I wish to admit a mistake. I apologize for this.
My mistake was to say the the loop flight time in an audio amplifier was 10 or so nano seconds. It’s incorrect. It’s more like 0.5 to 1 nano second, assuming electron flow of circa 0.9 c and physical distance travelled of 0.15 meters.
There is a phase difference between the input annd the feedback signal, but there is no delay. The first feedback electrons are arriving at the inverting input within 1 nano second of leaving LTP or DB output.
My mistake was to say the the loop flight time in an audio amplifier was 10 or so nano seconds. It’s incorrect. It’s more like 0.5 to 1 nano second, assuming electron flow of circa 0.9 c and physical distance travelled of 0.15 meters.
There is a phase difference between the input annd the feedback signal, but there is no delay. The first feedback electrons are arriving at the inverting input within 1 nano second of leaving LTP or DB output.
FR4 permittivity is around 4.7*epsilon0, so signal* propagation speed is around C/2.
*Certainly not electron speed !
Ok. Agreed. So it’s 2 nano seconds.
Point is, there’s no delay in practical terms at audio, so let’s stop this Martin Colloms inspired nonsense. If you don’t understand the difference between delay and phase . . .
Point is, there’s no delay in practical terms at audio, so let’s stop this Martin Colloms inspired nonsense. If you don’t understand the difference between delay and phase . . .
Last edited:
Time delay = Φ(ω)/ω, so it would be quite clever to have no time delay with a certain phase at ω ≠0.
But this whole electron discussion is BS, so I will also look for the popcorn.
Hans
Last edited:
No time delay Hans other than the minuscule loop transit time which is measured in nano seconds and irrelevant for our discussion.
Feedback current begins to flow immediately- loop transit time notwithstanding- and it builds to its peak value some time later and you see this as phase shift.
If this were not the case, the amplifier would be running open loop for the time it would take for Vin to equal Vo after the feedback divider. Which is exactly what the anti-feedback people say: the amp runs open loop because of the feedback delays around the loop.
See the Audio Note website for more of this kind of stuff.
😀
Feedback current begins to flow immediately- loop transit time notwithstanding- and it builds to its peak value some time later and you see this as phase shift.
If this were not the case, the amplifier would be running open loop for the time it would take for Vin to equal Vo after the feedback divider. Which is exactly what the anti-feedback people say: the amp runs open loop because of the feedback delays around the loop.
See the Audio Note website for more of this kind of stuff.
😀
Go back to Laplace and a simple RC low pass which has the same one pole transform as the basic amplifier with feedback. When you apply a step the output immediately responds.
The mathematical representation has no physical dimensions so as was stated c would be the limit in a real amplifier. One thing has been missed, the physical reality of how devices are actually fabricated often involves the parameters being distributed in three dimensions whereas the models are lumped. At the highest frequencies distributed R/C networks behave more like a diffusion of the signal which can mimic a delay in response. This is sometimes modeled as an excess phase. This is also at frequencies far beyond audio.
This idea that there is a true delay in a feedback system leads to nonsense like the feedback goes round and round. How physical reality actually works is not up for a vote.
Yes MUCH faster.
Gentlemen please, "audio" electrons, electrons vs the EM signal, this is drifting off into the non-physical. So to speak 😉
Last edited:
I have demonstrated this with a very small sim a step charging a C. It is quite clear very quickly that any change in signal immediately results in a change in voltage on the C, notwithstanding that there is a phase shift between source and C. Must have done it a dozen or more times.
And please (not you Scott) stop this nonsense about 'phase delay'. You either have phase shift or time delay.
Jan
Physical components have physical dimensions, traces on board have physical dimension. Speed of information is limited.
Delay from input and output signal in a physical amp is a fact.
The fact that we don't care in audio is not a proof of non existence.
Phase delay and group delay are certainly not nonsense, and have well known definition.
Good. Then this is my main point: Because the collector current is almost exactly equal to the emitter current which is equal to the current through Rg minus the current from the output stage, we have current feedback.
Please note that Middlebrook's DIT will tell you that voltage and current feedback co-exist in CFA's, but that current feedback usually predominates.
This is why I use the term "signal current" to distinguish it from "bias current."
A transconductor requires at least three terminals to operate: two for control and at least one additional dedicated exclusively to the signal being controlled. A resistor and the Early Effect that it models are two-terminal entities. A resistor is no more a transconductance than it is a transresistance.
Let us distinguish between "current feedback" and "CFA". Middlebrook does not tell you whether you have a CFA, the name of which has been assigned by industry to a gain block with a specific topology. Middlebrook tells you whether current or voltage feedback predominates at a specific point in given circuit. Does this distinction make sense? I shall assume that you accept MDIT unless you say otherwise.
I'm not sure what you mean by "single device input." To me, by definition, a CFA's inverting input is connected to emitters/sources/cathodes. Current on demand is inherent in such a connection. And as long as the impedance of the inverting input is less than that of the impedance of the feedback network that that input sees, bandwidth can be held more or less constant as closed loop gain is varied. These characteristics are inherent in every CFA. They are not inherent in VFAs.
I am not arguing precedent here. It matters not to me in which topology a particular feature appeared first.
Here is the infamous Martin Colloms article:-
A Future Without Feedback? | Stereophile.com
Here is another one that follows the same trend:-
Audio Note
Both talk about 'delays', feedback going around and around etc.
A Future Without Feedback? | Stereophile.com
Here is another one that follows the same trend:-
Audio Note
Both talk about 'delays', feedback going around and around etc.
Happy Holidays as well and thanks for going through the formalism. Given circuits isn't my first love within EE, there's a few cobwebs needing clearing. And thank goodness your handwriting is better than my own. 😀 (and glad your table is reinforced down in the measurement room, to follow up from before)
(Not directed anywhere in specific) I've long lost track of the threads direction and the arbitrary arguments about nomenclature versus "not caring what it's called but pay attention to the mechanisms". If "CFA" vs "VFA" is useful in term of giving a general idea of the system's behavior (from a family of topologies), then its served its purpose well enough. Aren't there more productive things to attack?