[Amps] Negative feedback passive input tetrode

[Manfred Mornhinweg]

Bill,

> Has anyone see or used negative feedback to better obtain lower IMD
> in a passive (resistor terminated) input tetrode amplifier?
> 
> Any thoughts or references or experience?

I can offer only some thoughts. I have no experience with such negative 
feedback in tubes, nor references.

My thoughts are that direct resistive plate-grid feedback should work, 
but would incur in high loss. The idea is this: In a normal tetrode 
grid-drive scheme, you would resistively load the grid to ground, and 
directly apply drive power, maybe through a broadband transformer or an 
L or Pi network, if the tube requires more drive voltage than the 
exciter provides into 50 ohm. When adding negative feedback, part of the 
resistive loading to ground is replaced by "resistive loading to plate". 
This provides the negative feedback, but the resistor value must be much 
higher than the value of the resistor to ground it replaces. The factor 
by which it must be higher is

(grid RF voltage + plate RF voltage) / grid RF voltage

And the power dissipation in that resistor is higher by the same factor.

A typical tetrode might run with a plate RF voltage 20 times as high as 
the grid RF voltage. If we would simply replace a 50 ohm grid load 
resistor by a 1050 ohm feedback resistor, the amplifier would work, the 
input SWR would remain fine, the IMD should improve, but if the grid 
load resistor previously dissipated 30W, the feedback resistor will now 
dissipate 630W! That's not acceptable.

So this simplistic scheme isn't workable. What we can do, of course, is 
raise the value of the feedback resistor, use a series resistor between 
the exciter and grid, and add the proper loading resistor at the input, 
to provide 50 ohm. This might produce a workable amplifier, but the 
higher the resistance values are, the more likely become HF and VHF 
self-oscillations due to additional phase shifts between those 
resistances and the tube's and stray capacitances.

Many solid state amps solve this problem by using a feedback 
transformer. This keeps the feedback and gate circuitry at low 
impedance, and also keeps the dissipation in thefeedback resistors low 
enough. But the drain of a typical solid state amp carries only 9 to 35V 
RMS voltage. Even a drain-to-drain primary winding in a 50V, 
kilowatt-class amplifier only sees 70V RMS. It's reasonably easy to make 
a clean-performing broadband transformer for that voltage. But tubes 
work with roughly 2000V RMS on the plates! I don't see how to make a 
broadband transformer for that voltage, that behaves cleanly (negligible 
phase shift) over the entire spectrum at which the tubes have gain. 
Winding enough turns on a ferrite core to handle 2000V RMS would incur 
in too much interturn capacitance and too much wire length. So I don't 
think that such transformer-based feedback can be used with tubes.

Maybe one should look into using a capacitive voltage divider for 
negative feedback in a tube amp. Basically the tuning cap of the output 
Pi would be tapped close to ground. For example by placing a capacitor 
in series on the ground side, that has 10 times the capacitance required 
in the tuning cap on each band. The grid would be connected to that 
point by a low value resistor. Power loss would be modest. But it would 
require more complex bandswitching.

In any case, only a relatively small improvement in IMD should be 
expected. At most 10dB or so.

There are better options: The gain of a tetrode is easily controllable 
via the screen grid. And the IMD produced by tube amps comes mainly from 
amplitude distortion, not from phase distortion. So it is possible to 
apply linearization by means of two detectors (connected to drive input 
and antenna output), an error amplifier, and a high voltage amplifier 
that modulates the screen voltage. Such a scheme should be able to 
produce a dramatic improvement in IMD, like 40dB or more. It would be 
more complex, but would have very little power loss.

Also predistortion can be applied to earlier stages. SDRs like the 
Hermes, ANAN, Red Pitaya, with PowerSDR software, can do that easily.

A catch is that external distortion reduction, be it via screen grid 
modulation or via predistortion, is effective only over the bandwidth of 
the distortion correction system. A screen modulation system can 
probably be made to cover several hundred kHz, while PowerSDR at this 
time has only about 40kHz correction bandwidth. Outside that bandwidth 
the IMD becomes HIGHER than it is without such correction! So, to use 
such distortion correction systems, one must start with an amp that is 
plenty clean enough by itself at frequencies well away from the 
operating frequency. The correction systems only work for the IMD that's 
relatively close-in. Plate-to-grid feedback instead works at all 
frequencies, but is far less effective.

Enough thoughts for now...

Manfred


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