (With major failures, the TO3P can auto-decap)

This isn't to say the devices aren't counterfeit, or that there isn't a supply sequencing problem or that BVds isn't being exceeded but that the gate drive is the final nail in the coffin.

Best o' luck

[Buriedcode]

All good info here, but perhaps I'm missing something - where is the schematic and the exact specs of the transformer? 

All too often there are posts about switching converters where the switch has fried, or the output isn't what is expected, or 'it gets hot' and people get lost in specifics.  Designing DC-DC converters without dedicated IC's (and their application schematics/layouts) to 'work' is generally difficult, and involves calculations and measurements to get the transformer specs right, as well as part selection.

I know many have copied a schematic off the web, with the same 'number of windings' on a flyback and got it to work, but that is mostly luck... those who haven't got it to work have just built a circuit with no idea how to debug it, or even what to try except 'use a bigger MOSFET' or 'put a resistor in series'. 

Whilst it is certainly possible to get fake discretes in this day and age, I don't think it's even in the top 10 things to check.  Peak drain voltage?  Primary current? Gate waveform/voltage?  Whats the transformer spec? Self-oscillating converters can be deceptively simple, often with the transformer specs being the most important (inductance, saturation current/knee etc..). Driving a self-oscillating converter with fixed waveforms without knowing the transformer specs is fumbling in the dark - the transformer itself is part of the feedback.

Sorry if it sounds harsh, I just don't want you to buy up lots of MOSFETS and pop them one by one.  If you have the facility to test components, then debugging is obviously going to be much easier - but I don't know what you have access to form your post.  You could have a masters in Power Electronics, or you could have no experience, copy a schematic and hook it up to a big PSU and expect something to happen - I don't know But most likely, like most here, somewhere inbetween.

[Siwastaja]

Whilst it is certainly possible to get fake discretes in this day and age, I don't think it's even in the top 10 things to check.

While you are completely right otherwise, I disagree with this particular point: replacing these parts (and everything else sourced from Ebay, unless you specifically know better) should be #1. In my experience, if you buy cheapish IR**** FETs from a Chinese Ebay seller, they are almost 100% guaranteed fakes. The typical symptom is an order of magnitude higher Rds(on). I have, stupidly, done this a few times, and have seen others do the same, with 100% fake rate so far.

Some Ebay parts are sometimes genuine, some product groups even at high rates, but IR power FETs, sadly, are not one of these groups.

Step #1 should always be: use genuine components, so that you know they match the datasheet. Otherwise, doing any math, any simulations, any circuit design is completely useless, since all datasheet values are totally meaningless. It's a complete guess game, only based on random experiments, no matter how experienced you are, or how willing to learn.

Step #2 is exactly what you have described: learning to understand how to design and measure these things, so that you don't need to resort to the "my last hope, totally random stupid things" like suspecting your component source .

Beginners need usable tools in order to learn instead of drowning in frustration. They don't need to be top-notch or expensive, but many Ebay semiconductors are totally out of question. They are not even cheaper, although when you buy from proper distributors, you need some planning with the postage fees.

[Buriedcode]

I didn't say it wasn't possible, or a factor, just one I don't think first off.  If I knock up a circuit and power it without checking anything first, and something pops, I don't immediately think "fake parts!".  Especially one that relies on oscillation from a possible unknown transformer to work and not draw a ton a current.

I can't say I've been stung by fake MOSFETS but I have no doubt  they are about.. it was more that people seemed to just skip over the large number of unknowns here.  Hell, people started going on about gate charge without knowing exactly what circuit this is going into, or even what its application is - Saying ZVS and posting a link to another schem doesn't cut it.

I wasn't trying to put off any beginner by suggesting one needs lots of expensive equipment, just that there is a reason these things are difficult to get working just by 'winging it'.  Changing parts willy-nilly because a part failed doesn't really determine why it failed, what the circuit is doing etc..

So, again, to the OP, post a schematic of your setup.  And as much info about the inductor, diodes, the transformer (I'm assuming you're building a ZVS driver for a TV/CRT flyback) as you can.  And if you're driving the MOSFETs with a function generator, then how is it a ZVS?

Edit: just re-read, and I probably sound like a dick.. just to be clear, I'm not saying anyone is wrong here.

[Neukyhm]

The attached picture is my circuit so far, simulated with OrCAD. I'm not using a TV flyback, I'm using a big ETD59 N87 core with 1.5mm gap, I know that it's not saturating.

The capacitor bank is also some big and expensive TDK capacitors rated for 2kV at high frequency.

I have to say that I first suspected that the MOSFETs were fake because the part number and International Rectifier's logo on the front of them was not as presented in the datasheet. Yesterday I went to a trustworthy shop here and bought more (and genuine) 250, they look different from the others bought on eBay.

Diodes are UF4007

I will try the new IRFP250 this weekend, if they die too because of peak voltages maybe I should use the IRFP350 instead.

Edit:
The IRFP460B (with a "B") seems better than the 350.