Once again and for all:
- There are exactly four types of negative feedback, four combination of output sampling and input adding: http://cas.ee.ic.ac.uk/people/dario/files/E22/L3-feedback amplifiers.pdf
- For all purposes, a "CFA" is series (voltage input) - shunt (voltage output) feedback. Which means the feedback network measures the output voltage and feeds back to the input a voltage correction signal.
- This type of feedback makes the amplifier a better voltage amplifier, since it increases it's input impedance and decreases it's output impedance. Typical example: any opamp in non-inverting configuration.
- A "current feedback amplifier" from this perspective is an amplifier that has input shunt connection, since such a connection would feed back to the input a current correction signal.
- This type of feedback would make a better transimpedance amplifier, since it would decrease the output impedance and decrease the input impedance. Typical example, any op amp in inverting configuration. Yes, that is "current feedback" at the input.
Etc...
- The above classification, known and used for a century, makes no assumptions on the amplifier input impedances, so the fact that a "CFA" circuit topology has a low impedance at the inverting input is totally irrelevant. If the feedback network feeds back a correction voltage at the input (as it does) it's still voltage feedback.
- Which makes the "CFA" voltage feedback all day and night. The fact that the circuit topology (which again, the feedback theory makes no assumptions about) has some special property which may be conveniently (but not mandatory!) analyzed using current summing instead of voltage summing, and that the same topology has certain properties (current on demand, etc...) does not change the feedback type.
- Which makes the "CFA" circuit topology a marketing driven naming convention. It does not refer, by any means, to the circuit feedback type, but to the rest of the particular circuit topology properties, which are indeed remarkable for a certain class of applications (not including audio, of course).
- There are exactly four types of negative feedback, four combination of output sampling and input adding: http://cas.ee.ic.ac.uk/people/dario/files/E22/L3-feedback amplifiers.pdf
- For all purposes, a "CFA" is series (voltage input) - shunt (voltage output) feedback. Which means the feedback network measures the output voltage and feeds back to the input a voltage correction signal.
- This type of feedback makes the amplifier a better voltage amplifier, since it increases it's input impedance and decreases it's output impedance. Typical example: any opamp in non-inverting configuration.
- A "current feedback amplifier" from this perspective is an amplifier that has input shunt connection, since such a connection would feed back to the input a current correction signal.
- This type of feedback would make a better transimpedance amplifier, since it would decrease the output impedance and decrease the input impedance. Typical example, any op amp in inverting configuration. Yes, that is "current feedback" at the input.
Etc...
- The above classification, known and used for a century, makes no assumptions on the amplifier input impedances, so the fact that a "CFA" circuit topology has a low impedance at the inverting input is totally irrelevant. If the feedback network feeds back a correction voltage at the input (as it does) it's still voltage feedback.
- Which makes the "CFA" voltage feedback all day and night. The fact that the circuit topology (which again, the feedback theory makes no assumptions about) has some special property which may be conveniently (but not mandatory!) analyzed using current summing instead of voltage summing, and that the same topology has certain properties (current on demand, etc...) does not change the feedback type.
- Which makes the "CFA" circuit topology a marketing driven naming convention. It does not refer, by any means, to the circuit feedback type, but to the rest of the particular circuit topology properties, which are indeed remarkable for a certain class of applications (not including audio, of course).
Last edited:
My comments relate to an observation and behavior that makes the CFA nomenclature more IMHO than some kind of marketing "fraud" no more no less. The "particular circuit topology" is all that matters to the performance to a set of specifications.
Taken literally the base resistance in the feedback network would not matter i.e. 1k and 1k for G=2 would give the same result as 10k and 10k. This is obviously not the case.
Still looking for a proof that a set of performance specs can not be achieved by either configuration, otherwise this whole argument is moot and frankly a waste of time.
Last edited:
Try to build a CFA with <1nV/rtHz noise, that would be interesting. Also ultra low offset and offset drift.
Of course, in the same process for an equivalent VFA. I wonder if a chopper "CFA" would make any sense.
P.S. I see one 0.9nV/rtHz LMH6554 LMH6554 datasheet 2.8 GHz Ultra Linear Fully Differential Amplifier | TI.com but with 11pA/rtHz, that's heroic... The CFA input buffer adds noise voltage to the input, without any voltage gain. The LMH6629 VFA has 0.69nV/rtHz and 2.6pA/rtHz.
