9
\$\begingroup\$

I am designing a pro audio-video product that runs from a nominal 24 V 5 A off-the-shelf computer "brick" that has worldwide certifications. My customer has very little tech savvy, so I want to protect my rather expensive PCB from someone connecting a higher-voltage supply that could destroy a Class D audio amp IC and numerous op amps.

The old-school version of the crowbar OVP circuit used an SCR or a TRIAC. However, I read that modern MOSFETs are far faster.

enter image description here

P-channel MOSFET Q1 in the circuit is a well-known reverse-polarity protection circuit, which works very well for me. The idea here is that when Z2 gets beyond it's knee and begins to conduct, a voltage will be applied to the gate of N-channel MOSFET Q2. When that voltage reaches, Q2's turn-on threshold, Q2 switches on and "crowbars" the excessive voltage to ground. Resettable PTC fuse F21 is guaranteed to trip at 10 A max. So if the sourced current exceeds that, F21 goes high-resistance until power is removed. Ideally, the circuit is not damaged and comes up normally if a voltage below 29 VDC is applied.

I didn't have a power supply with the 10A capable of tripping F21, but I did have a 0 - 40 VDC 5 A bench supply. Slowly increasing the input voltage to the circuit, Q2 is off until I reach 29.4 VDC. Q2 abruptly comes on, and the bench power supply goes to it's current limit point - about 5.4 A. I know F21 isn't supposed to trip at this current level, but the PSMN6R7-40MLD FET that I selected, with it's max 6.7 mΩ RDS-on should only be dissipating 168 mW with that continuous current load. I figured even if the fuse doesn't open, any 100 watt-ish high voltage power supply would go into hard current limiting, and my board would be protected. Here is a data sheet link for the FET: https://assets.nexperia.com/documents/data-sheet/PSMN6R7-40MLD.pdf

Quite unexpectedly, Q2 fails under the test conditions just described. There was no package rupturing, but with power removed and the FET pulled from the circuit, Rds measures 0.09 or 0.10 Ω. Additionally, Rgs = 17.5 ohms, indicating catastrophic gate damage. 27 V Zener Z2 should have insured that Vgs never got over about 5 V, and absolute max for Vgs is +/-20 V, so I don't see how I could get have gotten gate punch-through. Does anyone have a theory on the mechanism for destruction of this MOSFET?

\$\endgroup\$
6
  • 3
    \$\begingroup\$ The gate-source voltage on Q2 in your scenario is only 2.5V, which is not enough to turn Q2 fully on. Since it's only partially on, your quoted RDS-on doesn't apply. \$\endgroup\$ Commented Aug 8, 2023 at 1:41
  • 2
    \$\begingroup\$ It seems like you're expecting your MOSFET to behave like a switch, with a sharp on/off characteristic. But there's nothing in your circuit which would drive that sort of behavior. Rds will vary over a wide range once you reach the Vgs threshold, and the MOSFET will dissipate a lot of power. A MOSFET is not at all like an SCR or TRIAC which really do "switch on" and stay on once triggered (as long as the holding current remains flowing). \$\endgroup\$
    – brhans
    Commented Aug 8, 2023 at 1:41
  • 3
    \$\begingroup\$ Your set setup probably makes the situation worse because as your PSU enters current-limiting mode it'll reduce its voltage, which in turn reduces the MOSFET's Vgs - just to the point of equilibrium where Vgs is high enough to set the MOSFET's Rds just right to extract the maximum amount of current your PSU will provide - and dissipate all that power in the poor MOSFET. It might be rated for 40V & 50A - but not both at the same time. \$\endgroup\$
    – brhans
    Commented Aug 8, 2023 at 1:44
  • 1
    \$\begingroup\$ Could be your test set up. The way you're testing, by gradually increasing the input voltage, is going to put Q2 into a linear, high dissipative mode of operation. Q2 is not suddenly going to snap ON into it's low Rds region. \$\endgroup\$
    – SteveSh
    Commented Aug 8, 2023 at 1:52
  • 1
    \$\begingroup\$ You may need a separate comparator circuit to monitor for the OV condition, and then apply the needed voltage to Q2 to turn in on, hard. \$\endgroup\$
    – SteveSh
    Commented Aug 8, 2023 at 1:54

3 Answers 3

11
\$\begingroup\$

The old-school version of the crowbar OVP circuit used an SCR or a TRIAC. However, I read that modern MOSFETs are far faster.

I think you have missed the point of what a crowbar circuit does; it shorts the incoming supply and blows a fuse or causes something to momentarily disconnect.

  • A TRIAC can do this
  • A MOSFET can't
  • Your circuit is a glorified power Zener diode
  • The MOSFET will be in danger of thermal runaway

Why am I destroying MOSFETs in this overvoltage protection circuit?

MOSFET thermal runaway.

\$\endgroup\$
7
\$\begingroup\$

Your chosen overvolt circuit is only suitable for transient issues. Power dissipation goes sky high with linear mode making things worse. Why not switch off the DC to your expensive PCB when the input voltage exceeds a certain amount and turn on when the input volts is below the threshold? Use some hysteresis to ensure that your switching FETs don't go in the linear mode.

\$\endgroup\$
1
  • \$\begingroup\$ My only thought on that is, "What if the user plugs the DC jack directly into a wall socket? How high do you want to rate the blocking device, and how likely is it for your chosen rating to be exceeded?" I kinda like the idea of a reverse-polarity blocker (or a full-wave rectifier) for "normal" voltage, and a fuse and crowbar for anything beyond that, including reverse. One-time-use fuses are easy to get ridiculous voltage ratings for... \$\endgroup\$
    – AaronD
    Commented Aug 9, 2023 at 1:19
6
\$\begingroup\$

When your external power supply supplies current above the small amount that the zener conducts, the rest is passed by Q2. At that point, Q2's drain-source voltage is close to the input voltage (RDSON doesn't really come into this scenario), so its power is close to VIN*IIN -- this is what is blowing it up.

\$\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.