Michael,
I wonder why you think that your simulations are better than Bob's simulations, they are both after all just that? Simulations only seem to be as good as the information that you are entering and it is easy it seems to leave out parasitic values and other things that are not so obvious. What is wrong with SPICE, it is fairly universally used as a simulation program, but I do understand that many standard library models are incomplete?
I wonder why you think that your simulations are better than Bob's simulations, they are both after all just that? Simulations only seem to be as good as the information that you are entering and it is easy it seems to leave out parasitic values and other things that are not so obvious. What is wrong with SPICE, it is fairly universally used as a simulation program, but I do understand that many standard library models are incomplete?
Hi Mike,
Nice to have you back. Although it can be painfull, we often learn much from a debate with another very smart person. You are indeed a smart person and I have learned from debating you.
We have to agree to disagree on Miller Input Compensation (MIC). Yes, it usually refers to a compensating capacitor from the collector of the VAS to the input of the input stage. It is obvious that the connection formes a Miller integrator. You need to understand that a Miller integrator can include more than one stage inside its loop. With respect to pole-splitting, there is indeed pole splitting between the input node (here the input to the input stage) and the output node (the output of the VAS). The output is now at a quite low impedance.
However, there is a caveat with respect to pole-splitting, which I explained in my 1983 amplifier paper and in my book. That is, that the Miller loop so-formed is a feedback loop that itself must be compensated, albeit lightly. This loop can have a gain crossover frequency well above that of the amplifier, but it must nevertheless be compensated for stability. That is the purpose of the series RC across the collectors of the differential pair.
Miller Input Compensation it is.
Cheers,
Bob
Mike,
You are simply out of phase and contrarian to widely accepted ways in our business of describing things. This is perfect example of how you play minor semantical ambiguities into something that adds nothing useful to a discussion.
Cheers,
Bob
Hi Mike,
You need to understand that what you think is a "proof" is not always accepted to be that by others of equal experience and intellect. People can prove many things with the right set of wrong assumptions. Those with an agenda are even more adept at doing so. One need look no further than a typical government.
SPICE simulations are far from "virtually useless" for evaluating relative distortion performance, especially when you use good models for the transistors. Poor transistor models have often admittedly been a source of some inaccuracy, at least in a relative sense. It is certainly true that SPICE distortion simulations have their limitations compared to a real-wired-up amplifier, not the least source of which is a thousand real-world parastics and couplings that are not usually in the SPICE simulation. But many of those will differ from amplifier implementation to amplifier implementation of the same circuit. Make no mistake, one should be cautiois in depending too heavily on absolute distortion results from a simulation. In the end, there is no substitute for measurement and listening tests. In this day and age, NOBODY, in my opinion, should design an amplifier without SPICE simulation as part of the process. It is remarkable that you seem to view SPICE simulation with such disdain.
I'm sorry that it is very disappointing that a person with such "learning and experience" disagrees with you. Newsflash: There are many others.
Cheers,
Bob
Are you even aware that a current feedback amplifier must necesarily have series derived negative feedback?
See below for the basics:
http://en.wikipedia.org/wiki/Negative_feedback_amplifier
Last edited:
A question is posted in my contribution fT/ic ( see thread in this forum).
We think that linear amps will soon all be replaced by "switching" ( class D) simply because
the availability of "audio power BJts" will dwindle.
In Class D, timing error as well as noise become distortion.
With e.g. Hawksford error compensation (feed forward, not feedback) some of these class D errors can be compensated with a low power super fast class A amp.
A close to perfect compensation is not easy to achieve neither in theory nor in practice.
Class D amps are pretty much cheaper to produce than linear amps with same power and response. Insofar error-forward corrected class D are a nice idea for full active speakers avoiding problems with reactive loads and saving cost of passive crossover parts ( some of these such as special inductors tin foil capacitors etc. cost more than a 200 watts class D with switching power supply!)
If so there is another question, limiters.
This applies to subwoofers and in general bass speakers as the radiation resistance of the moving cone drops with frequency such that cone excursion must be limited.
The limiter problem can be solved using readily available digital processors , but not so easy if one wants to process the PWM signal.
