Bob Cordell's Power amplifier book

While I agree that what's found inside a casual CF IC opamp (which is mainly optimized for speed) is of little use for an audio amplifier, the issues can be addressed.

Like here. Beats hands down (in specs) many classic VFA opamps, including at voltage noise, PSRR, etc... before even getting to the speed specs (like the ridiculous 1900V/uS). Ok, the current noise is high-ish.

It would be interesting if you could get this device through your standardized test bench.
 
Last edited:
The few monolithic so-called "CFAs" that use two stages of gain also use minor loop compensation which loses the much vaunted "speed".

Any feedback amplifier needs compensation, independent of number of stages or feedback type. And the AD8001 (which is a two-stage CFA) is still "high-speed" at 880 MHz/1200 V/us. It's not low distortion, but as I noted this could be achieved with appropriate circuitry.

I fail to appreciate how and why the shortcomings of these miserable concoctions can be improved for audio.

So it would be very interesting to see Bob writing something up, wouldn't it?

Samuel
 
Beats hands down (in specs) many classic VFA opamps, including at voltage noise, PSRR, etc... before even getting to the speed specs (like the ridiculous 1900 V/uS).

But distortion performance is relatively poor. Better than many standard CFAs, but not nearly as good as a decent VFA.

It would be interesting if you could get this device through your standardized test bench.

Unfortunately not possible, as the trick with high noise gain would need a bit adaptation (another resistor in the feedback network).

Samuel
 
The few monolithic so-called "CFAs" that use two stages of gain also use minor loop compensation which loses the much vaunted "speed".

I fail to appreciate how and why the shortcomings of these miserable concoctions can be improved for audio.🙁

Isnt it ironic then that many of the so-called best amps in the world are cfb designs. It is also ironic that most of these designs originate from your homeland which shows that many dont agree with your view point. I know at least 4 major high end audio companies that literary only use CFBs types.

Like the blameless concept CFB types can be tweaked for higher performance and higher loop gains closer to that of VFB can be attained. Even using TMC or TPC in blameless type amps cannot match the THD performance that can be obtained at frequencies of 5 khz and up by a well designed CFB amp with less complication.
 
But distortion performance is relatively poor. Better than many standard CFAs, but not nearly as good as a decent VFA.



Unfortunately not possible, as the trick with high noise gain would need a bit adaptation (another resistor in the feedback network).

Samuel

I dont think there is any slew rate penalty in two stage CF amps, one only needs to look at AD8011 which is a typical example of the topology, it achieves slewrate of 3500v/us. There is a complexity penalty but then again there is huge bandwith gains.

If one wants to compare performance that can be achieved with CFB topology it is wise to look at TPA6120.
 
Last edited:
20 to 30 ppm at 20 kHz at full power (300 Watts into 8 ohms) seems relatively easy to achieve with CFA. Bandwidth and rise/ fall times superior to VFA.

So, what precisely is it that makes CFA unsuitable for audio? 😕
Michael for one keeps complaining he can't get enough feedback with a CFA topology, and that he can't use TMC or TPC (to get even more feedback).

Actual performance of the amplifier seems to be a secondary consideration. Pretty much a "feedback for feedback's sake" kind of requirement.

Plus some nonsense about 2'nd harmonic distortion. Apparently he hasn't figured out that an input stage like the one in your nx-Amplifier cancels even order distortion just as well as a long tailed pair.
 
I second that. I was reading it again yesterday and came away feeling again informed. Most importantly, it was nice to have information as opposed to just a series of opinions. I would also welcome some more info on common source output stages and more importantly, a more thorough practical exposition on the optimization of input LTPs. Thank you.

Writing down one's knowledge alongside measurements/evidence is different from simply writing down one's knowledge even if the person is right. It is sort of like the open-source software community. Anyone can check the analysis.

I think few of us stand blameless for doing something different. After all, a LOT of discussion goes on and running the analysis for each and every reply would consume a lot of time. However it may save more time than it takes because one wouldn't need to be driven mad before they show their analysis.

