Current Feedback Amplifiers, not only a semantic problem?

I would like to discuss about the first article below and its comments :

Michael Kiwanuka, AudioXpress, June 2017
"Current Feedback and Voltage Feedback Fallacies"
https://www.audioxpress.com/files/attachment/2621

Currently, invalid links :
http://www.waltjung.org/waltsblog/catego...mplifiers/
Some Useful Background Reading on Current Feedback Amplifiers | Walt's Blog 2014


Walt Jung, Sergio Franco, Michael Kiwanuka, AudioXpress, July 2017
"Current Feedback and Voltage Feedback Dispute"
http://www.audioxpress.com/article/curre...ck-dispute

EDIT december 2018. I added links having correlations with the topic :

Christian Henn, Burr-Brown, May 1993
"New Ultra High-Speed Circuit Techniques with Analog ICs"
http://www.ti.com/lit/an/sboa047/sboa047.pdf

E.M.Cherry, IEE Proc/-Circuits Devices Syst; Vol 147, No 6, December 2000
"Feedback amplifier configurations"
https://drive.google.com/file/d/1n030f9Lh8aQe0Hlc45VLJQtzjYJWwLNU/view

Walt Jung and technical staff of Analog Devices, 2005
"Op-amp applications handbook / Current feedback basics"
https://studfiles.net/preview/429636/page:5/

Cyril Mechkov, Technical University of Sofia, Lutz von Wangenheim, discussion started in 2013
What is the truth about the exotic current feedback amplifier? Is it something new or just well known old? Is it really a current feedback device?

Sergio Franco EDN August 23, 2017
"In defense of the current-feedback amplifier"
https://www.edn.com/design/analog/4458753/In-defense-of-the-current-feedback-amplifier

Christopher Paul, AudioXpress, July 2018
"Current Feedback: Fake News or the Real Deal ?"
https://www.audioxpress.com/files/attachment/2660
 
Last edited:
just accept "CFA" as an informal, "trivial name"

I think some of the objection to "cfa" is that the name doesn't fit the existing formal classification schemes - as a Voltage output device it is the Voltage Output that is sampled by the feedback

a more abstract, formal op amp classification: http://web.archive.org/web/20070128220917/http://people.ee.ethz.ch/~hps/publications/2000cas.pdf

but I don't think there is any real dissagreement on how the circuit works

but a lot about whether it is advantageous in audio poewer amplifiers
 
Last edited:
AX tech editor
Joined 2002
Paid Member
I think some of the objection to "cfa" is that the name doesn't fit the existing formal classification schemes - as a Voltage output device it is the Voltage Output that is sampled by the feedback

True, and I myself was also upset when I found out, and I wrote many angry posts here.
But if you take some sensible distance you realise that the term 'CFA' is firmly established for a particular opamp circuit configuration. It is not unusual for a term to have different meanings in different areas or contexts, and if the context is clear like in this case, the smart thing is to accept it.

Jan
 
Pass DIY Apprentice
Joined 2001
Paid Member
I'm dealing with the semantic issues of feedback naming in the article I'm working on now.
My preference is for the wordy but explicit convention which names how it is derived and how it is mixed back. In my research, I've identified a whole bunch of confusing names, some which sound identical but mean completely different things. It's not hard to understand how confusion creeps in. And don't even get me started on so-called Schade fedback :)
 
Member
Joined 2011
Paid Member
Why not name it after its most famous implementation?

We call the circuit in Figure 1 "the LH0002 buffer" after the NSC product.

Instead of "current feedback amplifier" why not call cicuits like Figure 2 "EL2020 type amps" after the Elantec product? I looked at a bunch of Comlinear datasheets but they don't have internal schematics.

_
 

Attachments

  • LH0002.png
    LH0002.png
    53.1 KB · Views: 3,154
  • EL2020.png
    EL2020.png
    86.6 KB · Views: 3,153
True, and I myself was also upset when I found out, and I wrote many angry posts here.But if you take some sensible distance you realise that the term 'CFA' is firmly established for a particular opamp circuit configuration. It is not unusual for a term to have different meanings in different areas or contexts, and if the context is clear like in this case, the smart thing is to accept it. Jan

Semantically

According to the comments [1], the CFA term is firmly "established since 3-4 decades" (Jung).

