Out of spec mosfets?

Hello, I just wanted to solicit anyone who may have some more experience dealing with mosfets. My current situation is this: I've designed and build a pcb that uses the FQP30N06L to turn on and off some 100w cob leds. I typically run these leds at 50w, and they are at roughly 50v. There is a teensy that controls the pwm to the fets (so 3.3v). The FQP30N06L is rated for 79w (with cooling), 60v, 30a, and works with 3.3v boards (also, my pwm frequency is 244 hz). I've got a fan over them, but no heatsinks due to space constraints. The FQP30N06L typically get pretty warm, my thermocouple measures 90-100 F which is in spec (175C max junction temp). I've ordered them from sparkfun in the past and always received the Fairchild branded ones.

Recently, I ordered more FQP30N06L from amazon since the shipping from sparkfun would be more expensive, and found they were no longer manufactured by Fairchild semiconductors because they have been bought by ON semiconductor. The ones from amazon immediately shorted on and burnt up. I thought maybe I had received some chinese ripoff product, so i ordered from mouser... same problem! I should note, however, that the ON FQP30N06L seem to work fine at 12v to turn some relays on and off (so basically, they only fail at higher powers).

I also ordered some IPP50N10S3L-16 along side the FQP30N06L and they seem to work fine! This is bonkers.

Has anyone else had experience with these mosfets or getting counterfeits that don't meet the data specifications? They all look identical, except the old ones have the Fairchild logo instead of ON.

I called mouser and they will be replacing the failed FQP30N06L with IPP084N06L3 G which is very similar but has a much lower RDs on of 8 mOhms instead of 35 mOhms, but otherwise similar specifications.

Here are pictures of the new FQP30N06L that don't work, and the old FQP30N06L by fairchild that have been running on my board for a while. Pardon the quality, my phone doesn't pick up the writing that well.

here are the relevant data sheets

FQP30N06L (2001 data sheet)

FQP30N06L (2013 data sheet)
https://www.mouser.com/datasheet/2/308/FQP30N06L-1306227.pdf

IPP50N10S3L-16
https://www.mouser.com/datasheet/2/196/Infineon-IPP_B_I50N10S3L_16-DS-v01_01-en-785441.pdf

IPP084N06L3 G
https://www.mouser.com/datasheet/2/196/Infineon-IPP_B084N06L3-DS-v02_23-en-1227150.pdf

Just to confirm, turn on a so called bad transistor (no PWM), measure the drain to source voltage.

Show us the drive cct. (schematic) for the MOSFET.

As I read somewhere few month ago - many modern switching MOSFETs was "optimized" by reducing die size. As a result under high load they are much more sensitive to a switching time. If they stay too long in a linear mode - they blow up.

larryd:
Just to confirm, turn on a so called bad transistor (no PWM), measure the drain to source voltage.

Show us the drive cct. (schematic) for the MOSFET.

I'd like to measure it but it gets dangerously hot very quickly. I'll see what I can do though.

alesam:
As I read somewhere few month ago - many modern switching MOSFETs was "optimized" by reducing die size. As a result under high load they are much more sensitive to a switching time. If they stay too long in a linear mode - they blow up.

Yea I suppose I could reduce the pwm frequency even more to reduce the time spent between full on or full off, but 244 Hz is already ridiculously low. I typically would use 20kHz plus for motors. I would imagine I'll start to run into flickering issues if I go too low.

Yea I suppose I could reduce the pwm frequency even more to reduce the time spent between full on or full off, but 244 Hz is already ridiculously low. I typically would use 20kHz plus for motors. I would imagine I'll start to run into flickering issues if I go too low.

It's not about frequency it's about rising/falling time. Cold you check a signal slope on MOSFET gate?

Show us the drive cct. (schematic) for the MOSFET.

Try to measure gate threshold voltage of the fets.

You need a 9volt battery, a 10k resistor, and a breadboard.

Connect source to batt-.
Connect drain to the gate and to one end of the 10k resistor.
Connect the other end of the 10k resistor to batt+

Measure Vgs(th) across drain/source of the fet.
Should be <=2.5volt for a logic level fet.
3.3volt logic needs fets with a threshold <=1.5volt.
Leo…

This happens all the time in professional electronics. Some buyer will come to you and say that he can’t get the same part from the same manufacturer but he can get an identical part from an other supplier, and they are exactly the same thing so can he use them otherwise the production line is stopped.