For scientific progress to be made, one has to prioritize scientific methodology over secondary considerations such as personal effort applied. IE, both sides of a discussion need to produce schematics and analysis, so their arguments can be verified. Unfortunately we shoot ourselves in the foot when we leave this job up to book writers; community-based progress can be so much faster than waiting for the next massive tome. It can consistently be ahead of the documentation of the time.

The problem in many cases I think is a mindset of economy over scientific progress. I think many are afraid of their ideas being improved on by others and "taken". There seems to be the perception that if one presents their knowledge as opinions without evidence, there is a degree of protection provided by plausible deniability and skepticism. Furthermore, most people seem to have some form of food anxiety, time anxiety, money anxiety, and so on. With all these restrictions, one tends to be either frozen with fear or applying all their thought to conserving resources, and in this condition it is difficult to sit down and allow a curiosity-driven, creative, nonlinear thought process to play out. Putting down these fears and getting some real work done is highly commendable, but at the same time there are pathological conditions where one does not feel anxiety simply because they cannot - and while this person may make a lot of progress due to slackened instincts, their pathological guidance is often able to mislead a great many people that look up to them due to their "guru" status. Such pathology is NOT as uncommon as is commonly believed, especially where there is a degree of insulating anonymity.

I just want to throw this in for consideration. Ultimately, we are human, and we are subject to personal motives unknown to even us because of our unconscious mind. I don't think a person can truly be scientific or objective without consistent introspection. I think proactive introspection is critical to scientific progress; without introspection we simply keep making the same mistakes over and over again. Applying scientific principles requires rigor, and rigor cannot be accomplished without introspection. I think in the scientific community at large there is a lack of proper mental hygiene, and this has resulted in the slackening of scientific standards, methods and even understanding of same.

I suggest that we develop a model for the presentation of engineering/design knowledge that separates evidence, analysis, conclusions, and speculations into different categories, and then use this model to drive the expansion of a wiki. It would be like a wiki where each page reads like a research paper. There can still be sections for anecdotal and unverified information, but they will not be mixed with verified information and analysis.
 
Writing down one's knowledge alongside measurements/evidence is different from simply writing down one's knowledge even if the person is right. It is sort of like the open-source software community. Anyone can check the analysis.

I think few of us stand blameless for doing something different. After all, a LOT of discussion goes on and running the analysis for each and every reply would consume a lot of time. However it may save more time than it takes because one wouldn't need to be driven mad before they show their analysis.

For scientific progress to be made, one has to prioritize scientific methodology over secondary considerations such as personal effort applied. IE, both sides of a discussion need to produce schematics and analysis, so their arguments can be verified. Unfortunately we shoot ourselves in the foot when we leave this job up to book writers; community-based progress can be so much faster than waiting for the next massive tome. It can consistently be ahead of the documentation of the time.

The problem in many cases I think is a mindset of economy over scientific progress. I think many are afraid of their ideas being improved on by others and "taken". There seems to be the perception that if one presents their knowledge as opinions without evidence, there is a degree of protection provided by plausible deniability and skepticism. Furthermore, most people seem to have some form of food anxiety, time anxiety, money anxiety, and so on. With all these restrictions, one tends to be either frozen with fear or applying all their thought to conserving resources, and in this condition it is difficult to sit down and allow a curiosity-driven, creative, nonlinear thought process to play out. Putting down these fears and getting some real work done is highly commendable, but at the same time there are pathological conditions where one does not feel anxiety simply because they cannot - and while this person may make a lot of progress due to slackened instincts, their pathological guidance is often able to mislead a great many people that look up to them due to their "guru" status. Such pathology is NOT as uncommon as is commonly believed, especially where there is a degree of insulating anonymity.

I just want to throw this in for consideration. Ultimately, we are human, and we are subject to personal motives unknown to even us because of our unconscious mind. I don't think a person can truly be scientific or objective without consistent introspection. I think proactive introspection is critical to scientific progress; without introspection we simply keep making the same mistakes over and over again. Applying scientific principles requires rigor, and rigor cannot be accomplished without introspection. I think in the scientific community at large there is a lack of proper mental hygiene, and this has resulted in the slackening of scientific standards, methods and even understanding of same.

