Local NFB on the PI?

[shaun]

Hi All. Hope everyone is staying safe, and I can’t wait for life to return to normal.

Anywho, I have a question about local negative feedback. I’ve looked at Robinette’s ideas for LNFB and also Merlin's ideas (posted below). My main question concerns whether LNFB can be applied to the phase inverter.

My thinking is this: If I apply LNFB to V1 in a simple amp design, such as a basic SE cathode-biased amp like a Champ, it would smooth the signal in the first stage.

But I suspect - in my limited experience - that a harsh treble signal is sometimes the result of the tone stack, whether it’s a single tone pot or a TB Princeton style stack. Therefore, if I can apply LNFB to the PI, it would correct much of the treble up to that point. (I know I could add global NFB from the OT secondary to the PI cathode, but I'm trying to understand the options with LNFB.)

The signal is 180 degrees out of phase between the plate and the grid, so I don’t see why LNFB would not work on the PI stage. But I also think the PI is more a current driver rather than a voltage amplifier – at least, that’s my limited understanding of its role. And I do mean limited ☺. Plus, I've never seen it done, so there must be good reasons why. I was hoping someone would be good enough to explain it to me.

Thank you, as always.

[mresistor]

I'm not sure if this will help but it seems to address your question, you can scroll down to (its a ways down) "Applying local negative feedback around the amplifier - SLPI combination."
where he talks about using lnfb on a split load PI .  He states it may lower distortion .


https://www.angelfire.com/electronic/funwithtubes/more_phase_inverter.html

[shaun]

Thanks mresistor! I've read a lot of Angelfire articles, and I've used some of his circuits to good effect, but I haven't seen this article yet. Much appreciated.

S

Update. Hmmm...just realized that in an SE amp, it's not really a PI (although it does invert), it's more of a driver. I see Angelfire addresses PIs in a PP amp, as sort of does Aitken briefly, when looking at the 7025 section in a Fender AB165. But I guess I'll experiment with LNFB on the driver in my SE amp just to see what happens. Hopefully, no explosions:).

[PRR]

"Phase inverter" covers so many different things (including your SE no-invert case) that it is wrong-thinking.

NFB won't smooth harshness on the incoming signal, "up to that point"; if anything, it reproduces "harsh" better.

The Cathodyne has no voltage gain so it can't do local voltage NFB. It "could" if the 6V6es+OT had voltage gain, but at typical speaker impedance they don't. (17V peak into the cathodyne may be 14V peak at speaker.)

[DummyLoad]

Thanks mresistor! I've read a lot of Angelfire articles, and I've used some of his circuits to good effect, but I haven't seen this article yet. Much appreciated.

S

Update. Hmmm...just realized that in an SE amp, it's not really a PI (although it does invert), it's more of a driver. I see Angelfire addresses PIs in a PP amp, as sort of does Aitken briefly, when looking at the 7025 section in a Fender AB165. But I guess I'll experiment with LNFB on the driver in my SE amp just to see what happens. Hopefully, no explosions:).

it won't go boom! - if you wire it up correctly. the Rin value is important - too high a value and the highs suffer (-3dB high roll-off knee dips deep into AF) the configuration of the schema in your opening post is sometimes referred to as a plate follower.

have never tried it with a LTPI or paraphase, i assume that it would be possible if the two halves were applied a loop to each and balanced. seems like the paraphase would be easiest. again, why? if you need more headroom in a phase inverter that has gain, use a lower mu tube - e.g, 12AU7, etc..   


a few LNFB scenarios listed using a 12AX7 with typical component values & B+: 300V B+ ; 100K Ra ; 1.5K Rk ; 22uF Ck ; 220K load.
-1dB NFB, Rin 47K - Rfb 1m, AV ~22 / 27dB
-5dB NFB, Rin 47K - Rfb 680K, AV ~14 / 23dB
-10dB NFB, Rin 47K - Rfb 360K, AV ~8 / 18dB
-15dB NFB, Rin 47K - Rfb 200K, AV ~5 / 12dB
-20dB NFB, Rin 47K - Rfb 120K, AV ~2.4 / 7.5dB


--pete

[shaun]

"a few LNFB scenarios listed using a 12AX7 with typical component values & B+: 300V B+ ; 100K Ra ; 1.5K Rk ; 22uF Ck ; 220K load.
-1dB NFB, Rin 47K - Rfb 1m, AV ~22 / 27dB
-5dB NFB, Rin 47K - Rfb 680K, AV ~14 / 23dB
-10dB NFB, Rin 47K - Rfb 360K, AV ~8 / 18dB
-15dB NFB, Rin 47K - Rfb 200K, AV ~5 / 12dB
-20dB NFB, Rin 47K - Rfb 120K, AV ~2.4 / 7.5dB"

Thanks Pete. A handy little chart for reference.

[Leevi]

One of the VOX AC50 circuits has some kind of NFB (or PFB) on the PI.


https://el34world.com/charts/Schematics/files/Vox/Vox_ac50_2.pdf


and 5D3


https://el34world.com/charts/Schematics/files/Fender/Fender_deluxe_5d3.pdf


/Leevi

[sluckey]

One of the VOX AC50 circuits has some kind of NFB (or PFB) on the PI.

and 5D3

Those amps have a paraphase inverter, not NFB.

[Leevi]

Those amps have a paraphase inverter, not NFB.

Thanks for the clarification sluckey.


/Leevi

[shaun]

I did some experimenting. Comparing Merlin's local NFB to Robinette's, I decide the latter was much easier to implement and a good place to start. So I tried it on V1 (12AX7) and found that the 30M resistor allowed too little feedback signal to pass; thus, little difference in sound. I tried 20M but settled on a 10M resistor going into a 0.047 cap. It helped tame the brittle treble.

I removed it from V1 and applied the same LNFB circuit to the cathodyne driver, a 12AT7, and I was quite impressed by the improvement in high end tone. Not fabulous, mind you, but a definite smoothing of the high end and an edge taken off the high trebles. So I'm pleasantly surprised and wonder why this isn't done more often. (Probably find out when my house burns down tomorrow:).

I also wonder whether my math is good for the 10M resistor's power rating - maybe someone could give me the nod if it is. Or laugh out loud if not:

125vdc (on plate) divided by 10,000,000 = 0.0000125 Amps of current.
0.0000125 * 125vdc = 0.00156 watts.

So I guess my 10M resistor should be okay at a half watt rating.

[sluckey]

I also wonder whether my math is good for the 10M resistor's power rating - maybe someone could give me the nod if it is. Or laugh out loud if not:

125vdc (on plate) divided by 10,000,000 = 0.0000125 Amps of current.
0.0000125 * 125vdc = 0.00156 watts.

So I guess my 10M resistor should be okay at a half watt rating.

It's pointless to try to calculate power dissipation in that LNFB loop because the capacitor blocks dc current from flowing through the resistor. Zero current means zero power. 

[shaun]

Boy, the things I don't know!
Well, I knew a cap blocks DC, but I guess I hadn't thought it through.
You really made me laugh, Sluckey. Thanks!

[PRR]

There can be AC power dissipation from large signal.

However 300 Volts of ANYthing in 10Meg is 0.009 Watts, so not worth further thought.

[shaun]

Thanks PRR. I was rather worried at first, then realized I should trust the math.