Joe Rasmussen "Trans-Amp" - 40 Watt Transconductance "Current Amplifier"

I have a parallel resistor calculator. It is called a calculator. :nod:

(R1xR2) divided by (R1+R2)
Ya, that works too, but by using R1, 1/x, +, R2, 1/x, =, 1/x., you only enter the values once.

I have a bit of code that I wrote 25 years ago for calculating R,L,C,F, but for parallel resistors, just add or subtract the reciprocals. For example, suppose you need 90K. then parallel 100K with: (90K, 1/x, -100k, 1/x, =, 1/x)= 900K (910K close enough, or 220+680).
 
Anyone measured the Output Z of this over 20 - 20k Hz?
Circuit analysis for this kind of amp (without the parallel to load components) yields:

Zo = (Ad +1)Rs + Za,

where Ad is the differential gain of the opamp, Rs the sense resistance and Za the intrinsic output impedance of the opamp. In this case Ad(20kHz) =~ 320 (50dB), Rs = 0.33 and Za =~ 0, so Zo(20kHz) =~ 100 ohm and capacitive (without the parallel components). The 0.1uF cap at 20kHz amounts to ~80 ohm and in effect parallel with the former yields ~44 ohms. With Rs =1 and C =22n (which are still realistic) we would get at ~170 ohm capacitive.
 
Zo = (Ad +1)Rs + Za,

where Ad is the differential gain of the opamp, Rs the sense resistance and Za the intrinsic output impedance of the opamp. In this case Ad(20kHz) =~ 320 (50dB), Rs = 0.33 and Za =~ 0, so Zo(20kHz) =~ 100 ohm and capacitive (without the parallel components). The 0.1uF cap at 20kHz amounts to ~80 ohm and in effect parallel with the former yields ~44 ohms. With Rs =1 and C =22n (which are still realistic) we would get at ~170 ohm capacitive.
Thanks for this ETM. I sorta expected this. Don't think 22n with that circuit is unconditionally stable.

Looks like I can't get Zo > 1k without something at least as complex as Hawkesford et al :stop:
 
So I am wondering, after reading this thread, is the Zobel really needed? Or can one get away without it if driver is a careful choice?
If the load impedance including cables is well behaving in the region around 1 MHz (where loop oscillatios can occur) without deep dips or strongly inductive behavior, then IMO the RC can be omitted, but this has to be checked with a scope. It's about the minimum gain requirement of the chip that must be met, but this is naturally specified for resistive feedback, whereas in this topology the load inductivity can cause excess phase lag in the feedback signal, which can endanger stability without compensation even if the gain itself is satisfied.

Another aid is to lower the sensing resistance for high frequencies, thus decreasing the feedback factor there, but this also comes with some phase lag in the transition region (e.g. from 50 kHz to 200 kHz).
 
Assuming the Zobel current does not return through the current sense resistor, it does not reduce the output impedance. The Zobel is there to stabilize the CFP part of the quasi output stack, which is a voltage source that reacts to the speaker current through the current sense feedback creating a synthetic high impedance. You could say that the impedance between the output and ground is near zero while the impedance from the output to the current sense is high impedance because the output voltage reacts to the current feedback. It does not react to any additional current to the Zobel network etc.