How much feedback depth you can cope with?

[BesPav]

Of course at a given UGF (realistically for audio power amplifier with known power transistors an ones of MHz's, for example say 3-7.)

And how much do we really need?

 

[MarcelvdG]

Fifth-order conditionally-stable loop?

 

[jxdking]

If everything is perfect as simulation, it will be stable.

Can you share the schematics?

 

[BesPav]

Fifth-order conditionally-stable loop?

Yes, now compensated to a gain of +26dB.

If everything is perfect as simulation, it will be stable.
Can you share the schematics?

Of course, but let's start from there:

Opamp + Common base

VAS input is a low impedance emitters of Q5/Q6 but they are driven from high impedance collectors of Q1/Q4.
And since drive currents can be easily summed we can just duplicate opamp+bjt stage picking additional stage's input from previous output. Voila, +80 dB.

 

[jxdking]

Opamp + Common base

That idea is not bad.
Maybe I would trade some NFB for more stability. Turn down the voltage gain of the output VAS to 5 ~ 10 times.

 

[MarcelvdG]

You will have to check large-signal stability (at start-up and during recovery from clipping), but that holds for any amplifier, even with first-order compensation.

 

[jan.didden]

Of course at a given UGF (realistically for audio power amplifier with known power transistors an ones of MHz's, for example say 3-7.)

And how much do we really need?

Normally the color of the label is the same as the color of the graph. Is this graph really for the shown equation?

Jan

 

[BesPav]

Normally the color of the label is the same as the color of the graph. Is this graph really for the shown equation?



Jan

Yes.
There are two steps in equation for using Tian's GNFB probe.

 

[BesPav]

You will have to check large-signal stability (at start-up and during recovery from clipping), but that holds for any amplifier, even with first-order compensation.

Clearly.
Recovery without special attention and countermeasures will be ugly and could make an amp unstable at all.

 

[bimo]

Of course at a given UGF (realistically for audio power amplifier with known power transistors an ones of MHz's, for example say 3-7.)

And how much do we really need?

ULGF around 1 to 2 MHz is easy to implement. 2 to 4 MHz can be done with careful pcb layout.

 

[BesPav]

Fifth-order conditionally-stable loop?

Moreover, it can provide aperiodic response, 100k square shown.
One more thing - even if you turn off input RC filter.

Can you share the schematics?

Try attached.