Negative feedback vs output impedance

[Circlotron]

Normally when you apply negative feedback to an amplifier it lowers the output impedance. What about the following situation? A valve class A output stage has a cathode resistor and a bypass capacitor across it. If we remove the capacitor, the varying cathode current makes for a varying voltage drop across this cathode resistor. This varying voltage subtracts from the grid drive voltage so is in effect negative feedback. The thing is, this feedback linearises the amplifier output current, not the voltage. So it raises the output impedance, not lowers it like negative feedback should. Any thoughts on this?

[T3sl4co1l]

Right, you can take feedback however you like.  Current feedback raises Zo, voltage feedback lowers Zo.

Around a single device, voltage feedback most often manifests as shunt feedback (from plate to grid, etc.; includes internal fields as in vacuum triodes and SITs (static induction transistors)).  In audio amps, this is done to reduce Zo and distortion; in RF amps it's done also for neutralization (which is to say, flattening the phase and frequency response too).

You can even take a proportion of both, so that distortion and gain are reduced proportionally, while Zo tends towards an ideal value set by the ratio of feedback types.  Or if one or the other feedback is positive (but overall gain is such that the system doesn't oscillate), it can even have a negative resistance characteristic.  Example, using positive current feedback to compensate for DCR in a motor driver (aka EMF control).

A friend has designed an audio amplifier with just such a characteristic, where Zo is continuously variable from CC to CV.  Nothing really ground breaking, just the clever application of an OTA or two.  Audiophiles love their oddball amplifiers, it's attracted a bit of attention as I recall.

Tim

[Circlotron]

Oh yeah, current feedback...  it’s been quite a few years since I fiddled seriously with audio amps. I actually made an amplifier with enough negative resistance to compensate voice coil dc resistance. Not worth the trouble IMHO. Did many interesting thought experiments with voltage drive vs current drive for loudspeakers. Came to the conclusion that given a sinusoidal signal, current drive gives sine cone force, therefore displacement, whereas voltage drive gives sine cone velocity. How they sound the same to one’s ear I don’t know. What I do feel sure of though is the most appropriate drive signal for a speaker with motional feedback is current drive. You are then sending the speaker a force signal, not a velocity signal. Somewhat comparable to the advantages of a current mode SMPS regarding eliminating a pole in the feedback loop. I’m unaware of anyone having tried that approach. Long time since I looked though.

[magic]

Right, you can take feedback however you like.  Current feedback raises Zo, voltage feedback lowers Zo.

Nice can of worms you have opened here
https://www.analog.com/en/analog-dialogue/articles/current-feedback-amplifiers-1.html

On other forums that shall remain nameless I have seen whole thread devoted to debating which of the two should be considered "current feedback"

[T3sl4co1l]

Right, you can take feedback however you like.  Current feedback raises Zo, voltage feedback lowers Zo.

Nice can of worms you have opened here
https://www.analog.com/en/analog-dialogue/articles/current-feedback-amplifiers-1.html

On other forums that shall remain nameless I have seen whole thread devoted to debating which of the two should be considered "current feedback"

Heh!  Note that a current feedback op-amp refers to something slightly different: the inputs are current mode, versus a conventional voltage mode op-amp which depends on the difference in input voltages.  The divider resistors in the given diagram, are still feeding back voltage, so will act to reduce Zo -- the trick is the amp core itself can be faster, in particular having bandwidth almost independent of gain, very helpful for some applications!

Tim

[penfold]

Right, you can take feedback however you like.  Current feedback raises Zo, voltage feedback lowers Zo.

Nice can of worms you have opened here
https://www.analog.com/en/analog-dialogue/articles/current-feedback-amplifiers-1.html

On other forums that shall remain nameless I have seen whole thread devoted to debating which of the two should be considered "current feedback"

Heh!  Note that a current feedback op-amp refers to something slightly different: the inputs are current mode, versus a conventional voltage mode op-amp which depends on the difference in input voltages.  The divider resistors in the given diagram, are still feeding back voltage, so will act to reduce Zo -- the trick is the amp core itself can be faster, in particular having bandwidth almost independent of gain, very helpful for some applications!

Tim

Or is one input voltage mode and the feedback terminal a current (avoiding confusion with a Norton opamp where both inputs are current)

[T3sl4co1l]

Yes, that too

Tim

[TimFox]

In my first formal electronics course, ca. 1969, I learned the following basic rules for negative feedback:
1.  Feedback from the output voltage tries to maintain that voltage, thus reducing the output impedance.
2.  Similarly, feedback from the output current tries to maintain that current, increasing the output impedance.
3.  Voltage feedback in series with the input (e.g. to the cathode of the input tube) increases the input impedance (but does not affect resistors not included in the feedback, e.g. the grid-to-ground resistor of this example).
4.  Current feedback in parallel with the input (e.g. a feedback resistor to the grid of the input tube) decreases the input impedance.
If I remember correctly, positive feedback (less than oscillation) inverts all four of these statements.  For example, bootstrapping the grid resistor from the cathode voltage to increase the input impedance.

[mawyatt]

Might include in your list that negative feedback can invert the feedback function. The classic T = G/(1+GH) = 1/(1/G +H) ~ 1/H as G >>1. This is often used to create various transfer functions like inverting a low pass to high pass, or high pass to low pass, or bandpass to bandstop, or bandstop to bandpass and so on.

Best,

[TimFox]

True.  Also, positive feedback can be used to "sharpen" the response of a frequency-response function (the basis of traditional "regenerative" receivers).

[mawyatt]

Or create an oscillator if the Barkhausen criterion is satisfied.

Best,

[mawyatt]

I recall in undergrad school in the 60s we were taught about Root-Locus techniques in Control Theory and remember plotting the poles and zeros of various systems. In the early days of DSP control systems a paper (can't recall the source) used an analogy of a sampled feedback system where the poles and zeros were plotted just like in an analog system, but the j-omega axis was actually a radius rather than just a vertical axis like in the standard Root-Locus diagram.

The j-omega axis radius was the sampling rate, so the j-omega axis would bend towards the poles and illustrate how the loop stability was affected (kind of reverse where to poles move towards the axis in analog systems) as the sample rate dropped. The analog Root-Locus was just a special case of this with a radius of infinity since the sample period is zero for continuous analog. With this technique it becomes clear how the sample rate affects loop stability is sampled feedback systems. If the sample rate was such that the axis intersects the complex system poles, then the Barkhausen criterion has been satisfied by the sampling function, and the system becomes an oscillator.

Anyway, thought this might of interest to some folks, it certainly provided an improved intuitive understanding of sampled feedback systems for myself

Best,