| davygrvy |
I was wondering if you guys could give me your technical opinions on this topology idea I have for a hybrid cathode driven triode amplifier. I plan on putting it together in a couple weeks.
Each half of the balance gain stage is essentially a little opamp with a DC shift. The jFET apparently lowers the noise generated at the cathode compared to a resistor. M3,4 maintains a constant voltage on the plate to reduce supply interaction and to allow the PMOS to run at a higher voltage. open-loop is about 75dB, but gain is set to 34dB. 40dB total when you include both sides. Tube draws ~500uA. The 12AX7A is just a simulation place-holder. I'll use the J/J ECC803S in the actual first off. The inter-stage transformer cancels 2nd harmonic quite well in the simulation and should negate any supply interaction because of the balanced feed. It also doubles the voltage on the drive side with plenty of headroom left.
This is the part where I've gone off the deep-end into no-mans land. What if I drove the cathode and used a tertiary winding (or a separate transformer) for feedback to the grids? SPICE says yes, it'll work. I might use 211s or 572s instead of the discontinued Svetlana tubes as shown.
From that shown, I can simulate 140W into the load with ~12dB of feedback and a 575Vpp cathode-to-cathode drive. Cathode MOSFETS are in series for voltage protection and should dissipate 18W each.
V2 is -100V to limit how far the grids can go positive (with respect to the cathodes). |
|
|
| Wavebourn |
I would cut a half of the 1'st picture: do you need numbers in specs to brag loosing even harmonics? With transformers you are getting already balanced ins/outs, what's the point in balanced amplification?
Also, what's for you use extra source followers in cathodes? Just to shift levels? Some positive bias on grids I believe will give more linear and less frequency dependent results.
About a second picture, why cathode driven? |
|
|
| kenpeter |
The nice thing about a triode is that can be linear without
GNF... And it hits the rails rather gently... To do so seems
to prefer a constant plate current, rather than a constant
plate voltage.... Else you will see the transfer curve of the
triode impressed on the load...
When you throw in sandy gain without a plan, it gets non-
linear real fast, and you end up needing a feedback loop.
Excessive global gain also pushes clipping hard to the rail.
In short, where does this circuit benefit by having a triode?
If it were a Pentode, would it sound any diff? If it were a
depletion mode FET would it sound any diff? The Triode's
say in all this (in your first circuit) is completely overruled
by the feedback loop...
The second circuit, I'm not entirely sure I understand yet.
It may or not benefit from gratuitous tubage... Hard to tell.
The voltages are certainly scary.... |
|
|
| kenpeter |
| quote: | Originally posted by Wavebourn
Also, what's for you use extra source followers in cathodes? |
Looks to me like his GNF loop drives the Cathode
from underneath... Thats not a problem I expect.
There are more important flaws what bothers me.
Too much loop gain. Too much GNF. Triode in its
least linear mode... More than one transfer curve
of distortion cumulatively impressed upon the loop. |
|
|
| Wavebourn |
| quote: | Originally posted by kenpeter
Looks to me like his GNF loop drives the Cathode
from underneath... Thats not what bothers me. |
The question was, "What's for"?
I guess, a tube in both pictures serves as a fast linear amp with a good gain that linearizes FETs. I am guilty: I use similar approach, but my FETs work really hard, like horses... Also, I use my tubes on more linear parts of curves..
|
|
|
| kenpeter |
IF: R6 and R7 were reduced to 10K and 150ohms...
Would R15 then be unnecessary?
The same could be said for the flip side of the amp. |
|
|
| kenpeter |
| quote: | Originally posted by Wavebourn
I guess, a tube in both pictures serves as a fast linear amp with a good gain that linearizes FETs. |
I don't see the tube (Circuit 1) linearizing anything.
GNF is the only linearizing element.
As for Circuit #2, I am still undecided...
I think this guy needs to draw various load lines
on a set of real triode curves until he "gets it".
