Is paralleling MOSFETs a good or a bad idea?

Question

In my circuit making experience when faced with a problem wherein I need to switch a high current load with a MOSFET, my go-to solution has always been finding a better MOSFET that can handle the current and has low enough resistance in order for thermals to become not a problem.

I am exploring the option of using 2 (possibly more) same model MOSFETs in parallel driven by the same MOSFET driver to guarantee simultaneous turn on. Is this a good idea?

Paralleling MOSFETs in my opinion would solve two problems at the same time. Paralleling them means that I would get theoretically half the "on" resistance which also means that thermals would also be reduced.

A counter argument to this would be if the MOSFETs don't turn on at the same time, for a brief period of time one of the MOSFETs will be subjected to currents possibly higher than its rating. A solution to this is have both MOSFETs be driven by the same driver pin. This will slow the turn on/off time depending on the capabilities of the driver. A further solution is to choose a MOSFET that carries the full current in the first place, meaning you are just solving the thermal problem.

In my opinion I have a good idea, but what bothers me is I have not seen any circuits that do this. That implies that my thinking might be flawed.

What are your thoughts about this?

Answer 1

I see no problem with it. It is also best for them to share the same heat sink and be switched at a reasonable speed and proper gate voltage. If one starts conduction more than the others it will get warmer and its resistance will increase putting more load on the remaining ones by conducting less. They share very well this way. They will be prone to oscillation which you have taken care of with with low value gate resistors. Not knowing your driver circuit you need to be cautious of the Vgs voltage and be sure it stays in range probably in the +- 15 volt range, depending on the MOSFET you use. I have seen this done many times over the years. In the early 2000's this was relative common mainly with N-Channel MOSFETs because of price and Rds(on) of the P-Channel MOSFETs. N-Channel MOSFETs were at that time generally pre trench and much higher in RDS(on).

Answer 2

It is perfectly fine and frequently done.

Even if the devices didn't turn on (or off) simultaneously, the (possibly) higher current for a brief time won't cause any damage.

Note that it is better to use separate gate resistors (as you have) than to combine into one.

You show a 24 V supply -- if you pull the FETs' gate to ground you will have VGS of 24 V which will exceed the max. rating of commonly available FETs.

Answer 3

The PTC characteristics of most FETs make this a valid application for digital or fast dVgs/dt switching, but not slow switching in the linear high power region as commonly ignored due to parallel internal geometric and weak link hot spots creating faults. That is due to the variance in Vt thresholds and not the RdsOn at rated switching voltages.

Answer 4

It's an excellent idea. As for unequal delays, consider a system where the drives are mismatched by 10 nsec. Then on turnon, one FET will have twice the nominal current - for 10 nsec. Same for turnoff.

If you are so hammering your FETs that a 2x current for 10 nsec is going to damage it, you're in for heartache anyways.

Answer 5

Make sure the MOSFETs on your board are all identical (part number, vendor, lot #). In a high power application, I had two different MOSFETS qualified for use on the bill of materials (BOM). Under test at the highest power, two of the four paralleled MOSFETs blew up. After visual inspection, my technician discovered the two that blew up were different parts than the two MOSFETs on the other side of the board. I believe that pair switched on a bit sooner than the other two and couldn't handle max current. I needed to write assembly instructions such that boards were populated with the exact same MOSFETs in all locations (as well changing the design to use six MOSFETs to have adequate margin).

Answer 6

This depends on context. It can be inferred from the schematic that this is a board-level design involving digital switching of discrete VDMOS power FETs. In that context, parallel devices are quite common. "close enough" matching of device characteristics can usually be achieved in practice by using components purchased from a single distributor, and if your're lucky you'll get devices from a single batch possibly made from the same ingot though that is of course never a guarantee. Usually the goal is to try getting components with same date codes and assume they're made from a single production run, and it'll be "good enough".

In cases where that is insufficient, it's technically possible to do your own device characterization and binning to achieve acceptable levels of uniformity, though that can be impractical and expensive for large-scale production.

In simplified MOSFET models where parasitic effects are minimal approximations, parallel devices with otherwise identical model parameters can be considered equivalent to a single device whose channel widths are the sum of of the widths of the devices. You'll get the same performance assuming same self-heating effects are maintained.

Answer 7

Parallel MOSFETs are generally satisfactory in the conduction mode when fully on.

Separate gate resistors are better for switching. Separate gate drivers are better still. Things do get nasty in the linear mode.

Answer 8

If you plan on using the mosfets in a switch mode power supply or a turning them on and off very quickly, be aware of parallel oscillations between the two devices. This is very bad and will a lot of times fail the devices. This is why you'll find that ferrite beads are in series with each gate in some designs. When it happens, the gate will begin oscillating out of control and turning the mosfet on and off in the Mhz range. The frequency at which it oscillates will be dependent on your parasitics.