If this remark is addressed to me, I do not think to have been ambiguous to merit it.
More arguments for the discussion, by Franco, and in the comments, by Feucht and by Aylward
Quest for the Ideal Transistor? | EDN
I've read the article by Sergio Franco and quote from it as as follows:-
"If you ground the input (Vi = 0), break the loop at the CFA's output, and inject a test signal into the feedback network to see what comes back to the CFA's inverting input, you see just a current - no voltage is fed back, as the input buffer keeps Vn fixed at 0 V. So, you can only refer to this state of affairs as "current feedback". " (comments referring to fig. 6B in his article)
If that's not current feedback, I don't know what is.
Secondly, this from Kevin Aylward
"The fundamental reason for fast large signal response of a standard CFA is because it has a class AB input stage, which allows for fast charging of its compensation capacitor. Period. It has absolutely nothing to do with its use of current feedback, which buys nothing in terms of speed as it still has to swing large voltages with high gain. A class A input stage current feedback amp would be just as slow as a standard voltage feedback amplifier."
Its absolute nonsense. The question of the CFA front end being class AB came up in the original thread on this subject about 4 years ago, and it was claimed that the CFA going into class AB operation in the small signal regime was bad for audio. I simulated and quickly realised it was very easy to keep the front in in class A operation all of the time. Secondly, the charging rate of the compensation capacitor is set by the feedback operation - nothing to do with class AB operation. My sim showed peak charging currents - while still in class A - of 8x the standing current and there would be no problem increasing this figure.
Dennis Feucht is an exceedingly capable analog designer, but he has not produced an argument that sweeps away Franco's description of the circuit, or those that have existed in semiconductor literature for the last 30 years - they still stands IMV. As for Aylward, no further comment (Is this our dear Waly by any chance?).
This whole CFA/VFA battle - which I assume started out here on DIYAudio 4 years ago since I cannot find any prior references to this dichotomy of views anywhere - seems to have been given wings by Mikeks and is now poisoning the wider analog design community.
"If you ground the input (Vi = 0), break the loop at the CFA's output, and inject a test signal into the feedback network to see what comes back to the CFA's inverting input, you see just a current - no voltage is fed back, as the input buffer keeps Vn fixed at 0 V. So, you can only refer to this state of affairs as "current feedback". " (comments referring to fig. 6B in his article)
If that's not current feedback, I don't know what is.
Secondly, this from Kevin Aylward
"The fundamental reason for fast large signal response of a standard CFA is because it has a class AB input stage, which allows for fast charging of its compensation capacitor. Period. It has absolutely nothing to do with its use of current feedback, which buys nothing in terms of speed as it still has to swing large voltages with high gain. A class A input stage current feedback amp would be just as slow as a standard voltage feedback amplifier."
Its absolute nonsense. The question of the CFA front end being class AB came up in the original thread on this subject about 4 years ago, and it was claimed that the CFA going into class AB operation in the small signal regime was bad for audio. I simulated and quickly realised it was very easy to keep the front in in class A operation all of the time. Secondly, the charging rate of the compensation capacitor is set by the feedback operation - nothing to do with class AB operation. My sim showed peak charging currents - while still in class A - of 8x the standing current and there would be no problem increasing this figure.
Dennis Feucht is an exceedingly capable analog designer, but he has not produced an argument that sweeps away Franco's description of the circuit, or those that have existed in semiconductor literature for the last 30 years - they still stands IMV. As for Aylward, no further comment (Is this our dear Waly by any chance?).
This whole CFA/VFA battle - which I assume started out here on DIYAudio 4 years ago since I cannot find any prior references to this dichotomy of views anywhere - seems to have been given wings by Mikeks and is now poisoning the wider analog design community.
I understand your urge to celebrate a "victory" over the infidels, but by calling Kevin Aylward's statements "nonsense" and Dennis Feucht "not producing any argument" you are crossing the limits of common sense. And you dare to call others for behavior!
BTW, it's not over yet. The debate is now way over what can be accommodated on diyaudio (or any other public forum), watch EDN for follow ups.
BTW, it's not over yet. The debate is now way over what can be accommodated on diyaudio (or any other public forum), watch EDN for follow ups.
Stay honest Waly. He backed up his statement by factual arguments, so he was factually correct to call it nonsense.
Stay honest Waly. He said 'Dennis Feucht not producing any arguments to sweep away SF's arguments', which is factually correct.
You running out of tech arguments Waly? That's no excuse for crossing the lines of basic honesty.
Jan
There are no nonsense arguments and, as much as you would like, nothing is just black and white. Have you bothered to read and understand Kevin's argumentation or, given the prof. EDN article, you believe it's not worth of?
Have you bothered to read what Dennis Feucht says about CFAs or, given the prof. EDN article, you believe it's not worth of?
Yes, I'm myself running short of technical arguments for the use of this public forum. Arguing with a professor that can't reply, with the background noise provided by amateur and fan club crowds, is anything but productive, so watch EDN for follow-ups.
Meantime, what about you checking out some famous CFA amplifiers and see what is the Rn value and if the prof. criteria is always met? And I would appreciate if somebody could point me under which of the 4 types of feedback does the CFA topology fall under.
What I DO note is that 'the prof' etc DO give arguments and facts and figures, while you can't apparently get any further than asking if I read and studied the stuff.
If you think they are wrong, how about YOU provide facts and figures to that?
I will watch EDN (am a subscriber) but unless you get anything more solid than 'victory', 'not over yet', 'infidels', I doubt if it will be worth my time, except possibly as entertainment.