We think that linear amps will soon all be replaced by "switching" ( class D) simply because
the availability of "audio power BJts" will dwindle.
In Class D, timing error as well as noise become distortion.
With e.g. Hawksford error compensation (feed forward, not feedback) some of these class D errors can be compensated with a low power super fast class A amp.
A close to perfect compensation is not easy to achieve neither in theory nor in practice.
Class D amps are pretty much cheaper to produce than linear amps with same power and response. Insofar error-forward corrected class D are a nice idea for full active speakers avoiding problems with reactive loads and saving cost of passive crossover parts ( some of these such as special inductors tin foil capacitors etc. cost more than a 200 watts class D with switching power supply!)
If so there is another question, limiters.
This applies to subwoofers and in general bass speakers as the radiation resistance of the moving cone drops with frequency such that cone excursion must be limited.
The limiter problem can be solved using readily available digital processors , but not so easy if one wants to process the PWM signal.
Further we are convinced that the old view , the separation of amp, speaker, and acoustical environment, is in the professional area long dead. Here we have a "system view", that is amp,speaker, and acoustical environment, and particular features of auditive sense are one system.
At the amp side this requires it is a current source whose output impedance
is negative in real part and in complex part the conjugate complex of the speaker impedance in case and listening room.
At the amp side this requires it is a current source whose output impedance
is negative in real part and in complex part the conjugate complex of the speaker impedance in case and listening room.
I am not sure I understand, or perhaps this is a dispute about nomenclature or perspective. How will the return ratio analysis of the 3 different amplifiers be used to establish your thesis? Can you provide an example?
Who has used a Middlebrook probe and erroneously called it a Tian probe?
Best to reply in a new thread and leave this one to Mr Self's use.
Best wishes
David.
Dave Zan,
Hear hear you just stated that very well. This thread is supposed to be about subject matter that we want to see added to Mr. Self's book and not a place to argue about our personal theory or dislikes of his analysis. Save that for your book reviews, and let's get back to the question at hand what do you want to see added to the next edition of the book that was missing or could use more detailed analysis. If Bob Cordell and Doug Self can so easily agree to disagree about this subject and still not interject their personal bias into the discussion we should be able to do the same.
Hear hear you just stated that very well. This thread is supposed to be about subject matter that we want to see added to Mr. Self's book and not a place to argue about our personal theory or dislikes of his analysis. Save that for your book reviews, and let's get back to the question at hand what do you want to see added to the next edition of the book that was missing or could use more detailed analysis. If Bob Cordell and Doug Self can so easily agree to disagree about this subject and still not interject their personal bias into the discussion we should be able to do the same.
TMC is neither just "Output-Inclusive Compensation" (OIC), nor ordinary Miller compensation (OMC). At AF it behaves as OIC, while at HF it behaves like OMC. The transition between these two modes is an essential property of TMC. For this reason it is dubbed transitional Miller compensation.
I'm not sure what you mean by 'learnt', but I did not copy the idea from some else. I invented it by myself. As the idea of TMC is so simple and rather obvious, at least for someone 'skilled in the art' of electronics, I was amazed that I couldn't find any application of it. Precisely for this reason I sent you three emails about this topc, eleven years ago by now. You only responded to the first email. Regrettably, I don't have it any more, but I still remember you told me that Baxandall has invented it already. You also told me that you re-invented the idea, though the results where not, as you called it, 'exciting'. Because your results were a bit disappointing I sent you two more emails to discuss the possibility that the distortion reduction from TMC might be masked by other distortion sources. Since you didn't respond to these emails, I concluded that you weren't interested in TMC anymore... until, about ten years later, and to my amazement, you published an article on this subject. Needless to say that I'm not that happy with this state of affairs.
According to Cherry, a few hundred of his amps were built and they were stable.
Now you know how I came across the idea, and you should know that all the time.
Cheers,
E.
This is pretty much my view. I don't see that "current feedback" amplifiers promise anything for audio use, so time spent investigating them is unlikely to be time well spent. I have decided that the topic will not be in the 6th edition.
