LTP NFB question

[catalin gramada]

I think is more easy to implement if I move the "shunt" feedback resistor up to cathodes, meant reversing the tail / shunt position. The feedback return should not be necessary the ground , leaving the OT output floating, despite someone may claim safety consideration...just an ideea.

[Helmholtz]

I think is more easy to implement if I move the "shunt" feedback resistor up to cathodes, meant reversing the tail / shunt position. The feedback return should not be necessary the ground , leaving the OT output floating, despite someone may claim safety consideration...just an ideea.

That would complete change your NFB divider and feed excesiive NFB signal.

[Helmholtz]

Yes, I remember that earlier discussion.

The circuit proposed by uneumann is an extension of the same principle of feeding a feedback signal to the tail.
Only he decouples NFB and tail signals and feeds more feedback signal to the tail.

I think bootstrapping the tail resistor is a good explanation for the balancing effect.

But the result in the amp should be tested for stability as the "speaker signal" is frequency filtered by OT and speaker.

[catalin gramada]

That would complete change your NFB divider and feed excesiive NFB signal.

Thought we are talking about how to modify the grid/cathode nfb ratio and not about the amount of nfb injected in circuit. I like more the ideea to use cathode/tail divider rather than to use an alternate nfb path so I give it a chance to test it. Considering the shunt as a part of a tail I may return the nfb wherever is convenient for a proper balance

[Helmholtz]

My comment of post #63 referred to your idea of moving the feedback shunt resistor on top of the tail resistor.
I think the tail resistor must sit above the NFB shunt.
Otherwise no desirable "bootstraping" of the tail resistor.

Check the proposal of uneumann in the other thread.
That gives the possibility to more or less independently adjust NFB ratio and tail signal.

[catalin gramada]

My grids are tied directly to the ground

[Helmholtz]

My grids are tied directly to the ground

That together with your negative supply shouldn't be a problem with uneumann's circuit.
Just change the lower tail wiring.
You might need different resisror values.

[catalin gramada]

That together with your negative supply shouldn't be a problem with uneumann's circuit.
Just change the lower tail wiring.
You might need different resisror values.

Of course. There is a ratio which may be determined in respect with the plates unbalance saying my outputs are unbalanced 3.5 percent please? He used 25 percent as nfb ratio but did not found how was determined in respect with the output

[Helmholtz]

There is a ratio which may be determined in respect with the plates unbalance saying my outputs are unbalanced 3.5 percent please?

Don't think there's a simple formula.

[circlotron]

My apologies if this has been asked, but have you tried applying balanced feedback? Disconnect OT common from ground and install virtual center tap on secondary, then run feedback to both PI inputs from each secondary leg? Seems like one could easily tweak values until good balance is achieved.

Along the same lines, have you tried applying negative feedback to the input of the power amp (1st triode grid instead of second), like one would with an inverting op amp stage?

[GainFreak]

Invalid file. Mesa used...

Here's a screenshot.

[Helmholtz]

Along the same lines, have you tried applying negative feedback to the input of the power amp (1st triode grid instead of second), like one would with an inverting op amp stage?

As the speaker output is in phase with the PI input, this would require inverting the OT phase. So far no problem.
But the problem with the proposed wiring is that it causes low input impedance as it makes the input grid a virtual earth node.
Another consequence is that the closed loop gain of the power amp will depend on the peamp source impedance.

[Helmholtz]

I think I now understand why the NFB increases the PI imbalance.

The simple reason is the inherent imbalance of the LTPI.
It is known that the forward signal applied to the left side (input) grid produces a larger (inverted) signal at the left plate and a somewhat lower (non-inverted) signal at the right side plate.
Means that a grid signal always produces a larger amplitude at the plate of the same tube.

Now when a NFB signal is applied to the right side grid, the same thing happens in reverse:
The amplified and inverted NFB signal at the right plate is larger than the non-inverted NFB signal at the left plate.
More NFB content means less total amplitude as the NFB signal is opposite phase to the direct/forward signal (holds for both triodes).

Consequently the right side plate signal gets more diminished by the NFB than the left plate signal => more imbalance.

If not clear enough I might try to explain in more detail tomorrow. Too tired now.

[catalin gramada]

OK.
the circuit (outputs marked y and b in respect with channels on scope)





and the test done with nfb injected- with and without nfb divider. First with nfb grid in common mode (100nf cap directly to shunt). Second with nfb grid into divider. I did it empirical with a series 1k to not load to much then tailoring for the best balance I got. 560 ohm seems allright.
To restore the same amount of nfb in circuit is still necessary to make the series nfb resistor half or less and well...is a mess. Once I replaced 1k series nfb resistor with 470 ohm the outputs comes unbalanced again. I had to change again the shunt of divider from 560 ohm to 2.2k to run outputs in balance and get another amount of nfb in circuit than I expected
So I may conclude the ratio between nfb to the grid and nfb to cathode have no consistence but also depends by the amount of nfb we want to inject in circuit. That meant if want to use more or less nfb we have to alter the cathode/grid nfb ratio as well if wanna preserve outputs balance.
The balance is determined by the ratio of cathode/grid nfb and the ratio is determined by the amount of feedback we want to use. May be also more variables as the supply voltage or tubes matching etc ?...have no ideea did not tested yet. But at first sight seems to be a equation with to many variables to be something easy to use...

[catalin gramada]

Next try:
It is much better if we drive an alternate nfb path directly from OT and not from the shunt. Like this




This is easy to manage and think is a winner.
So the nfb should be injected first to the grid and only there. Then we create another path from OT into the tail and tweak it till balance both outputs. Trying to drive from tail shunt and mess the whole circuit.
This one It works and balance very easy without to mess with global feedback injected in second grid.