I have a MOSFET consistently blowing up. 550V part, 230V in, buck configuration, transformer drive. The symptom is always the same: g-s measures short. What may be causing it? Is it gate drive overvoltage? Thanks
If with 230V you mean it's running off of a 230VAC directly rectified mains circuit then a 550V device is IMHO below marginal.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
It's running off 230V DC, there is no AC in sight....
Is there anything else running off the same 230V DC supply? It only takes one narrow switching spike to blow the gate oxide.
-- Bill Sloman, Nijmegen
-
It's a good question. There is a spark discharge generated by the same board. It sends electrical "tremors" everywhere (I have seen TDS3054 blink). I do not think that it causes the failure as I have been working on this circuit for a while and have never seen shorted gate before. The only recent change I can think of is using four (instead of two) MOSFETs (STP20NM50, if it matters) in parallel. One of them develops g-s short very quickly. I do not see Vg-s coming even close to (maximum) 20V. So I wonder what it takes to short gate oxide...
It only takes a wee spike and this can also come onto the gate via the gate-drain capacitance (Cgd). In such an environment it is best to have fast clamps on the gates. Diodes against +12V and -12V or something like that. If it's fairly slow stuff you might get away with a TVS but remember that they have a large capacitance.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
d
xt -
OK, my suspicion that gate is poked through is correct..... I do not understand why.... I have SMAJ12CA across g-s.
The MOSFETS (four of them) Cin is much larger
e
ed
xt -
Does each one have its own gate resistor? High-voltage mosfets love to oscillate at 15 to 25MHz RF when they're in the linear region. Especially mosfets in parallel. 100V drains swings are not uncommon during oscillation. This can happen during the 100ns, etc, of slow switching. RF oscillation can cause high voltages across the source bond-wire inductance, and expose a mosfet gate-source to damaging voltages. Solutions are to switch faster, or to add ferrite beads to the source pins, etc.
Those ought to do it. Are they connected with very short paths to the FETs?
[...]-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
ck
be
fied
text -
The MOSFET drive looks like Fig 2 here:
Please show a schematic or LTspice of this circuit. You could have some flyback issues in the driving circuit, you could have a high Voltage pulse getting in the circuit. A bidirection TVs diode across the GS could help..
Or, you could simply be over driving the gate !
-- Regards, Joerg
Where did you place the TVS in there?
-- Regards, Joerg
"gmail" domain blocked because of excessive spam. Use another domain or send PM.
Only bits and pieces of this thread are getting through here, so forgive the 'reconstruction'.
You reference a gate drive circuit from IR AN-950. You have driven two parrallel parts successfully in a buck circuit, but four parallel parts give problems.
First thing....a failure in one part requires pretty careful repair of gate drive circuitry in a parallel switch. Repeated failures (same position?) suggest possible undetected damage.
Second...the specific gate drive circuit chosen may be impractical for driving paralleled devices without modifications to anticipate the distributed gate loads, threshold voltages and individual Cgd switch effects.
Fets share only under saturated conditions, not dynamically, while switching with 'identical' gate voltages. The first part to turn on will tend to drive the others off through individual Cgd paths. The drain node will tend to be driven by the one device. Other fets in parallel will form gate threshold voltages after the Cdg discharge.
If the drive is pulse-formed and statically maintained, as in the circuit of AN-950, you may find that the gate voltage finally stored on the four devices is not as anticipated.
Suggest you back off and study reconfirm two-device circuitry (gate drive integrity, current and temperature balances etc), then add one extra device at a time, while monitoring effects of doing so. There will be additional requirements as the gate drive load becomes larger and more complex.
RL
i think RL might be right here. if your FET gate driving current is not strong enough, multiple FETs in parallel might would increase the total gate capacitance and slow down the switching. when the FET are in linear region, they don't share current too well, and one might take more than the other and just over heat.
what happens if you reduce the load current?
if you want to drive the FET across isolation, why not use a digital isolator and a driver? see fig 1 of:
tomrei
043.pdf
The circuit generates a spark. I haven't tackled CPU shutdowns caused by EMI yet (I have mentioned earlier that I have seen EMI-caused oscilloscope screen blinking). I do not want anything on this board that can latch. Some DC fans stall near this circuit (not that it matter much - amused me first time I saw it). No, no digital control... I want to retire but cannot afford it... ;o)
If the oscilloscope is in close proximity or attached to the active power circuit, it can be affected by noise generated there, paricularly if powered from the same AC line.
This interference could be a symptom of the failure mechanism, as it may be an indication of unusual di/dt, avalanching or oscillating mosfets, or severe reverse recovery in rectifiers. Layout or simple test interconnection can aggravate transmission of this noise.
Odd if switching is as moderate in speed as you have indicated. Are your probes/scope capable of conveying fast rise times? (Model and type number?)
RL
:
The probes and oscilloscope are fast enough for the application (Tek.... do not remember part # off the top of my head; 500MHz passive probe, 500MHz portable oscilloscope, the one with floppy drive). The circuit ignites spark gap (~10kV pulse that generates all the EMI) and then controls discharge current (~100kHz switching, EMI is negligible). EMI has never caused MOSFETs blow ups. The problems arose when I increased (was asked to) current (from 25 to 75A) and added few mosfets in parallel. I looked more carefully at g-s voltage yesterday (Fri). It was wandering outside of +-20V max. Oooops! I tried carefully adding resistors and beads togates. The swings were getting smaller, I left for the weekend. There are parts that look better (on paper): lower Rds-on, higher current, higher maximum Vg-s. I might try them.... Thanks! MK
Ground clips can pick up quite a bit.
That would triple any conducted/magnetic emmissions that were there before ;-)
You might want to think about small TVS devices from gate to source. Ok, they'll add capacitance but the FETs will probably have several thousand picofarads by themselves already.
Most likely also more Cgs.
-- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.