[Amps] Sad News for Eimac Users

[Manfred Mornhinweg]

Gene,

> I am puzzled as to why the solid-state devices seem to be more linear 
> than tubes ("fire-FET's"?).  As I understand it, FETs are square-law 
> devices that generate mostly low-order and second-order distortion 
> products, while tubes are 3/2 power-law devices that generate more and 
> higher order distortion products.  Does anyone out there with a better 
> comprehension than I have of the physics in this matter have an 
> explanation? 

I will try to elaborate on this, but instead of thinking in terms of 
square law and the order of distortion products, I prefer in this case 
thinking simply about the instant voltages and currents along the RF 
waveform. Not that the waveform is important, as a low pass filter will 
sinusoidalize it anyway, but for analysis purposes it's good to think 
about the RF waveform, because any distorsions there translate to 
distorsions of the envelope waveform, which is of course not corrected 
by any filter you can install between a power device and the antenna.

The base curves of output current (plate or drain) divided by control 
voltage (grid or gate), which you could measure at low frequency, are 
basically:

- For a triode: Severely nonlinear and very dependent on plate voltage

- For a tetrode: Quite linear except for a certain area of the curve, 
and not very dependent on plate voltage, but highly dependent on screen 
voltage

- For a pentode: Quite linear and quite independent on plate voltage, 
but highly dependent on screen voltage

- For a MOSFET: Severely nonlinear but quite independent on drain voltage

 From these characteristics it's clear how each device needs to be used. 
With a pentode, often no further linearizing is required. You just apply 
proper grid bias, a VERY STABLE screen voltage, some RF drive, and you 
get a reasonably clean output that can be put on the air directly. With 
a tetrode it's much the same, as long as the bias and drive are selected 
such that the nonlinear part of the curve is avoided.

With triodes and MOSFETs it's not that easy. These are simply too 
nonlinear to be used without "ironing flat" their curves. This is done 
with negative feedback, which lowers the gain but improves linearity. 
With triodes, the time-proven method to add a lot of negative feedback 
in a simple way is running them in grounded grid configuration. A very 
low gain results, typically just 10-15dB, often even less, along with an 
acceptable degree of linearity, but usually not a very great one.

RF power MOSFETs cannot be easily run in grounded gate, because of the 
extremely low input impedance and poor power gain that results. So the 
negative feedback has to be added in other ways: Source degeneration, 
which is already built into MOSFETs intended for linear service; Overall 
feedback from the drains to the gates, sometimes with frequency 
compensation; and sometimes more complex methods are used, such as 
inductive drain-source feedback (bootstrapping). These techniques result 
in good gain and linearity.

Modern MOSFETs have extremely high power gain, and this allows adding a 
lot of negative feedback and still having enough gain left over for 
practical use. It is this large amount of negative feedback that ends up 
giving well designed MOSFET amplifiers their high level of linearity.

It would be possible to make a very low distortion tube amplifier, by 
using a pentode in grounded cathode configuration, which has good 
linearity and very high power gain by itself, and then adding strong 
negative feedback to cut down the gain to 12dB or so. But it seems that 
nobody is doing this. Probably there would be stability problems, 
because tubes have such a poor ratio of output capacitance to plate load 
resistance, that there is no option but to use them with narrow band, 
moderately high Q, tuned impedance matching circuits, to absorb that 
capacitance. These circuits introduce enough phase shift to bring an 
amplifier with strong negative feedback to the point that at some 
frequencies the feedback might rotate into positive, and whops, you have 
an oscillator!

MOSFETs instead have a much better ratio of output capacitance to drain 
load resistance, allowing them to be used with broadband matching 
transformers on HF, or very low Q tuned circuits, on VHF and UHF. This 
makes them far more stable in feedback amps.

A particular linearity problem with MOSFETs that has to be addressed is 
that their internal capacitances vary very much with voltage. They are 
basically varicaps. Specially the output and reverse transfer 
capacitances vary strongly over the RF cycle, causing distortion. The 
higher the frequency, the worse this effect becomes. The designer of a 
solid state amplifier must be aware of this issue, and handle it 
properly. If he "thinks tubes" and assumes these capacitances are 
constant, an amplifier with poor linearity will result. Negative 
feedback helps a lot in reducing this form of distortion. Swamping with 
external capacitance also helps some. But in many cases it can be 
advisable to avoid driving the drains very close to the sources, because 
that's when the problem becomes worst. So this requires leaving some 
headroom between the power supply voltage and the RF voltage at the 
drains, and this costs efficiency. That's why a low distortion linear 
amp using MOSFETs often can't match the plate efficiency of a pentode 
amp with similiar distortion performance. It makes up for that by not 
needing any heater power.

Manfred

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