F6 Amplifier

Here is an F6 design that is fully symmetric in the spirit of a circlotron. The output power supplies float relative to ground, but not relative to each other. The harmonic distortion is primarily 3rd order. It will accept either balanced or unbalanced inputs.

I have simulated it with LTSpice and it looks very good.

The power spectrum plot in for 1 watt output at 1kHz into 8 ohms.
The best solution for the input until now.
And what about the use of a circlotron input stage?
Then the p-ch jFET is dispensable and I can use other types for the N-channel like the follow:
http://www.toshiba.com/taec/components2/Datasheet_Sync/53/6937.pdf
or even a Sziklai darlington (CFP) instead a single jFET (e. g. consisting of a BF245 and a 2N5401).
The disadvantage of a second power supply isn't an issue - so I think.
 
Last edited:
This thread is absolutely fascinating !!!
I have to go back and re-read a large portion to see if I can get my head around it.
I do know that I became intrigued with the "transformer interstage" when I built my F4 headphone amp (posts around 3400 in F4 thread) and I am now hooked.
Have the transformers and R100s - now will have to look for a chassis :)
Thanks to everyone
Back to reading
 
The only difference from a circlotron that I know about is the position of the power supply V1 relative to the output FET.

  • In my circuit, there is the clockwise loop: Rload - V1 - R4 - J3
  • In a circlotron the loop is: Rload - R4 - J3 - V1
All remains the same (except the bottom of the bias circuit remains attached to the bottom of R4).

An appropriate name for your circuit in post#1055 is: The L. H. Quam Circlotron. Clearly, the F6 Amplifier thread was, and continues to be a source of your inspiration.
 
Here is an F6 design that is fully symmetric in the spirit of a circlotron. The output power supplies float relative to ground, but not relative to each other. The harmonic distortion is primarily 3rd order. It will accept either balanced or unbalanced inputs.

I have simulated it with LTSpice and it looks very good.

The power spectrum plot in for 1 watt output at 1kHz into 8 ohms.
A note about post #1055. If the node Out- is grounded, then the circuit reverts to a non-floating version. Here is the difference in harmonic distortion floating vs. grounded PS, particurly the lowering of the 2nd harmonic:

Floating PS, 1kHz, 1 watt into 8 ohms closed-loop gain approx 5.3:

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 2.034e+00 1.000e+00 -0.50° 0.00°
2 2.000e+03 1.651e-05 8.116e-06 -170.34° -169.84°
3 3.000e+03 1.622e-04 7.974e-05 5.35° 5.85°
4 4.000e+03 8.365e-06 4.113e-06 179.94° 180.44°
5 5.000e+03 7.235e-06 3.557e-06 -179.05° -178.55°
6 6.000e+03 5.574e-06 2.740e-06 -179.96° -179.46°
7 7.000e+03 4.780e-06 2.350e-06 -179.99° -179.49°
Total Harmonic Distortion: 0.008041%

Grounded PS:

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 4.013e+00 1.000e+00 -0.48° 0.00°
2 2.000e+03 6.831e-04 1.702e-04 153.26° 153.74°
3 3.000e+03 5.623e-04 1.401e-04 16.21° 16.69°
4 4.000e+03 3.786e-05 9.435e-06 -80.85° -80.37°
5 5.000e+03 3.080e-05 7.674e-06 -165.27° -164.79°
6 6.000e+03 1.509e-05 3.761e-06 168.18° 168.66°
7 7.000e+03 1.023e-05 2.549e-06 177.28° 177.76°
Total Harmonic Distortion: 0.022085%

This is the best THD under the same conditions from my previous design which has grounded PS:
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 4.042e+00 1.000e+00 -0.12° 0.00°
2 2.000e+03 6.355e-04 1.572e-04 24.32° 24.44°
3 3.000e+03 5.170e-05 1.279e-05 27.04° 27.16°
4 4.000e+03 4.809e-05 1.190e-05 -179.51° -179.38°
5 5.000e+03 3.709e-05 9.175e-06 179.91° 180.04°
6 6.000e+03 3.073e-05 7.602e-06 179.74° 179.86°
7 7.000e+03 2.631e-05 6.510e-06 179.79° 179.91°
Total Harmonic Distortion: 0.015877%
 
Well, we are all trying to figure out Papa's P6 design. I must retract my "circlotron-like" design because it is not novel in any of its concepts.

My question is whether Papa has devised a way to get circlotron symmetry without floating power supplies.

The asymmetry with my design in post 941 http://www.diyaudio.com/forums/pass-labs/216616-f6-amplifier-95.html#post3133892 is mostly due to the transformer inter-winding capacitances. The floating power-supply design balances the effects of those capacitances eliminating most of the 2nd harmonic.
 
Hi all,

Based on the findings so far from the many valuable contributions on this thread, plus key hints from mr. pass (#1051, #1058), this is something that may be worth playing with for folks with some parts on the bench and impatient. The final one we see from NP may well be very different but with so few parts, it wont be hard to change things around and have some fun meantime.....