So you check the data sheet and they look the same, but you have been caught out like that before so you ask for samples and try the new part. Even if it appears to work you must pass it through all the factory tests and do some good measurements on it before you say OK, otherwise you will get a bite in the bum.

The part is not out of spec it is just that the your circuit is badly designed and relies upon the part being towards the better end the spread of ranges given in the max - typical - min values on the data sheet. You should always design a circuit to work at the worst end of the spread.

You have the wrong MOSFET for the job entirely.

With a “roughly 50V” supply a 60V MOSFET is not good enough, go for at least 80V rated, 75 at a pinch.
If you are driving the gate with 3.3V you need a MOSFET rated to work at 3.3V gate drive - the FQP20N06L
is not rated for anything below 5V gate drive, and given the fact its not rated for 4.5V, but for 5V, I’d treat
that as an indication is not really good enough for 5V!

[ in fact, looking again at the datasheet, the typical plateau voltage is about 4V so its definitely
not going to work at 3.3V unless you select-on-test for parts with abnormally low threshhold
voltage ]

MarkT:
With a "roughly 50V" supply a 60V MOSFET is not good enough...

But OP is driving COB LEDs.
10 or 12 LEDs in series could act like a ~20volt zener diode for very low currents.
Meaning the drain will never see the full supply voltage.
Leo..

When it is off, it will see the full voltage.

When it has just switched off and any inductive elements are still ringing, it will see more than the supply voltage.

larryd:
Show us the drive cct. (schematic) for the MOSFET.

Imgur

I hope this is helpful, I’m not great at schematics. The pcb is really simple, I just drew it up in easy eda and had it printed so I wouldn’t have to wire everything. There are 18 mosfets total, but i only drew two on there… you get the idea though.

MarkT:
You have the wrong MOSFET for the job entirely.

I wonder if sparkfun cherry picks the lower threshold ones and just sells those, since they market them for 3v controllers. That might explain why the ones from them worked and others didn’t.

alesam:
It’s not about frequency it’s about rising/falling time. Cold you check a signal slope on MOSFET gate?

I wouldn’t know how to check that and probably can’t. Wouldn’t a higher frequency cause the mosfet to heat up more quickly if the rise/fall times are too high for it to fully turn on or off? I may be misunderstanding this.

Wawa:
Try to measure gate threshold voltage of the fets.

You need a 9volt battery, a 10k resistor, and a breadboard.

Connect source to batt-.
Connect drain to the gate and to one end of the 10k resistor.
Connect the other end of the 10k resistor to batt+

Measure Vgs(th) across drain/source of the fet.
Should be <=2.5volt for a logic level fet.
3.3volt logic needs fets with a threshold <=1.5volt.
Leo.

Thank you for the great suggestion. I was concerned about sticking my hands near 50v. I did this and with a 3s lipo at 11.2v and a 100k resistor (all i had around) and measured the voltage to be 1.510-1.513v with the original FQP30N06L from amazon, and the ones from mouser measured around 1.528v to 1.535v. Here are some pictures of the setup. Imgur: The magic of the Internet

MorganS:
When it is off, it will see the full voltage.

When it has just switched off and any inductive elements are still ringing, it will see more than the supply voltage.

MarkT:
60V MOSFET is not good enough, go for at least 80V rated, 75 at a pinch.

I guess I should have linked to the power supply for the leds. Here it is https://www.meanwell.com/Upload/PDF/LRS-350/LRS-350-SPEC.PDF
I think the takeaway here is that i probably need 80v+ mosfet? I’m not sure that I understand what “inductive elements” are going to cause the voltage to spike above the maximum regulated output of 52.8v. I don’t know much about electronics though, clearly.

Change gate to ground resistor to ~10k

Also between the gate and the 10k add a 220 ohm resistor.

It's not about frequency it's about rising/falling time.

Wouldn't a higher frequency cause the mosfet to heat up more quickly if the rise/fall times are too high for it to fully turn on or off? I may be misunderstanding this.

Yes you have misunderstood. Read the original quote again.

One way of blowing up MOSFETs is to switch too slowly. Then it spends a lot of time (microseconds) in the condition that wastes a lot of power. That heats up the MOSFET until it pops.

