LTP NFB question

[catalin gramada]

So, to get a conclusion maybe someone may pronounce why it work better balanced with the nfb injected in common mode despite the fact mostly of texts states it was a Fender mistake used by circumstances ? This experiment clearly shows for same amount of nfb the circuit is balanced (or better balance) with shunt in the tail and out of balance just in the grid please.
I don't know if injected nfb just in the grid and used a CCS will force into the balance, but it works very well as it is, The shunt in the tail do something magic to compensate the nfb in grid without to compromise the common mode as seen.

[Helmholtz]

A small part of the feedback signal injected at the tail reaches the cathodes, which are low impedance inputs.
Other than a grid signal a cathode signal doesn't get inverted at the plate.
The result is increased gain of the right side triode and decreased gain of the left side triode, so better balance.

[catalin gramada]

A small part of the feedback signal injected at the tail reaches the cathodes, which are low impedance inputs.
Other than a grid signal a cathode signal doesn't get inverted at the plate.
The result is increased gain of the right side triode and decreased gain of the left side triode, so better balance.

Looks smarter than to mess with the plate loads matching...

[Helmholtz]

Looks smarter than to mess with the plate loads matching...

...and explains why the method is used so often.

[catalin gramada]

So is a subtle balance between the amount of nfb injected to grid vs the amount of nfb injected to cathode to keep everything in balance. Can we make it variable in a way to modify this raport -grid vs cathode nfb injection- for a certain amount of nfb we want ? Say I want to inject little more to cathode in respect with the grid or litle bit less in grid in respect with cathodes to do a perfect balance ?

[Helmholtz]

I think so.
Increasing the value of NFB shunt resistor in the tail should intensify the balancing effect.

[catalin gramada]

I think so.
Increasing the value of NFB shunt resistor in the tail should intensify the balancing effect.

Are you suggesting say 10k / 1k will do better balance than 1k / 100 ohm considering same nfb amount please ?

[Helmholtz]

Are you suggesting say 10k / 1k will do better balance than 1k / 100 ohm considering same nfb amount please ?

Yes, I think so. What matters is the ratio of NFB shunt resistor to total tail resistance.
You might have to lower the value of the the upper tail resistor to preserve operating conditions/plate voltages.

[catalin gramada]

Yes, I think so. What matters is the ratio of NFB shunt resistor to total tail resistance.
You might have to lower the value of the the upper tail resistor to preserve operating conditions/plate voltages.

So to get sensible diferences meant the order of magnitude of the shunt should be comparable with the tail meant changing the shunt from 10 to 100 ohm will not do the same diference as changing 1k to 10k as time the tail is in tens of kilos range. I trying to think how the common mode is affected by that.

[GainFreak]

This LTP reminds me of MESA Dyna-Watt type of amp.

Mesa 20 20 poweramp.pdf

[catalin gramada]

Invalid file. Mesa used in 400w bass models but they did something else. They did not used the nfb shunt in the tail so severe unbalanced may occur as it was show. Still the schematic shows they used a CCS which..., may be, force well the PI to stay in balance (as it is more effective than what I did). I never measured a Mesa so I cannot confirm if it works or not from outputs balance point of view.
Some other models did not used global feedback at all.
Found also a pic of a Turner amp using CCS in the tail and nfb only in the grid so I may suppose is a solution which work. The outputs shows 5.3V on the both sides. So yes, using ccs is possible.
But if don't use ccs and just a resistor to get common mode the nfb should be applied on the tail as well to compensate as it was shown. If you put a small value like in you Mesa example the outputs are severe unbalanced. Use nfb only in the grid and will be more.The only solution is to mess with plate loads. What they did with 68k on one side and 120k on the other.

[Helmholtz]

So to get sensible diferences meant the order of magnitude of the shunt should be comparable with the tail meant changing the shunt from 10 to 100 ohm will not do the same diference as changing 1k to 10k as time the tail is in tens of kilos range. I trying to think how the common mode is affected by that.

Feeding the NFB to the tail means common mode drive.
The total tail resistance and the cathode(s) input resistance form a voltage divider.
The closer to the cathodes you feed the signal, the more balancing effect it should have.
But if you overdo it, feedback ratio willl suffer.

Don't forget, it's all theory.
I mean I love theory but it's so easy to overlook something. The proof is always in real measurement.

[catalin gramada]

Same in Mesa 400 but no nfb used here.

[SoulFetish]

I apologize if this has already been mentioned, but you can keep your original feedback (or whatever) injected into the non-inverting input, and take a small amount of the speaker output and inject it into the tail as positive feedback. Positive feedback here will bootstrap the tail, and can be used to balance the outputs.

edit: there is a discussion on this here -
https://music-electronics-forum.com/...-design/49744-

[catalin gramada]

I apologize if this has already been mentioned, but you can keep your original feedback (or whatever) injected into the non-inverting input, and take a small amount of the speaker output and inject it into the tail as positive feedback. Positive feedback here will bootstrap the tail, and can be used to balance the outputs.

edit: there is a discussion on this here -
https://music-electronics-forum.com/...-design/49744-

Hey. Thanks. This will be an interesting reading and was what I thinking about to make it variable. Still at the first sight the Helmholtz ideea makes things simpler because a divider between the tail and input cathode resistance already exist and just tailoring the resistor values will get the optimal raport.
I will get a good reading about suggested topic. Thanks.