It's worth bearing in mind that a very high slew rate (much greater than required for 20 kHz at full throttle) is not a desirable thing. If a fault causes HF oscillation, a modest slew-rate will limit its amplitude and minimise the harm to the output stage. If the slew-rate is deliberately made asymmetrical, then the oscillation will cause a DC offset that will trip the DC protection and disconnect the erring amplifier from the speakers.
Hello Bob
Wouldn't a series RC be more properly called lag-lead compensation, as it becomes resistive at high frequencies? I think this is what it was usually called when it was in the form of a series RC between the input stage collectors. Of course, with current-mirror input loading its now not so easy to do.
How about Transitional Output-Inclusive Miller Compensation? TOIMC?
At what date did you invent it?
I do keep all my emails, but I have to admit that finding one that is eleven years old would be a bit of a mission, so I hope you will permit me to rely on my memory. I recall that the correspondence was about an example of Output-Inclusive compensation that was published by Michael McLoughlin as a Letter to Electronics World in April 1992. I sent in my comments on his plan in August 1992. Firstly it used an under-biased output stage that created big crossover spikes. Secondly it used only small-signal transistors, with Cherry-type non-transitional Output Inclusion, (I gave the 1982 Cherry NDFL reference) and I already knew that it could not be scaled up to use power output devices without the certainty of oscillation. Nonetheless I built an improved version and did manage to demonstrate some reduction in distortion, but the amount was indeed not exciting, and other matters took priority. I am surprised to hear that I did not respond to two emails, and I am sorry you are not happy about that. I try to be meticulous in answering enquiries, as I feel that courtesy is so important in technical debates.
Well at least two of them weren't, because their despairing constructors asked me to fix them. (this was the April 1983 ETI design) All the component values were correct but stability was not to be had; I think using that Butler amplifier stage was a mistake. I entirely accept that these examples may have been outliers but they were built independently by two different people, so I suppose it is at least a data point to note.
But I would like to know when you invented it. I think we should try to record the history of audio as it happens.
Last edited:
And btw what is the highest slew rate that CDs MP3 digital radio provide?
The "fastest" source in this respect is a direct cut venyl LP. Or a high speed analog tape. What really matters is a high open loop bandwidth but not in terms of slew rate but in terms of a smooth spectrum of harmonics with NFB.
Good.
I cannot for the life of me comprehend why the marketing men in the IC industry were allowed to call them "current feedback" amplifiers when it's patently obvious they're nothing of the kind.
Voltage feedback amplifiers they are and will always remain, contrary to Cordell.
Last edited:
It's called current feedback because the current through the feedback resistor directly drives the TIS stage via the common base summing junction stage - some of it is of course shunted to ground by the lower resistor in the feedback leg. There is no gain stage interposed between the feedback and the TIS as in VFA's. Thus, CFA I believe is an apt name.
Discounting CFA's as a valid topology for audio like this seems to me to be unhelpful. If you want to chase lowest distortion at the expense of all else, then maybe. But CFA 's I find are simpler, fast by nature, and yes, subjectively sound very good.
Like all things, there will be tradeoffs, and for VFA's superior PSRR and perhaps lower distortion, the CFA can offer greater circuit simplicity, higher slew rates and wider inherrant circuit bandwidths.
Rather than take a stand either way, I think an open discussion on the pros and cons of both would be more valuable in your book.
Discounting CFA's as a valid topology for audio like this seems to me to be unhelpful. If you want to chase lowest distortion at the expense of all else, then maybe. But CFA 's I find are simpler, fast by nature, and yes, subjectively sound very good.
Like all things, there will be tradeoffs, and for VFA's superior PSRR and perhaps lower distortion, the CFA can offer greater circuit simplicity, higher slew rates and wider inherrant circuit bandwidths.
Rather than take a stand either way, I think an open discussion on the pros and cons of both would be more valuable in your book.
Last edited:
On the contrary diyAudio is full of current feedback themes and still rising. Your book will completely ignore one of the most important topics, greatly improved to date, solid state amplifier designs based on modern semiconductors can offer, so misleading. Definitely not buying something that crippled.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.