However, there is a lack of fair acknoledgement that :
- it was already established much earlier to mean the control of the output current of an amp by negative feedback.
- it is probably still now the main usage of the term.

So the term current feedback can be considered as having been appropriated by marketing services (Steffes : "there is certainly a bit of marketing slang involved there") to describe a topology which was a novelty because its uses of a push-pull for an inverting input but was not a novelty at all in the way how the difference between the input signal and the feedback signal is obtained.

I extract a few words written by Peter Baxandall in Michael Talbot-Smith's book
Audio Engineers Reference Book Focal Press

2.8.8 The 'current feedback amplifier' notion
The 'current feedback amplifier' term has appeared in op-amp manufacturers (Ogden, 1992) since 1985, and also in audio power amplifer context (Alexander, 1990). It implies that a new basic principle is involved, but really it is an old one in disguise.
[...]
It is felt that the term 'current feedback' in the above connection is a rather unfortunate one, since this is a long established term employed when negative feedback is derived from the voltage across a small value resistor in series with a load for the purpose of obtaining a high output impedance or, in combination with normal voltage feedback, an output impedance of some specific value
[1]http://www.waltjung.org/waltsblog/current-feedback-amplifiers/useful-background-reading-on-current-feedback-amplifiers/#comments

***

What shall we call an amplifier that actually has current feedback then?
In what context : function intended of the feedback or topology of an input stage ?

***

Technically

Now technically, the term CFA for topology is highly questionnable.
I prefer to describe the input stage than how it is supposed to work :
single device, diamond or differential.

Note that if a diamond input stage is said to belong to the topology of current feedback amplifiers, a single device input stage should also be said to belong to the same topology. But nobody does it. This is a first clue of the weakness of the expression.

Feedback relies on the difference between the signal at the non-inverting input and the signal at inverting input. The input signal at the non-inverting input is a voltage. So the feedback signal can't be a current, it's a voltage.

The circuit which operates the difference is a single transistor or an arrangement of transistors to provide better performances but keeping the main fundamental property of a single transistor which is to be a transconductance device, voltage controlled (by its Vbe) current source (Ic).

These are two more clues about the weakness of calling CFA an amplifier using a diamond circuit input.

Now what is the exact role of the current in the topology called CFA ?
 
AX tech editor
Joined 2002
Paid Member
Forr, I fully agree with you regarding the semantic part. It was unfortunate that this term was hijacked. I wonder whether it was a delibered move, or maybe by someone who was not aware of the original meaning, could also have been the case.

For the technical part, an easy way out is to note that in a 'VFA' the feedback is returned to a common emitter stage which generally is assumed to be voltage driven.
In a 'CFA' the feedback is returned to a common base stage which is generally assumed to be current driven.

I say 'generally' because you can always come up with specific cases that are different. And as we all know, you can't have current without voltage (unless Z = 0).
But as a naming convention to make a broad distinction between two quite different circuit topologies, I believe CFA and VFA are usefull and non-ambiguous identifiers. Has been for 30+ years, and I am not aware of a single case of confusion on this in the context of opamp structures.

Jan
 
Last edited:
You could do what philosophers do in such cases: first define what you mean by a certain term, then use it.

By the way, the term current feedback has at least three meanings. First it meant series feedback at the output, then it changed into series feedback at the output combined with shunt feedback at the input, then it changed into plain old voltage feedback using an amplifier with low open loop impedance at its negative input.
 
Last edited:
AX tech editor
Joined 2002
Paid Member
Feedback relies on the difference between the signal at the non-inverting input and the signal at inverting input. The input signal at the non-inverting input is a voltage. So the feedback signal can't be a current, it's a voltage.

Now you are back at semantics, trying to use the word ' difference' in an overly literal sense. We all know that the output of the ' subtraction stage' depends on the voltage at the non-inverting common emitter input, and the current at the inverting common-base input. Calling this output ' the difference between...' is usefull and clearly understood, even if technically you can't subtract a voltage from a current.