Last edited:
Obviously so. But the classic feedback theory doesn't make any assumptions about this, it's all about what's sampled and what's added as a correction signal.
some light observations ---
The internal topology allows for very high SR. Much higher than is needed for audio. However (there is always a- however), When compared to Volt-Mode operation topologies set for a guess of .5-1v/usec and compared to CMA, the CMA sounds better. Maybe why the top models of nearly all competitive brands are CMA -- mislabeled IMO as CFB. So, we went off and compared. SR does seem to need a spec update --- .5-1v/usec is too slow. But 1000v/usec isn't needed either. The New and higher number should be maybe 5-20 times higher to eliminate distortion creep due to approaching SR limiting. So, either mode of operation Volt or Current - mode, if updated seems to sound very similar.
Regarding the ease of compensation of a CMA --- a simple comp cap is all that is needed for stability. No that is Not maxing out or optimizing the amp's performance at the freq extremes, but the CMA has such wide BW OL that it isn't an issue for an amp handling LF audio. Thus, the comment that it is easier to comp.
THx-RNMarsh
Last edited:
If it looks like a CFA, works like a CFA and smells like a CFA, it is a CFA.
If you can make a functioning audio amplifier with CFA topology (and there are hundreds of exemplars out there reaching back 30 years or more) then its one of a number of possible solutions is it not.
To paraphrase your position (and Mikeks) ''CFA's for audio amplifiers is nothing but marketing nonsense"
Again, a completely unsubstantiated claim no better than trying to say there's no such thing as a CFA.
There are many other fine engineers out there that you could look up to . . .
If you can make a functioning audio amplifier with CFA topology (and there are hundreds of exemplars out there reaching back 30 years or more) then its one of a number of possible solutions is it not.
To paraphrase your position (and Mikeks) ''CFA's for audio amplifiers is nothing but marketing nonsense"
Again, a completely unsubstantiated claim no better than trying to say there's no such thing as a CFA.
There are many other fine engineers out there that you could look up to . . .
Last edited:
I said save noise two days ago. I thought we were talking PA's for line stages and pre-amps CFA's are silly.
To aid the discussion, here is the stuff I put up about 4 years ago that compared both topologies. note, no claims made, just sims to highlight the strengths/weaknesses of the topologies. Both circuits very simple and 10 transistors each (VFA shows a current source - that would be a 1 or 2 transistor current source - neither here not there for demonstration purposes.
Moral of the story: if you want ultra low distortion and high slew rates from either topology, you are going to have similar performance characteristics in terms of loop gain profile and you are going to have to tame the OPS phase shift which will dominate your compensation design efforts.
Moral of the story: if you want ultra low distortion and high slew rates from either topology, you are going to have similar performance characteristics in terms of loop gain profile and you are going to have to tame the OPS phase shift which will dominate your compensation design efforts.
I don't mean that I am saying across the FB resistor, either. ... other ways that are simple. C across FB R in CMA doesn't work. At best you'll get a huge peak in FR... at worse it will osc.
-RNM
Last edited:
If you normalize both for low distortion and high slew rates, then yes both will be essentially the same. In the consumer world, they have in the past (present?) Not been of equal. Then they may sound different.
Maybe my designs are not the norm? I have such wide BW OL in CMA (CFB) that the phase margins needed to be stable is easy to do even if a bit heavy handed ...... yes, I loose a lot of potential BW but has not mattered, I have a lot that can be thrown away and still have ultra low distortion at high audio freqs.
??
-RM
Last edited:
Sorry, I missed that. If you refer to audio AC applications, then I 1000% agree. There's nothing special that can't be achieved with either "CFA" or "VFA".
Only that CFAs sound better, according to some TTYH experts. That CFAs (and in particular the Alexander amp) were chastised a few years ago as having a "horrible screechy sound" is of course irrelevant.
Only that CFAs somehow provide "a constant loop gain" across the audio band, according to some other experts, like that would ever matter.
Only that CFAs somehow allow more loop gain, therefore improving the distortion numbers over a VFA, according to yet other experts.
Only that CFAs potential mega slew rates are improving the sound, according to the golden ear brigade experts.
Only... for all of the above and more fetish stories, links are available upon request. The diyaudio search engine could also help.
have you internal schematic of TPA6120? or something else what is closer to this IC?
