1
\$\begingroup\$

I am using the PMV20XNER as a simple switch. I am driving a load of about 8 Ω from a 24 V power supply. A schematic is below:

schematic

When I apply 3.3 V to the gate (which should be enough to turn this MOSFET fully on according to the Nexperia PMV20XNER datasheet), sometimes it works and I measure ~1.5 A through the load. However, seemingly randomly, sometimes the MOSFET begins smoking and it cracks into two pieces within seconds of applying 3.3 V. Othertimes, this part simply shorts between drain and source. I am at a bit of a loss as to why this is happening, but I have run through several theories.

  1. I am not fully turning the MOSFET on

  2. The temperature is raising due to high thermal resistance. According to this figure in the datasheet, I should have an RDSon of worst case about 25 mΩ. So the power dissipated in the device is 1.5A^2 * 25 mΩ = 56 mW. Even at worst case thermal resistance (~250 K/W), this leads to a temperature rise of ~15 degrees, which should be fine.

    Extract from Nexperia PMV20XNER datasheet

    Image source: Nexperia - PMV20XNER datasheet

  3. The Spirito Effect. I don't fully understand this, but my understanding is at low gate voltages, an increase in temperature causes a decrease in the threshold voltage, so individual cells have a lower threshold voltage -> higher current -> higher temperature and it forms a positive feedback loop until the cell dies. However, short of using another MOSFET, I am not really sure how I could avoid this.

Does anyone know which of these (or somethign else) it might be? I am thinking of trying either the PSMN2R0-30PL or the PSMN022-30PL instead of this, but I really want to understand what happened so I can avoid it in the future. Thank you!

EDIT: The load is nichrome wire, so is is slightly variable length and is between 8-20 Ω. Sorry about that, I was using the worst case example but I should have been more clear. The specific instances I was testing was where the load was 16 Ω, but it can be as low as 8.

\$\endgroup\$
8
  • \$\begingroup\$ sometimes it works and I measure ~1.5A through the load it should have been 3A unless it's 50%-PWM-driven. Is the gate drive pulse coming from GPIO a 50% PWM or just a constant 3.3V pulse? \$\endgroup\$ Nov 22, 2023 at 11:22
  • \$\begingroup\$ @RohatKılıç Sorry, I edited the post, I was unclear. \$\endgroup\$
    – Chris
    Nov 22, 2023 at 15:44
  • \$\begingroup\$ I can't see anything wrong with that as a schematic, so I'd be checking the layout, or for other mistakes. Double-check resistor values, pin orientations, the gate voltage, etc \$\endgroup\$
    – LordTeddy
    Nov 22, 2023 at 22:43
  • \$\begingroup\$ Did you configure the GPIO as output with really push/pull - or is there a pull-up resistor internally of e. g. 50 kOhm -> measure UGS = UR2, that would be then quite low like 1.5 V in such an example! \$\endgroup\$ Nov 22, 2023 at 23:57
  • \$\begingroup\$ Ah, how do you apply the 3.3 V - with a mechanical switch or by touching a cable to the right pin? Both cause contact bouncing - and your R2 of big 51 kOhm and the parasitive capacitance of Gate to source (big 1.15 nF! typical at 0 V UGS) gives an RC device with a time constant of quite big 51 us! So the UGS is falling probably too slowly from 3.3 V to 0 V --> in one moment the UGS is so low, that the UDS is 12 V and the current is 1.5 A = 18 W - and if it is bouncing several times - too much pulse power... transistor is cracking!! \$\endgroup\$ Nov 23, 2023 at 0:09

1 Answer 1

3
\$\begingroup\$

Now after some comments I am quite sure, that your problem is contact bouncing. While you touch the GPIO or even the gate of the MOSFET with a cable, you create very random contact bouncing. In reality the opened periods after the first contact can easily last between hundred microseconds to several milliseconds! And there can be several bounces, lasting up to around 50 ms until it settles (depends on switch type...) See here and here. Let's take figure 3 of latter link as an example of my explanations: Example of real world contact bouncing

  • The first high peak left is short - perhaps with R3 and the Cgs (between Gate and Source) it is so short, that it doesn't charge the gate to 3 V, perhaps only 0.8 V - so it does not switch on the MOSFET completely - perhaps even on, but in the worst case such, that ID is 0.75A and UDS 12 V = 9 W!
  • and during the long low period(s) afterwards (e. g. around 500 us after the first 3 high spikes) the Cgs discharges - no matter if it is charged with 1 or 3 V) and passes the worst point slowly. So the MOSFET is under high load for several microseconds. Here a picture how the Ugs can look like similarly with bounce (with R3 and Cgs and 3 V the scales should be adapted to 100 us/div and 0.6 V/div): Similar curves for Ugs due to contact bouncing (but other scales)
  • and as you pointed out correctly about the temperature coefficient, when the MOSFET heats up, the required UGS drops - so while the Ugs is dropping with the R2-Cgs time element, the heating up keeps the MOSFET even longer in the bad area with high power!

You wrote, that after switching on, it starts to smoke after a few seconds.

I do not know how MOSFETs behave at overload. But perhaps it destroys only a part of the MOSFET, but with time it starts to destroy the whole one. <-- this is an assumption, and bad in such a scientific topic like electronics. (ESD can have similar effects to semiconductors, immediately ... month(s) later destroyed)

So destruction is even more probable with such a MOSFET/circuit that

  • switches off at very low Ugs tending to 0V
  • the slower dropping of Ugs with the RC around 0V
  • the even worse behavior when heated up!
  • the quite high parasite capacitance of ca. 1.15 nF (because MOSFET has low UgsOn!) together with R3 of 51 kOhm -> high time constant of ca. 52 us
  • the manual contact bouncing, creating unpredictable number of concentrated power pulses to the MOSFET - until it is perhaps (partially) destroyed.

Solution Idea: In your case I would switch the input voltage with the GPIO - by programming the processor - or use a proper debouncing circuit before the GPIO pin (e. g. retriggerable monoflop of e. g. 0.1 s) Then nothing cracks, burns ... anymore.

If the GPIO output is also configured in push and pull configuration, after the falling edge the Cgs is discharged through R2 (and a bit by R3) and Ugs drops in 1 us to ca. 1 V and in about 5 us to 0 (now with your cable and touching, it lasts ca. 52 us for dropping to 1 V and 250 us until it reaches 0 V Ugs - R3 and Cgs)

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.