What is your point Forr? It is as if I hear MK talking ;) Don't be a stooge, you're too good for that.

Anyway, first saying you want to use a diamond config as the defining feature, and then declare that it isn't, seems not very useful!

Jan
 
For the technical part, an easy way out is to note that in a 'VFA' the feedback is returned to a common emitter stage which generally is assumed to be voltage driven. In a 'CFA' the feedback is returned to a common base stage which is generally assumed to be current driven.
It is only a common base stage for 0V at the non-inverting input.
But this input receives a voltage signal.

Steffes (I give a small extract below) consider the stage as
having a voltage follower output function but they cascode the “error current” from the feedback network.
Same as a single transistor input stage then.

(Omitting the small resistors used to set the quiescent current), if the feedback network is disconnected from the amp output, the transistors under scrutiny are simply common-emitter , emitter-degenerated by the bottom resistor of the newtwork. Their emitters voltage reproduces the signal input voltage within a few millivolts, just like emitter-followers.

When the feedback is reconnected, the dynamic load (dynamic = resistor + feedback voltage) of the emitters changes. To maintain the copy of the signal input voltage, the emitters deliver more or less current to the constantly changing load; under the control of the Vbe's... just as do the input stages of the other feedback configurations (single device, long-tail pair or, not yet mentionned, Rush circuit).

I say 'generally' because you can always come up with specific cases that are different. And as we all know, you can't have current without voltage (unless Z = 0).
But as a naming convention to make a broad distinction between two quite different circuit topologies, I believe CFA and VFA are usefull and non-ambiguous identifiers. Has been for 30+ years, and I am not aware of a single case of confusion on this in the context of opamp structures.
[...]
What is your point Forr? It is as if I hear MK talking ;) Don't be a stooge, you're too good for that.
I enjoyed to read MK's article but did not wait to be upset by the term CFA which puzzled me as soon I saw it, more than thirty years ago. I already posted my point of view on the question in a post here maybe ten years ago. Albeit arriving at simlar conclusions, my arguments somewhat differ from those of MK, at least the way I develop them, don't you think ?

Steffes said:
Some Useful Background Reading on Current Feedback Amplifiers | Walt's Blog 2014

The real magic of what we have come to call a current feedback op amp is the dual use in the emitter followers at the inverting input. Not only do these transistors provide the voltage follower output function for the buffer across the inputs, but they cascode the “error current” from the feedback network into the inverting node up and down through the current mirrors to the high transimpedance node that is then buffered to the output by the output stage. This “error current” is very real and how the parts operate – hence the current feedback name.

Note : cascoding a current is a puzzling thing.
 
Last edited:
Note : cascoding a current is a puzzling thing.

The better wording is "superimposed" instead of "cascoding".

There is a current coming out of the buffer (driven by the signal source) and another current going into the buffer (driven by the feedback network).
And both are superimposed and form the (small) driving current for the current mirror. And this is why they decided to call it "current feedback".
 
AX tech editor
Joined 2002
Paid Member
It is only a common base stage for 0V at the non-inverting input.
But this input receives a voltage signal.

It does not have to be zero. In a common base stage, the voltage at the base is the reference ('common' ) for the input signal at the emitter. But that does not mean it can not have an independent input signal so that the final output is a function both of the voltage at the base and the current at the emitter. It is still a common base stage.

Jan
 
Disabled Account
Joined 2014
It's funny to see how people that obviously don't understand the difference between VFA and CFA are trying to tell everybody that what they are saying is the "True news" and everybody else are telling the "Fake news".:eek:
It's in fact impossible to argue with such people, so I'm not going to try.
Have fun:rolleyes:
Cheers
S
 
I think some of the objection to "cfa" is that the name doesn't fit the existing formal classification schemes - as a Voltage output device it is the Voltage Output that is sampled by the feedback

Exactly so. I did not go further than Mr. Jung rebuttal; it appears to me he is making a few mistakes:

- Nobody appears to debate 3 or more decades of successful CFA applications. In fact the CFA topology, even if it doesn't require any special instruments to analyze (it very well falls within the 4 feedback types known since Black) deserves a separate denomination, to conveniently talk about the special properties of this circuit topology.

