Parallel FETs

When I built the GW3UEP 630m transmitter, I made one small modification by adding a second FET in parallel with the first. Normally when dealing with RF devices (tubes, RF transistors) in non-switching modes, adding a second device in parallel with the first, always guarantees a significant increase in output power. This is not the case with FETs when parallelled in switch-mode amplifiers.

Second FET added to GW3UEP 630m transmitter

Adding the second FET reminded me of an interesting e-mail that I received last year from Minto, PA3BCA, who uses parallel FETs in all of his LF transmitters.

"Adding (parallelling) FETs in an existing TX configuration will indeed not lead to a large increase in power output. It is easy to explain why. The FET (in class D or E) acts as a simple on/off switch, with a little bit of series resistance (the Rds_on). And parallelling switches will increase the current (and thus Po) only slightly by halving the Rds_on, but this is negligible.

A calculation for instance with my PA: 50 Volts, 600 Watts.
Say the Rds_on of a FET=0.2 Ohm, so for a single FET the dissipated power is 70Watts. Now with two FETS in parallel the Rds_on is only .1 Ohm. Dissipated power is then only half that of the one-FET configuration, ie. 35 Watts. So 35 Watts less loss in the FETs. But even if all of the 35 Watts so gained becomes output, this is only a 5% gain in output (600 + 35 Watts). Not easy to see on amateur class output meters.

The big advantage of parallelling FETs therefore is something else. I did it to protect myself from sloppiness, clumsiness and stupidity. FETs are very easy to parallel. When heating up, the Rds_on increases so dissipated power is nicely distributed over the FETs. By parallelling the FETs, the total dissipated power halves (in this case from 70 to 35 Watts). This means that the heatsink can be smaller or stays cooler. The dissipated power per FET decreases by the number of FETs squared. With two FETs, the dissipated power in a FET is only 1/4 of the single FET configuration. In this case only 17.5 Watts instead of 70 Watts per FET.... 70 Watts dissipation per FET is too much for comfort, there is no headroom and if the temperature gets too high, the FET will die quickly.
And secondly, my PS delivers 20+ Amps when short circuited. A single IRFP360 that is already hot will then certainly die (for instance user error by applying forward bias to the gate). Two IRFP's in parallel can easily handle the 20 Amps.

So user error (no antenna, variometer way off, screwdriver dropping from fingers and connecting the gate to + 12 volts, had all this happen) until now has never resulted in a FET dying.

The flip side is that the Input capacitance and the reverse transfer capacitance also double. That is the reason I added an extra high-current totem pole to drive the FETs. The BC337 / BC227 just were not up to the task. I found it a cheap price to pay! "

In my particular case (30 Watts in / 25.5 Watts out), using a single FET resulted in 4.5 Watts of total dissipation while going to two FETs could result in a total dissipation of just over 2 Watts, or 1 Watt per FET. Even when keydown for long periods of time, the FETs are cold.

While parallelling FETs would certainly be more rewarding when applied to a high power switching amplifier, even small amplifiers can enjoy some benefits of the extra addition.