There is a place (music) for linearity, and a place
(clipping) for creative non-linearity. The triode
gives us both in very good proportions. Other
abuses without solid reason are suspect.
I support sand 99.995%. But you can't do sand
right until you first understand what the Triode
has to teach us about good sound.
The second thing to learn is what a square law
device (JFET or Beam Power Tube) has to teach...
Perhaps the third thing to learn is what excessive
global gain, poorly restrained by excessive global
feedback has to teach us about bad sound...
He won't get anywhere by making a really big
op-Amp and sticking a tube in just for giggles..
Unless he's just trying to destroy tweeters. |
|
|
| Wavebourn |
| quote: | Originally posted by kenpeter
IF: R6 and R7 were reduced to 10K and 150ohms...
Would R15 then be unnecessary?
|
Both R15 and J1 may be eliminated for good. |
|
|
| Wavebourn |
| quote: | Originally posted by kenpeter
I don't see the tube (Circuit 1) linearizing anything.
GNF is the only linearizing element.
|
Yes, but if most of the gain is provided by fast linear tube GNFB linearization may be much better than when the same GNFB applied to slow dirty sand only. The whole generation of engineers grew up on the idea, "No matter how dirty it is inside, a GNFB will linearize everything". They laugh when hear about power amplification, they amplify a voltage only assuming that GNFB reduces output resistances and increases input resistances such a way so their non-linearities as if don't exist!
By the way, I recently played with such an idea, but without cascode: one triode and one P-type FET in common source, with feedback to cathode from drain. Nice results on low enough frequencies and high enough currents!
I needed a good fast 10W driver. Ended with a tube triode and source follower. |
|
|
| kenpeter |
R15 might stand in for a load in his H-Bridge? No center tap.
I don't understand his choice of output transformer either... |
|
|
| kenpeter |
| quote: | Originally posted by Wavebourn
one P-type FET in common source, with feedback to cathode from drain. Nice results on low enough frequencies and high enough currents! |
Yeah, local feedback from drain is like FETron, a triode emulation.
Good set of "plate" curves (with Linear Mu) for the same reason.
I can't find the schematic on my drive, I may have offloaded to
a memory stick... It was posted (by someone else) over in Solid
State in my thread about "triode emulation 100% sand". FETron
was far simpler to implement than an overzealous schematic I
had drawn to do the same thing. I give FETron a thumbs up!
You used that as your output stage? With a real triode and FET
follower as the driver? |
|
|
| Wavebourn |
| quote: | Originally posted by kenpeter
You used that as your output stage? With a real triode and FET
follower as the driver? |
As a class A driver in A+C amp. But the FET follower was a bit complicated: stable voltage and stable current (i.e. one more source follower on top, and counter-modulated CCS on bottom)
But it may be off-topic... |
|
|
| kenpeter |
FETron is sort of an unfortunate name for the cicuit described.
As it seems there was once a real part by that name, a simple
high votage depletion mode JFET that plugged in like a tube.
Not quite the same thing as the circuit. But hey, I didn't name
either the part or the circuit... |
|
|
| davygrvy |
| quote: | Originally posted by Wavebourn
I would cut a half of the 1'st picture: do you need numbers in specs to brag loosing even harmonics? With transformers you are getting already balanced ins/outs, what's the point in balanced amplification? |
Supply rejection, 2ndH cancellation and 6dB more level. The other half of the dual triode would be unused. Actually, headroom is the real answer. It takes a lot of drive voltage to move the output triodes with a Mu of 3
| quote: | | Also, what's for you use extra source followers in cathodes? Just to shift levels? Some positive bias on grids I believe will give more linear and less frequency dependent results. |
Yeah, that wasn't needed. Silly me.. a DC shift in a feedback point made an obvious bad result I didn't see at first.
| quote: | | About a second picture, why cathode driven? |
Why not? is the best reply. With cathode drive the work the PMOS guys do adds to the output power.
|
|
|
|