BTW, we all already agreed that the CFA term in this opamp context is different than the '4 feedback configs' context. You are right, it is different, it is a deviation, it will not change, deal with it. Very weak argument Waly.
Jan
If you think they are wrong, how about YOU provide facts and figures to that?
I will watch EDN (am a subscriber) but unless you get anything more solid than 'victory', 'not over yet', 'infidels', I doubt if it will be worth my time, except possibly as entertainment.
BTW, we all already agreed that the CFA term in this opamp context is different than the '4 feedback configs' context. You are right, it is different, it is a deviation, it will not change, deal with it. Very weak argument Waly.
Jan
And how would you suggest doing this on the text forum format? If there's any way to do this that I am not aware of, I'm full in for it.
Meantime, I'm still waiting for your calculation regarding the Rn in various CFA audio power amplifiers. You will have a big surprise.
I find it funny people claiming having me on their ignore list, but still going through the trouble of replying to my posts.
I posted up in this thread a few days ago the canonical feedback forms showing how they relate to VFA and CFA. You can build current controlled output or voltage controlled output circuits with both CFA's and VFA's.
An inverting VFA is not a current feedback amplifier (whether the controlled output is a current or a voltage) since [ideally] no current flows into the - input. A CFA is a current feedback amplifier since the feedback current flows into the - input just as SF has explained.
An inverting VFA is not a current feedback amplifier (whether the controlled output is a current or a voltage) since [ideally] no current flows into the - input. A CFA is a current feedback amplifier since the feedback current flows into the - input just as SF has explained.
Last edited:
How about a blog or article like the other serious people?
I am not going to do that. I read SF's reasoning which seems sound to me. If you find it not so, please give your version.
Irrelevant.
Jan
🙂 I started it here. So beat me up. I wondered why so many of the mfr used CMode amps in the top level products and that most people thought they sounded better....and what was the reason for that......my own circuit years ago (1970's) was prior to IC use and was the beginning of that movement. I only wanted to eliminate capacitors with a direct coupled design but ended up with a current mode operation. I don't take credit for the new mode of operation... it was a nice accident though. I wanted to eliminate the coupling caps and make a more perfect amplifier.
THx-RNMarsh
Last edited:
That may happen soon. You will be required to read prof. EDN article using paper pen and pencil. I'm afraid would it not be signed "prof" you would be much more circumspect about the conclusions.
Frankly, I don't think you can. But surprise me.
You offend me. It must be because obviously you don't know me.
Jan
Preview: One of the error sources that Prof. is correctly handling is the finite output impedance of the buffer. Yes, that only can't in general account for the differences, although for the audio CFAs that's good enough.
The other error source which Prof. did not handle is the assumed x1 gain in the input buffer.
Prof said:
Which is correct, as long as you don't consider that a real world CFAs X1 buffer does NOT have Vd=Vi-Vn=0 but likely exactly about the 2.5mV delta required by the VFA model (that is, a gain of 0.9975). They are emitter follower and the x1 gain is only an approximation, which is also depending on the current. And this approximation is only getting worse at HF.
I'm sorry, I can't get more into the details here, again, watch the Internet. I'm not in the blogging, sorry again, life is to short to filter out spam.
So far I see only agreement, which I guess is progress. IIRC SF did also mention the not-equal-to-exact-one gain?
Jan
Jan
It is not global feedback, the loop is broken.
The circuit can play the role of a non-feedback virtual earth.
Approximately.
The original design used 4 bipolar transistors and was extremely linear..... I used +/- 24vdc supplies and had to get the output to near clipping (22v) to get above the HP339A residual of .0018% THD+N. No high z values used in circuit ... only 100 ohms and 1K ohm. My generator only went to 20MHz and the output was still going strong and stable with load. There was no circuit like it prior to publication. I got a phone call from engineer at Comlinear who asked a lot of questions because he revealed they had been working on same design or similar and were clearly upset because they were going to patent it. Comlinear made an important change and got the patent in 1982. This is covered in my Linear Audio headphone amp article. Since then.... diamond I/O has been added as well and more refinements..... all of which we discuss here now. The newest designs (esp cmos) continue to evolve far out into RF+. I have my eye open and mouth shut. But I know when someone here is off base and a lot of it is. Thus I give a clue/comment and suggest some books.
From a design which started as a problem solver (eliminating electro coupling caps) we are here today with advance designs in IC packages.
However, for audio, simpler often seems to get best results in cost, reliability and performance and the 'old' basic design still works well. That HPA with very well matched transistors produces thd below AP 2722 analyzer.
In another forum, the pursuit of distortion measurements was developed out of my need to be able to measure this HPA. As it evolved and we got to ever lower detectable levels, I was able to trim the circuit values to get further distortion reductions..... now I have a system of new gen sources and test gear capable of measuring accurately harmonics to -160dbv. Another case of need pushing designs.
THx-RNMarsh
From a design which started as a problem solver (eliminating electro coupling caps) we are here today with advance designs in IC packages.
However, for audio, simpler often seems to get best results in cost, reliability and performance and the 'old' basic design still works well. That HPA with very well matched transistors produces thd below AP 2722 analyzer.
In another forum, the pursuit of distortion measurements was developed out of my need to be able to measure this HPA. As it evolved and we got to ever lower detectable levels, I was able to trim the circuit values to get further distortion reductions..... now I have a system of new gen sources and test gear capable of measuring accurately harmonics to -160dbv. Another case of need pushing designs.
THx-RNMarsh
Last edited: