"patrick-turner" wrote in message
...
On Wednesday, 7 November 2012 10:33:50 UTC+11, Alex Pogossov wrote:
The recent thread about the SET 300B based amplifier got me to question:
Why triode? 1) Obviously, a triode 300B boasts low distortion say 2% at
6...8W output and reasonable damping WITHOUT GNFB, but the penalties a
low plate efficiency, complexity of filament supply, cost, possible
self-destruction if shorted load, full volume and fixed bias.
The use of 300B satisfies those who insist only real triodes in pure class A
should handle their ear input signal. All later developments in audio amps
after 1930 were tainted by economic considerations and silly arguments about
efficiency. Some folks still buy what other laugh at.
2) A common pentode / beam tetrode class A1 SE (say, two 6L6 or EL34 in
parallel) design needs GNFB to achieve the same benchmarks, but more
efficient, cheap and will not blow the tubes. A penalty here is probably the
GNFB loop which encompasses the OPT and would require shelving or other
elaborate compensation, which only Mr Turner can properly master to achieve
stability over the wide range of the load impedance.
I'm not the only expert on tailoring open loop gain with shelves while
keeping BW wide as possible into pure R loads and while allowing any pure C
load to be used without more than 3dB peaking in response above 25kHz.
People have been using GNFB around pentode amps since they were invented.
But the pentode gives a spray of multiple H in its THD, and RDH4 compares a
6F6 with 2A3 if you wish to know the typical differences. GNFB around OPT
and pentodes reduces all H products, and about 20dB is needed so that a
typical SEP giving 12% THD at -1dB is reduced to about 1.4%. Many people
would rather have 4W from a 2A3 than the same 4W from a 6V6, and not have to
use so much FB, if any.
3) Yet there is another possibility -- use pentodes (or beam tetrodes) with
local feedback. In effect the goal here is to reduce the gain (from G1 to
plate) of the stage to about the typical mu of a rival triode. In other
words, replace the internal electric field local NFB in a triode by an
external local feedback around a pentode. One way is to use cathode feedback
with the cathode winding having about 1/4 of number of turns of the plate
winding. A drawback here is a special transformer to be made. Another
simpler method is to arrange a resistive local feedback. Throw say R1=470K
from the output tube plate to grid, through a DC blocking cap of course.
Feed the input signal to the grid via a series R2=150K resistor. Thus the
output stage will resemble an inverting op-amp, the gain of which is defined
by the ratio of R2 / R1. This is because an inherent gain of a pentode from
grid to plate is high. Say for two EL34s in parallel it can be up to 50.
(More practical to throw R1 from the output tube plate to the driver tube
plate without a DC blocking cap.) Thus the local feedback can reach
15...20dB. As a result: a) Drive voltage will be comparable to 300B; b)
Distortion will be same if not lower than with 300B; c) The local NFB will
be unconditionally stable, since it does not include OPT; d) Output
resistance referred to plate (damping) will be (1/Gm) * ((R2 / R1) + 1).
With a paig of hi-gm tubes as EL34s it will be even lower than with 300B; e)
Will yield higher efficiency with the same DC input power; f) No real need
for resonance damping circuits across OPT primary; g) No GNFB will be
needed; h) Not prone to self destruction with shorted load. A drawback
though is a low input impedance (about R2). However, with a cathode follower
in the driver stage it will not be an issue. I am wondering if such
"op-amped" pentode stages are common. What am I missing? Regards, Alex
I have made 30W+ SEP amps with both 4 x 6AC7 and 1 x 13E1 and they are fully
detailed at
http://www.turneraudio.com.au/se35cfb-monobloc.htm and
http://www.turneraudio.com.au/monobl...12version.html
These amps employ local NFB in OP stages based on the Acoustical Connection
where a substantial % of OPT primary turns are used in the cathode circuit.
I also like to sometimes use some screen signal, ie, the tubes have
operating voltages conforming to Ultralinear, so that odd no H are reduced
to be similar to triode and then the reasonably high gain of UL still allows
the CFB windings to then reduce OP tube gain to about 3.3, so that to make
200Vrms Va to Vk, some 60Vrms must be applied to the OP grids. But where
there is a high amount of CFB, say 33% as I did with 13E1, and if g2 is
taken to a fixed Vdc, then you have UL action because of the signal between
screen and cathode.
Then I find the OP stage works like triode but with lower Ra and THD and the
driver can be made to be linear as is convenient to do if you have a choke
dc feed to driver anode, and you still get 2H cancelling. Then GNFB can be
quite low, but total GNFB + local voltage NFB is about 20dB, but the result
in THD is better than if the tubes were set up with cathode AND screen both
bypassed to 0V and there was only 20dB GNFB. I don't have to worry about my
ideas being stolen or copied, I'm a just a fukken nobody who promotes fancy
ideas that are damn difficult to implemement, and which all cause apolexy in
accounting and tranny winding departments of amp making companies, even in
China. Some ppl have tried making made amps using my ideas, but most give
up. And those that get them working without FB then struggle with stability
caused by them having a different OPT to what I've used, or they change
other things and they just NEVER understand 1,001 basic issues such as
motorboating, or that their amp is a band pass filter around which one uses
NFB. There is much that cannot be fully explained, and blokes have got to
realise that if they **** about with amps they end up with oscillators if
they just don't get it fully. No wonder SET triode amps are popular with the
largely ignorant DIY Brigade of the Unskilled.
Alex:
I was aware of your cathode feedback design. It requires a specially made
transformer. What I am wondering, do people often use output tubes as
"inverting op-amp" with resistive feedback from plate to grid to make a
rather deep and always stable local NFB around the output stage?