Some points:

a) Since the biasing network is ground referenced, I have moved it to be closer to the one done by Ilhquam, (but still R biased here - can use voltage reference I guess if desired). The bias resistor network is high impedance so as to avoid any unwanted AC feedback paths in the op stage. If you go for voltage ref bias (led, zener, tlc etc) I think you'd still want to use a relatively high r network around it (no R should give a path of less than 20K or so to the gate)

b) Input stage I just used 100% per the simplified schematic (apologies to NP for my hack job on the schematic). I built it with no resistors in the drains and it seemed to work fine, no heating of the jfets either. Supply rails for V2 I just connected to V1. At 24 V rails and Idss of 10mA or so, that would mean max of 240mW which is a bit high so have to watch the Idss of your part else drop the rails a few volts. Mine were 6.4mA or so.

c) OLG should be very high so think that CLG should be circa 30x or 30db. Rf can be adjusted to suit. I suggest don't increase Ra much unless the tx has a very high DC-resistance as that would increase the driving impedance on the primary.

This should be a pretty spiffy amp once it grows up and matures :)

I will be travelling this next week so no soldering iron time but will be the first thing I try when I get back. Cheers!
 

Attachments

  • Test-F6-Schema.png
    Test-F6-Schema.png
    55.4 KB · Views: 468
Last edited:
Why is the bias network changed vs the previous version I tried since we now needs caps ? Simple - if we use the previous scheme, the bias resistor network needs to be low impedance. Why ? because if we made the network using high value resistors, the open loop freq response suffers very dramatically as the fets gates have no low-impedance path to source to discharge. The effect is very dramatic if you try it. In the schematic shown above, the low impedance AC path to source is provided through the caps and through the secondaries. The low value of gate stopper is also in aid of the same objective..... just in case anyone was wondering :)
 
The one and only
Joined 2001
Paid Member
a) Since the biasing network is ground referenced, I have moved it to be closer to the one done by Ilhquam, (but still R biased here - can use voltage reference I guess if desired). The bias resistor network is high impedance so as to avoid any unwanted AC feedback paths in the op stage. If you go for voltage ref bias (led, zener, tlc etc) I think you'd still want to use a relatively high r network around it (no R should give a path of less than 20K or so to the gate)

I think this won't give you the perfection of symmetry that others here
seem to be striving for - the bias source will subtly modulate the signal.

:cool:
 
Thanks for the explanation. Does this bias network leave the possibilty of fiddling with transformer loading. I see that you have a pot therefor tweaking, but this will be geared to bias level and not loading, correct?

I think the way I'd do it would be
i) Set the pot for the correct bias'
ii) Check freq response - if no major peaking at the top end you're done.
iii) If peaking, then try a 20k pot across the secondaries and adjust to taste for flat freq resp, nice 10khz sqr wave etc. I think the transformer experts on this list can probably tell you off the top of their hats what sort of compensation you need if you show the the freq response and sqr wave (provided the tx is a common one of course)

In any case, papa's comment has me stumped now unless its just the bottom bias network that's causing the subtle bias modulation asymmetry he mentioned. That's an easy fix. If its the top one .... well that would be a tough nut for me. Perhaps a ground referenced CCS with 'infinite' impedance might be the answer ....

With this circuit papa has found a way to achieve perfect electrical symmetry but whether it will be visually symmetric is another (probably irrelevant) question....
 
Back to basics

I would like to refocus on the F6 Conceptual Schematic in post #1, and comments from Nelson and Zen Mod on the underlying circuit. Some of us are exploring alternatives that probably diverge considerably from that design. That isn't a bad thing. We all learn from stimulating our neurons occasionally, rather than being spoon-fed a design to replicate.

I would like to ask Nelson to respond to a few guidelines about the design.

  • Is the F6 primarily an exercise for the readers? That would not be a bad objective. The design introduces components not seen in the F5 (and most earlier) designs, such as transformers, capacitors, and issues related to loop stability.
  • Will a single 4 winding transformer and two 2 winding transformers give similar results?
  • Does the F6 provide symmetry to provide significant 2nd order distortion cancellation? The effects of inter-winding capacitances are the main issues here.
  • Do the +V1/-V1 power supplies float relative to ground?
 
Official Court Jester
Joined 2003
Paid Member
bottom

feed it from stable potential

either gnd or pos. leg


but that biasing scheme is flawed in few things more ;

you really need to have stable bias ref voltage (either voltage reference or resistor bypassed wit cap ) , then biasing net must be connected to modulating point through resistor of some higher value

as I wrote previously - last few Pa's articles (simple amps ) , same as BA1 FE detail are enough
 
Mr. Pass. You called Conceptual F6 a Classic. Todate, no one has asked you: what is the precedent schematic of this classic old design you used to resurect it as Conceptual F6? Please show us a specific schematic of this old classic. It will be valuable to DIYers to compare with their ongoing designs to revive it.