[Amps] SOTA Solid State amplifiers

[Manfred Mornhinweg]

John, and all,

> There is another group online on AM phone that does a lot of class E
> transmitter development, and they too are using plastic switchmode devices in
> new ways.

Using class E (or D, or F) on AM is easier than on SSB, because AM isn't 
sensitive to phase changes, as long as the whole AM signal shifts phase 
together. SSB instead is very sensitive to this. So AM can tolerate some of the 
AM-to-PM conversion that typically happens in MOSFETs, while with SSB it has to 
be carefully handled, to reduce it to a level low enough that doesn't cause trouble.

I have been in contact with some experimenters who don't realize that getting 
very good envelope linearity is NOT good enough for SSB - and that in addition 
to it, good phase linearity is needed.

So, for a truly linear, multimode amplifier we can only partially apply the 
technologies used for AM.

Regarding plastic switchmode FETs, I'm placing all my bets on them, because of 
the huge price advantage. But it's good to know how to do it! Many experimenters 
start the wrong way, by choosing the biggest FETs they can find. And there are 
some switchmode FETs that can handle 600V, 50A, and 600W, while costing just 3 
bucks! The problem with these is that they are too slow. They might work well on 
160m, poorly on 80m, and be unacceptable on any higher band.

So, my experimenter friends, here comes a little bit of knowledge, which _is_ 
important for using cheap plastic switchmode FETs at RF:

- You must look for _small_, low power FETs, and use enough of them in parallel, 
to get the power you want. That's the only way to use them throughout the HF range.

- These should be high voltage, low current FETs.

- Select those that have low enough capacitances, specially the Miller 
capacitance needs to be low.

- Select those that have low enough gate resistance (not always specified, 
sometimes needs to be tested).

- Select only those that have a fast enough dV/dt rating.

- RdsON is uncritical. Generally the ones that have worse RdsON are the best 
ones for RF.

- Transconductance should be low, as far as possible, for linear modes.

The things that limit high frequency response of such a FET are:

- Source lead inductance. To minimize it, use FETs in the smaller plastic 
encapsulations, and connect them extremely close to the body. Using many FETs in 
parallel means having many source leads in parallel, reducing total source 
inductance. A practical rule is that to operate up to 30MHz, you shouldn't have 
much more than 1 ampere of RF current in each TO-220 source lead. And that 
requires a high voltage supply, to get lots of power from a reasonable number of 
FETs. Higher than 50V, in any case.

- Capacitances and gate resistance. The gate-source capacitance must be pulled 
up and down at RF rate, and worse than that, the Miller capacitance has to be 
pulled up and down by the whole drain swing, through the gate. This requires a 
lot of gate current. FET gates are _not_ high impedance, at RF! And the gates 
have an internal resistance, that forms a low pass filter together with the 
capacitances. Don't try to operate a broadband amplifier beyond this cutoff 
frequency! A resonant, narrow band amplifer can work a bit above that cutoff 
frequency, but gate drive becomes hard and lossy.

- FET slew rate (dV/dT). Each FET has a limit as how fast it can slew. It will 
limit the slew rate, by turning partially on even when the gate is at low 
voltage, through several mechanisms. At some high enough frequency, instead of a 
sine wave on the drain you start getting a triangle wave, and the efficiency 
drops. This is a big limitation, and most switchmode FETs have trouble getting 
to 30MHz, due to this. Sometimes it's a good idea to run a 600V FET at 100V, to 
get one that is fast enough to slew through the whole swing withing the allowed 
time at 30MHz.

I don't yet fully understand myself these FET slew rate limitations, so I would 
be grateful for any insight that exceeds what can be found in data sheets and 
application notes on the web.

- RF circuits must be designed to never apply negative drain voltage, because 
that biases the FET's internal diode on, and those diodes are always too slow to 
turn back off in time, at RF rates.

-

There you have it. Now start experimenting! ;-)

Manfred

========================
Visit my hobby homepage!
http://ludens.cl
========================