I need to protect a MOSFET bi-directional switch made with 2 of IRF540 (VDS 100V) against potentially over-voltages on inductive loads. enter image description here

Because isolator is slow (say ~1mS), I am concerning about the moment when SW just opens or closes into inductive load (a motor).

The plan is to put some uni-directional TVS across mosfets with VBR < VDSS.

Should be bi-directional (because load can generate negative spikes)? Should I connect the common? My guess is yes because in case of a event, I don't want to let the 2'nd mosfet to handle that too.

Any comments really appreciated!


2 Answers 2


Should I connect the common? My guess is yes because in case of a event

You don't need to connect to common because one of the MOSFETs will be reversed biased and its body diode will be therefore forward biased.

Because isolator is slow (say ~1mS), I am concerning about the moment when SW just opens or closes into inductive load (a motor)

Well, when activating, one of the MOSFETs is going to be vulnerable and the other will conduct through its body diode and be less of a problem but you do need to check that the body diode can handle the same current as the load when passing through the off-to-on situation.

By the looks of it the IRF540 is going to be OK in that respect.

But generally, and putting numbers on it, if the load is 20 amps and this activates in 1 ms linearly, the voltage will fall from (say) 50 volts to 0 volts and the peak power will be about halfway i.e. 10 amps and 25 volts = 250 watts. You could use a simulator to get a more precise figure (and I recommend that you do) but, a short estimate is 250 watts for 1 ms.

Looking at figure 8 in the data sheet suggests that it will be OK at these levels but you have to put real numbers onto this graph that represent the true operating conditions (and not just my guesstimates): -

enter image description here

The plan is to put some uni-directional TVS across mosfets with VBR < VDSS

The TVS would need to be quite beafy in order to handle the back emf at the current being taken by the inductive load when it is disconnected. This might be 10 amps (for instance) and the energy stored in the inductive load might keep pushing that current into the TVS for a few tens of milliseconds. If the TVS voltage was 70 volts, that's a power of 700 watts sustained for several tens of milli seconds.

I'd probably consider one of these: -

enter image description here

It has a stand-off voltage of 60 volts at 2 uA (only you can decide if this is sufficient) and it clamps at 96.8 volts with a peak current of 52.7 amps. If you draw a line between 60 volts and 96.8 volts and moved along that line a linear proportion of the current you will need (possibly 10 amps) compared to 52.7 amps, you can estimate that at 10 amps, the TVS will clamp around 67 volts.

Again, these are my numbers and you need to use your own. Anyway, that's a peak power of 670 watts and well within the 5 kW rating but the next graph will help you understand how long it will survive: -

enter image description here

As you should be able to see, a 10 ms pulse of nearly 2 kW is on the limit so, providing your pulse isn't much in excess of 10 ms you should be OK. But you can parallel devices providing you used worst case limits but I would recommend using a simulator for this. I designed a triple-parallel TVS surge suppressor like this.

Do not use MOV devices as an alternative - the MOV will be called into action for each on-off cycle of the load and MOVs/varistors only have a limited number of surge cycles before they fail short-circuit. A TVS (of the appropriate voltage rating) is, more than likely, the most reliable solution but choose with care and choose a beafy one.


If the gate control is slow you risk thermal problems when the fets are not saturated yet. Switching loss will be too high for the fet.

The voltage transients from inductive loads can be caught but a snubber circuit, as often applied to relay coils or thyristors.

TVS diodes will most likely fail quickly for this purpose, they only take their rated power for microseconds. You'd be better of with varistors, and even then it's questionable.

  • \$\begingroup\$ good point with MOVs, thank you. One more thing, 1 MOV across entire SSR or put on each device (I guess one) \$\endgroup\$
    – yo3hcv
    Apr 28, 2020 at 8:43
  • 3
    \$\begingroup\$ No, please do not recommend MOVs for this application because they are not intended to be fly-back protection devices that are called-into-action regularly. TVS diodes are a far superior device to use. MOVs will fail because they have a limited number of surge cycles. \$\endgroup\$
    – Andy aka
    Apr 28, 2020 at 8:59
  • \$\begingroup\$ Good point, I recall that MOVs tend to degrade when stressed regularily. They do not recover. \$\endgroup\$
    – stowoda
    Apr 28, 2020 at 9:38
  • \$\begingroup\$ @Andyaka aisde a snubber mov's can be applied. You're right though, the maximum surge power of a mov is a single lifetime event. \$\endgroup\$
    – Jeroen3
    Apr 28, 2020 at 9:41

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.