Bob,
I agree, and although the unltralinears do sound good, I think
that cathode feedback cures more sins of the output transformer,
and does not rob power like the screen grid ultralinear setups
seem to. it gets its extra voltage from the power supply and the
tubes handle the extra current when needed. I think a good test
of the effects are to pass a low frequency square wave.
Like ultralinear it also needs a separate winding, but only
capable of making a peak voltage equal to about half the peak
grid voltage. Like the screen winding scenario, the correction to
the cathode bucks the gain to a certain definite level when all
is perfect (top of the square wave is perfectly horizontal),
decreases it during time when the HF response through the
transformer is too great (top of the square wave slopes up during
the half cycle), and increases it during time when the LF
response is poor (top of the square wave slopes down during the
half cycle) such as when the transformer is deficient in iron.
Did I explain this correctly?
When the output waveform of the transformer does not match the
input to the tube grids, the negative feedback to the cathodes
corrects the instantaneous grid-cathode voltage with the goal of
straightening up the tube-transformer combination to where it
is very linear. Higher plate voltage helps this work better, and
more grid drive voltage is needed, and if grid current is drawn,
the driver must be exceptionally well regulated, and the power
supply needs to have the capacity for high peak currents (alot of
demands).
A triangular wave is good for judging this. The straighter the
lines of the triangle, the more linear the stage. I have a couple
articles on doing this with old PA amps here:
http://www.montagar.com/~patj/mi12188a.htm
http://www.montagar.com/~patj/aph1050.htm
(or at least the audio experimenters like them) Truthfully, with
this mod, you can actually hear the difference, it will clear up
a muddy old amp right quick. Maybe would be good for the speech
amp. It also gives the amp very good regulation.
On the APH-1050's, which have good iron, I was passing a 10Hz
squarewave through two of them (one unmodified, one modified),
and on the modded one, you could really hear the laminations go
'thunk' each time the polarity was reversed, and watch the plate
voltage waveform greatly distort in order to cause the current
through the transformer to remain true to the input signal. The
unmodded one came nowhere close to faithfully reproducing a
square at 10Hz. In the modded one, there was some ringing on the
leading edges, but tuning the overall feedback loop capacitor's
value minimized it. I had to ditch the 6L6's though and use
6CD6's so there would be plenty of current capability.
The only issue with filament type tubes and this method is having
to use a separate filament transformer for each tube in the PP pair.
But back to the 4-400's since there may be questions about how
the tubes would sound having the control grids driven along with
the screens, maybe the other option would be to connect the grids
to the center tap of the filament transformer for no drive, and
just drive the screens with audio, like a zero bias triode setup.
I have not decided exactly what to do for fidelity since I am set
up for 3-500Z's right now. I may have to see if a speech amp
modified for cathode feedback and greatly oversized will be
capable of good regulation and 3-4 times the required power will
do it, assuming I can place it inside an overall feedback loop
from the modulated HV supply to the speech amp input.
The original question came up as the 3-500 and 4-400 fit the same
sockets and use the same filament supply.
Patrick
-----------
From: "Bob Bruhns" <***@erols.com>
Subject: Re: [AMRadio] 3-500Z vs 4-400A
Hi Pat,
The 20K resistor idea was published in an old article on getting
120 watts from a pair of 807s, but I have been told it really
didn't produce that much power, it did some really odd things at
low audio levels, and it really didn't sound good.
The idea about tying the control grid and screen of a 4-400
together and using it as a triode might be better, but I suspect
that the 3-500 will be a better triode than a kludged 4-400.
Tying the screen of a 4-400 to the plate will *drastically*
reduce plate voltage limits, and plate current will be much lower
at acceptable grid potentials (instantaneous E-grid *below*
instantaneous E-plate) - so it would be impractical, because the
power output capability would be massively reduced. To do
ultra-linear right would require a separate, center-tapped
screen winding on the mod transformer.
I am in favor of the 4-400, but with active screen drive to
improve linearity. Some kind of circuit could take instantaneous
control-grid drive and produce a corrective pre-distortion to the
screen voltage, aimed at making the grid-plate transfer as linear
as possible, and making overload graceful.
With lower voltage tubes such as the 6L6, 807 and 8417, it is
actually possible to accomplish this with simple resistors in
series with the screens. But the big, higher voltage tubes have
so much secondary screen emission that such resistance can not be
tolerated. So with higher voltage tubes, a more complex active
screen control is required.
But when the smoke clears... you will have the output and gain of
a tetrode, but with the linearity of a triode (or better). With
decently designed feedback, you can have the source resistance of
a triode as well.
I've done it with 6L6s and 8417s. Some day maybe I'll try it on
4-400s or something.
Bacon, WA3WDR