- Mr. Jung talks about "the unique CFA circuit topology" which exactly correct, but doesn't help the cause. The CFA is indeed a particular circuit topology, but not a feedback type, as inferred by the CFA denomination.

- Mr. Jung talks about the "fast dynamic, due largely to the current-on-demand behavior of the CFA input stage". Correct, but I've personally shown several implementation of the current-on-demand behavior in what it would be universally accepted as a "VFA". One of which is the Stochino amp, with only one extra capacitor required to implement this much praised feature. Another one is a standard CFA topology with the inverting input buffered, which transforms it in a pure "VFA" while keeping the current-on-demand property, see the LM6171, etc...).

- Mr. Jung talks about the "relative constant closed loop bandwidth, regardless of the closed loop gain" which is again correct, but is once again the property of a particular circuit topology (which could be also called "diamond buffer input stage topology" instead of "CFA", but which would be of course way to complex for the marketing departments). It would be also correct and relevant to add "for a certain range of closed loop bandwidths". The property holds indeed for closed loop gains of, say, 0...12dB but no longer holds significantly for closed loop gains of, say, 30...40dB.

- I absolutely agree with MK that the CFA topology is simply moot when it comes to audio, it is just another buzzword. Set aside that only a few years ago, the CFA topology was chastised right here on DIYAudio as having a "horrible screechy sound", the usual high closed loop gain required in audio amplifier (20-30dB) makes any CFA topology advantages blur away, while the high slew rate (as a result of the current on demand property) is simply useless. Nobody was ever able to prove mathematically or experimentally, under controlled conditions, that more than 1V/uS per 1V of peak (audio) output is required.

However, I fully understand the need to concisely denominate as "CFA" the unique properties of this particular circuit topology, and that those properties are indeed very useful in certain circumstances. Beyond the marketing departments, even competent engineers are using it as a reference to it's properties, rather than kind of fundamental paradigm in feedback theory and practice.

If I wouldn't be a lazy MF, or if I would have more interest in audio beyond that of a periodic lurker, I would probably send to AudioXpress a rebuttal to rebuttal but otherwise, being both, I'll leave it here on DIYAudio.
 
Last edited:
I looked at a bunch of Comlinear datasheets but they don't have internal schematics.

_

It so far predates Comlinear this is a waste of time. The current feedback name was applied to some amplifiers that had some uniquely different properties, time for people to let go of this.

EDIT - @Waly those skilled in the art design circuits to achieve desired results and generally know all this stuff and mix and match topologies to get what they want. I'm not sure where this controversy arose but I think it was from folks with narrow ultra-academic ways at looking at these things, i.e. why do they care?

I shudder to think of the results, how droll, 100's of millions of $ of ADSL drivers shipped of so called CFA's oh the poor duped customers.
 
Last edited:
I'm not sure where this controversy arose but I think it was from folks with narrow ultra-academic ways at looking at these things

No, it arose from those calling the CFA the best thing from sliced bread, when it comes to audio reproduction. And from those claiming that the CFA is some sort of secret sauce to audio performance. And from those believing a "CFA" somehow allows more loop gain to linearize an amplifier with a certain GBW (without any stability margins penalty).

BTW, because somebody else mentioned single ended CFAs, the Curl standard topology is exactly a "CFA" and can be analyzed as such. Just an example of yet another "CFA" topology, that AFAIK not even the author ever officially mentioned it as such (although he could rightfully do so), too bad the audio review chimps were at the time not overly excited about "CFAs".
 
Last edited:
No, it arose from those calling the CFA the best thing from sliced bread, when it comes to audio reproduction.

Except those folks don't have enough of a clue to mount an interesting argument/discussion. I was talking about the academic definitions of series/shunt feedback etc. that some flog endlessly. The fact that some of the proponents of CFA's here wouldn't know a series from a shunt is beside the point.

BTW my Laibach Divided States of America DVD happens to be sitting about a foot to my left (by sheer coincidence of course). I wonder if Dick would consider it world music.
 
Last edited: