Nelson, thanks for the input. There IS a difference, and many here would not have thought that there would be one. I really would like to see the spectra difference, as I am having trouble evaluating the differences in the distortion waveform. There is more to this than most people realize, and sometimes just measuring is better than speculation about it.
No, we do not claim they are equivalent. Clearly they are not, because one senses voltage and reduces output impedance while the other senses an output current proxy and raises output impedance. Not the same, but both are negative feedback: sampling the output and using that to reduce the effective input at some earlier point in the circuit.ThorstenL said:
You may "prefer" to redefine technical terms but that reduces communication, and may cause confusion in others. Of course, by redefining 'negative feedback' to mean just a subset of negative feedback you can then produce circuits which include a different subset of negative feedback and claim that they are 'zero feedback'.
JCX, I fail to get your point. I have provided examples of three circuits that
anyone here might build. The case of degeneration is typical - I have used
it myself on many occasions, and the case of loop feedback is also typical.
I repeat my point: Degeneration and loop feedback are different. If
you want to call degeneration a form of feedback, I have no argument, but
I dislike seeing them confused in a technical discussion, particularly when
the point put forward is along the lines of "degeneration is just as bad as
loop, so you guys who think you have no feedback amps are fooling
yourselves and/or misleading the poor customer"
anyone here might build. The case of degeneration is typical - I have used
it myself on many occasions, and the case of loop feedback is also typical.
I repeat my point: Degeneration and loop feedback are different. If
you want to call degeneration a form of feedback, I have no argument, but
I dislike seeing them confused in a technical discussion, particularly when
the point put forward is along the lines of "degeneration is just as bad as
loop, so you guys who think you have no feedback amps are fooling
yourselves and/or misleading the poor customer"
I seem to recall that claim on a number of times on this thread, and I
consider it a simplification unjustified in an argument about re-entrant
distortion.
Unfortunately, I am at the coast and my own website is refusing my
password, otherwise I would simply post the MicroCap files and anyone
would be able to play with them. As it is, you can simply insert those
schematics into either MicroCap or LTspice and perform the same analysis.
Note that I have allowed 2 seconds for the 1 KHz waveform to settle,
and you can see the scaling factors on the plot labels.
If you perform the spectral analysis you will in fact see some 3rd creeping
in with the case of degeneration - it's just to small to see easily.
Also of note, in the case of 100% degeneration, as with a Source follower,
you will see the 3rd clearly compared to the second, but of course both
will be at quite low levels.
Nobody is confusing them. We know they are different, but they are both feedback so it is misleading to claim that one of them is not. I don't believe feedback is bad, although if badly used it can have bad effects. A weakness of degeneration is that it does not sample the actual output but a proxy. Degeneration, like other forms of feedback, can convert low order distortion to smaller amounts of high order distortion. One trick which can be done is to use a small amount of degeneration to cancel the 3rd-order in the exponential response of a BJT.Nelson Pass said:
Now, what about expanding and contracting 3'rd harmonic, DF96? ( The answer to your 'trick' is: Re=1/2Gm) But if you degenerate more, then you get contracting 3'rd, If you use none, you retain expanding third harmonic. Which is worse? Which is better? Does it matter? What about a differential pair, where does the 3'rd come from?
We are in full agreement.
Now that nobody has noticed it, I have to confess that a good chunk of what
you see in the difference waveform of my loop example is due to phase shift of
the fundamental. I thought surely that JCX would have caught it.
I still insist on the differences between degeneration and loop.
Yes, of course degeneration is a form of negative feedback, in the sense that it can be fit into the feedback model. However it differs from loop feedback in that it does not require a causal sequence of events in unidirectional time.
Entropy is just a word for an arrow of time. Re-entrant distortion components imply a causal path through time (around the loop). Proof of the difference is the (relative) lack of re-entrant distortion components in the degenerative case. (And this isn't really strictly true either, just a closer model. All we have are maps, not the world.)
"Camelot!"
"Camelot!"
"Camelot!"
"It's only a model."
Thanks,
Chris
This line of reasoning is dangerous. Beware of the feedback going round and round this has no basis in reality. There is a steady state solution and NO re-entrant distortion.
To clarify, a simple third order non-linearity inside a feedback loop has all odd order distortions present in the steady state solution. All these intuitive concepts like re-entrance, delay, etc. are not useful. The actual propagation delay at audio frequencies is such a tiny portion of a wavelength of any frequency of interest that it is ignorable.
Last edited:
however impractical in hardware my sim showed I can sense the output variable - diff V across the collector R
scale it - (feedback loop Beta ring a bell?)
"subtract" it from the input signal by putting the controlled source in series with the Q - in exactly the "degeneration" resistor position
I have a "loop" from the measured output variable through the controlled source to the input of the amp
the input Z, gain, distortion harmonic profile are, at the visible resolution of the plots, identical for my "thought experiment" loop feedback and resistive degeneration
if you want a "bigger loop" you can move my E to be in series with Q2 base, with Q2 emitter gnded - the feedback loop still works, outputs still look identical, Ib differ slightly due to Early effect
scale it - (feedback loop Beta ring a bell?)
"subtract" it from the input signal by putting the controlled source in series with the Q - in exactly the "degeneration" resistor position
I have a "loop" from the measured output variable through the controlled source to the input of the amp
the input Z, gain, distortion harmonic profile are, at the visible resolution of the plots, identical for my "thought experiment" loop feedback and resistive degeneration
if you want a "bigger loop" you can move my E to be in series with Q2 base, with Q2 emitter gnded - the feedback loop still works, outputs still look identical, Ib differ slightly due to Early effect
Attachments
I should have known better than to have used the word "re-entrant", with all of its baggage. If I could go back in time and change it to "that change in distortion spectra accompanying loop feedback; see Baxandall and Bruno P" I would, but I can't. My argument is not based in time delays but in causal order. This does this, causing that to do that, etc. Loop feedback then plugs that into this. I simply contend that this causal chain doesn't apply to cathode/emitter/source degeneration.
Thanks,
Chris
Last time I did the maths I think it was Re=1/3gm, assuming an ideal exponential response. Expanding or contracting 3rd, which is worse? I don't know. Other things being equal, I guess less 3rd is better than more 3rd. In practice people rarely seem to use exactly the right Re to cancel 3rd, but it might be useful in an amp which has no other feedback.john curl said:
Differential pair 3rd? I haven't done the maths, but I guess there are two contributions: the BJT themselves, and the mixing of 2nd from each of them. Its getting near my bedtime here so I don't want to work it out tonight!
Simple algebra says you are wrong, as Scott has said. You can't escape causality so easily.Chris Hornbeck said:
10dB feedback
Hi Nelson,
Seein as nobody else can be bothered to do a simple analysis of your ccts, I've done a bit.
Your degenerated case does indeed have 10dB of loop gain.
However, the explicit loop feedback case has only 0.22dB of loop gain. The 10dB reduction in overall gain is aided by the potentail divider formed by R5 and R4||R3.
Alex
Hi Nelson,
Seein as nobody else can be bothered to do a simple analysis of your ccts, I've done a bit.
Your degenerated case does indeed have 10dB of loop gain.
However, the explicit loop feedback case has only 0.22dB of loop gain. The 10dB reduction in overall gain is aided by the potentail divider formed by R5 and R4||R3.
Alex
"I simply contend that this causal chain doesn't apply to cathode/emitter/source degeneration."
The same charge flows through the cathode/emitter/source and through the degenerating resistor. It's not possible to distinguish, from the "outside", how much resistance is inside the active device and how much is a separate resistor.
It's not that degeneration escapes causality, but rather that causality happens to the degenerated device as a single step in the causality chain.
Edit: at least down to the level of charge flow.
Thanks,
Chris
The same charge flows through the cathode/emitter/source and through the degenerating resistor. It's not possible to distinguish, from the "outside", how much resistance is inside the active device and how much is a separate resistor.
It's not that degeneration escapes causality, but rather that causality happens to the degenerated device as a single step in the causality chain.
Edit: at least down to the level of charge flow.
Thanks,
Chris
Last edited:
V Rusty and more haste less speed
Sorry Nelson, I was right and wrong the frist time:
Taking the amount of feedback applied as being (1+AB) where A is the forward path gain, and B is the feedback path gain. (I kept posting AB)
the feedback in the degeneration is 9.7dB
the feedback in the loop case is 8.49dB
Can you post your IRF540 model?
Alex
Sorry Nelson, I was right and wrong the frist time:
Taking the amount of feedback applied as being (1+AB) where A is the forward path gain, and B is the feedback path gain. (I kept posting AB)
the feedback in the degeneration is 9.7dB
the feedback in the loop case is 8.49dB
Can you post your IRF540 model?
Alex
- Status
- Not open for further replies.