So a constant voltage supply.

I don't see any current regulation/limiting for the COB LEDs.
Leo..

I'm not sure that I understand what "inductive elements" are going to cause the voltage to spike above the maximum regulated output of 52.8v.

An inductive element is anything in the switching circuit that contains an inductor or coil, like a motor. Even simple wiring has some inductance.

Current flowing through a wire generates a magnetic field and when that current is turned off the magnetic field collapses. This means the magnetic lines of force cut through the wire or coil and act as a generator, producing a voltage / current in the opposite direction of the current that originally caused the magnetic field. The voltage this produces can be many times the original voltage used to produce the field in the first place. The normal way to suppress this and stop it causing damage is to put a reverse diode across the load.

snowskijunky:
Imgur: The magic of the Internet

I hope this is helpful, I'm not great at schematics. The pcb is really simple, I just drew it up in easy eda and had it printed so I wouldn't have to wire everything. There are 18 mosfets total, but i only drew two on there... you get the idea though.

I wonder if sparkfun cherry picks the lower threshold ones and just sells those, since they market them for 3v controllers. That might explain why the ones from them worked and others didn't.

No chance of that, its a commodity item.

I wouldn't know how to check that and probably can't. Wouldn't a higher frequency cause the mosfet to heat up more quickly if the rise/fall times are too high for it to fully turn on or off? I may be misunderstanding this.

Thank you for the great suggestion. I was concerned about sticking my hands near 50v. I did this and with a 3s lipo at 11.2v and a 100k resistor (all i had around) and measured the voltage to be 1.510-1.513v with the original FQP30N06L from amazon, and the ones from mouser measured around 1.528v to 1.535v. Here are some pictures of the setup. Imgur: The magic of the Internet

I guess I should have linked to the power supply for the leds. Here it is https://www.meanwell.com/Upload/PDF/LRS-350/LRS-350-SPEC.PDF
I think the takeaway here is that i probably need 80v+ mosfet? I'm not sure that I understand what "inductive elements" are going to cause the voltage to spike above the maximum regulated output of 52.8v. I don't know much about electronics though, clearly.

When switching high currents rapidly the stray inductance of the wiring can be enough to produce
significant voltage spikes (aka inductive ringing) - even without that using a device at 83% of its
absolute maximum voltage is asking for problems, there's so little margin of safety.

I've had a look at the Sparkfun site and they seem to sell FQP30N06L and RFP30N06LE under the same
page and these are different devices with different threshold voltages. Have you checked the actual
parts you have?

MarkT:
No chance of that, its a commodity item.When switching high currents rapidly the stray inductance of the wiring can be enough to produce
significant voltage spikes (aka inductive ringing) - even without that using a device at 83% of its
absolute maximum voltage is asking for problems, there's so little margin of safety.

I've had a look at the Sparkfun site and they seem to sell FQP30N06L and RFP30N06LE under the same
page and these are different devices with different threshold voltages. Have you checked the actual
parts you have?

Yes, if you look in the first post I have some pictures of them. Its a bit difficult to make out the writing in the picture, but they are definitely the FQP30N06L. Typically sparkfun tells you which one they are shipping at the moment.

larryd:
Change gate to ground resistor to ~10k

Also between the gate and the 10k add a 220 ohm resistor.

Adding 221 ohm resistor managed to make the FQP30N06L work!... at least it hasn't blown up yet. I don't have any 10ks though unfortunately, so its still a 100k pull down. The only slight issue is that there is still some inductance. When the other cob led that is run parallel is turned fully on, the off one turns on very slightly. I limited the wire length to 5ft, but thats about as short as I can go. This only occurs with the FQP30N06L. However, the IPP50N10S3L-16 does not have this issue with the 221 ohm resistor...this one is rated for 100v, though. I'm thinking that the IPP50N10S3L-16 is going to be the way to go since that will give me some margin for any overvoltages too.

I also found this rather informative video talking about gate resistors, for anyone else who might find this thread useful. Why We Must Always Use A Gate Resistor - N Channel Mosfets - Gate Loop Ringing (read show more) - YouTube

I have used 5-220 ohms in the past for similar problems.

Yes ringing can cause all kinds of problems.

Time to buy an oscilloscope so you can see what is happening :wink: .

Also, a 10k to GND will help discharge the gate input capacitance.