Rock and Flame: Wavebourn Barracuda

[Wavebourn]

Well;
I decided to reveal some secrets to the fellow community.

My previous "Barracuda" used 6N6P on IRF730 steroids giving 30W of pristine clean power. It worked in class A+C with deep feedback.

The current one uses vintage 12L6GT tubes on STW15NA50 steroids and gives 100W of very clean tube sound, in class AB.

Each 12L6 has 100 Ohm + 100 Ohm in series in cathode.
From the point of their connection goes one more 100 Ohm resistor to the gate of MOSFET with 6.2 Ohm resistor in the source. I.e. the tube sets 8V on cathode, so MOSFET sees 4V on it's gate (too much, especially for the hot MOSFET, right?)
Now, attention!

Excess voltage on gate is sucked out by one transistor current source: emitter to ground, collector to the point of connection of all 3 100 Ohm resistors. Base shunted on AC by 20 uF capacitor through 10K resistor is connected to the source. This servo keeps current through the MOSFET stable. Against 100 Ohm it's dynamic resistance is huge.

Now, the question: why sound is so pristine clean?

It seems to me that the effect is very similar to well known "Ultra Linear", but in my case anode voltage is modulated by output instead of screen grid one being modulated like in "Ultra Linear".

Also, I can answer George's question about difference in THD curves BJT VS MOSFET: MOSFET has higher transconductance, so "sneaky ultralinear feedback to plate" (let's call it SULFB ) is deeper.

Here is the photo:

[Cauhtemoc]

Originally posted by Wavebourn
[B]Also, I can answer George's question about difference in THD curves BJT VS MOSFET: MOSFET has higher transconductance, so "sneaky ultralinear feedback to plate" (let's call it SULFB ) is deeper.

Isn't the general rule that BJTs has higher transconductance than MOSFETs?

[Wavebourn]

Originally posted by Cauhtemoc


Isn't the general rule that BJTs has higher transconductance than MOSFETs?

My general rule is "Everything has to be calculate in each particular case".

If resistor value in cathode is 100 Ohm 1 mA variation of cathode current leads to 0.1V variation of voltage. If transconductance of the MOSFET George used is 4A / Volt it means current amplification factor is 4000/10=400
I doubt his BJT had beta close to 400.

By the way, right now paralleling MOSFETs I obtained 300W of clean power from the same tubes. Looks like I've found a golden mine... Poor output tranny is hot, also much better heatsink is needed...

[kenpeter]

I am assuming the 12L6 must be Triode strapped?
Sneaky 4th circuit feedback to linearize the FET.
Is this schematic reasonably close to correct?

[Wavebourn]

Originally posted by kenpeter
I am assuming the 12L6 must be Triode strapped?
Sneaky 4th circuit feedback to linearize the FET.
Is this schematic reasonably close to correct?

Yep, except G2 is powered from stiff 120V source.

I did not measure output resistance yet, but I expect it to be very linear and quite low...

[kenpeter]

If G2 isn't tied to the plate, I don't understand how plate
feedback to the cathode is possible? A fixed voltage at G2
defeats that sneaky mechanism.

4th circuit doesn't normally work on Tetrodes or Pentodes
for the same reason it wouldn't normally work on FETs or
BJTs (aside from triode strapped or triode faithful emulations).

The constant current characteristic of the plate allows no
intrinsic feedback (or at least very little).

Looks like you simply have a follower driving a MOSFET.
Why again was G2 tied to 120V?

[Michael Koster]

I don't think we're seeing the entire picture here.

It looks from Ken's schematic like a cathode current mirror/multiplier. The signal current ratio is 100/(6.2 + 1/Gfs) or thereabouts.

The actual amp is push-pull from the photo and comments.

It doesn't say there's no NFB...

I can imagine this working as a current-boosted PP pentode amp with loop feedback.

Michael

Very cool, Anatoliy! I have also been thinking about simpler circuits.

[Wavebourn]

Originally posted by kenpeter


The constant current characteristic of the plate allows no
intrinsic feedback (or at least very little).

Do you remember recent discussion about plate driven tubes?

Look at pentode or ray tetrode curves: they are not parallel.

Looks like you simply have a follower driving a MOSFET.

Also, pentode drives a load, and MOSFET delivers feedback to it's plate.

Why again was G2 tied to 120V?

Do you remember Mr. Miller.
Imagine a tube with huge transconductabce...

Originally posted by Michael Koster
I don't think we're seeing the entire picture here.

Agree. There are many factors involved.
Except "Anode feedback" due to enormous transconductance of MOSFET much less of grid voltage variation is needed. Also, transconductance of MOSFET goes up with current. I think all that factors summing give such unbelievable results... But the whole picture needs to be observed more thoroughly.


[quote]
It doesn't say there's no NFB...
[/auote]
When the picture was taken I already had GNFB and phase splitter. Speaking of "No GNFB" I meant my first experiment, with output stage only driven my backwards connected output tranny.

Very cool, Anatoliy! I have also been thinking about simpler circuits.

Thanks Michael! Are you going to try them?

[Michael Koster]

Anatoliy,

Is this a way to think about the composite load line of tube+mosfet compound device?

"Imagine a tube with huge transconductance..."

Michael

[Wavebourn]

Originally posted by Michael Koster
Anatoliy,

Is this a way to think about the composite load line of tube+mosfet compound device?

Load on HUGE resistance, higher with more current exponentially.

Edit: also, 6L6 curves have to be scaled if to use for exact voltages/current I used for a different tubes.

Edit again: I will take curves later, now have to go on a folk festival (unfortunately I can't show this amp there!)