I suggest that we develop a model for the presentation of engineering/design knowledge that separates evidence, analysis, conclusions, and speculations into different categories, and then use this model to drive the expansion of a wiki. It would be like a wiki where each page reads like a research paper. There can still be sections for anecdotal and unverified information, but they will not be mixed with verified information and analysis.

Kean, I am completely with you.
 
Plus some nonsense about 2'nd harmonic distortion. Apparently he hasn't figured out that an input stage like the one in your nx-Amplifier cancels even order distortion just as well as a long tailed pair.

EVEN order harmonic cancellation, if you please, and, NO, Bonsai's nx amplifier's input stage is incapable of cancelling even order harmonics like a current-mirror loaded diff. pair.:smash:
 
Last edited:
20 to 30 ppm at 20 kHz at full power (300 Watts into 8 ohms) seems relatively easy to achieve with CFA. Bandwidth and rise/ fall times superior to VFA.

So, what precisely is it that makes CFA unsuitable for audio? 😕

I am thinking 10ppm THD or less at 20KHz. Diff. input stage VFAs give more than adequate bandwidth, slew rate and small signal rise and fall times if properly designed.
 
No. :smash:
No:smash:
And No :smash: again

It sims at 20 to 30 ppm at full power at 20 kHz.

It has adequate gain and phase margin (would I design an amp that is marginally stable? C'mon!)

Bandwidth and slew rate are manifestly better than VFA, as would be expected.

I will build it, and write it up in the next few months (after my new pre is completed).

And Waly is absolutely right, the difference between 20ppm or 30ppm and 10 ppm distortion has no bearing on the perceived sound. Zero.

However, a little AFEC may also be applied to get it down to 2-3 ppm at full power 20 kHz. 😉
 
How about cancelling all harmonics? That seems so much better than just cancelling half of them.

Bear in mind that the cascomp developed by Pat Quinn at Tek deals with this issue in a rather elegant way. The cascomp is covered in my book. I also show how the cascomp principle can be applied to JFETs. I seem to recall that the cascomp has a 3-dB noise penalty associated with it, however.

Pat's mentor at Tek was Barrie Gilbert.

Cheers,
Bob

The Cascomp seems to be one way, but maybe it could probably be done with less active components.

Is there a circuit like this for the VAS? For the output stage?
 
Bob, please take time out in your new edition to cover CFA's in some depth. Your unbiased, un-opinionated and practical writing style is a breath of fresh air. I for certain will buy the second edition to sit alongside my 1st.

Thanks for your very kind words, Bonsai. I will be the first to admit that I have not studied or built any CFA power amplifiers. So at this point I'm not in a strong position to comment on them, and would also have to think carefully what kind of power amplifier I would call a true CFA.

I do have some previous experience with what I think people would agree are CFAs. My first job at Bell Labs in 1971 was to design equalizers for the transmission of Picturephone signals. We needed to equalize a mile of twisted pair to within about 0.1dB out to 1 MHz, where the loss at 1 MHz was about 60dB. The equalization was adaptive with the use of a pilot tone at just above 1 MHz. The fixed equalizers were CFAs with a feedback network that was a large combination of R's and C's, and we depended heavily on the property of the CFA that the feedback factor could be made manageable over a large equalization gain range.

A later experience was with a Class H DSL line driver which, if I recall, involved a CFA architecture as part of it. I suppose that architecture would be something like what would lead to a power amplifier that uses a CFA architecture.

In any case, I'd have to give some thought to it to identify whether it really brought anything to the table in order to decide whether to educate myself enough to include a discussion on it in the second edition.

If pursuing CFA for higher slew rate alone, I would probably not see that as needed value-added. For example, the 50-watt MOSFET power amplifier with error correction that I reported on in 1983 achieved 300 V/us, which I think is more than adequate for audio.

Cheers,
Bob