A question: you are designing a power amp and trying to optimize compensation. What would you rather have: 40 deg of phase margin with 11 dB of gain margin, or 45 deg of phase margin and 7 dB of gain margin?
Jan
Jan
Simulate with capacitive charges will gives-you the answer ;-)
On my side, i use this method: Optimize Miller cap for max flat bandwidth possible with no bump at HF. Then, insert a low pass input filter to kill any overshoot on squave waves (both little and large signals).
Last edited:
The discussion of the circuitry is sound enough, but it appears the inventor merely asserts electromechanical damping without any evidence for it.
I see he failed to pay maintenance so it lapsed pretty early on.
too, "A phase margin of 60 degrees is also a magic number because it allows for the fastest settling time when attempting to follow a voltage step input ".
Jan, as a first step, I'd shoot for at least 60 degrees phase margin on an audio power amplifier designed to drive commercial speaker loads. Different story of course if you have a a DAC output buffer where you might tolerate lower phase margins and instead focus on settling time.
Personally, I shoot for at least 60 degrees and preferably better than 10 dB gain margin. Anything better than this and you are home and dry. And, in all cases, an output inductor and a Zobel network can help when driving capacitive loads.
Personally, I shoot for at least 60 degrees and preferably better than 10 dB gain margin. Anything better than this and you are home and dry. And, in all cases, an output inductor and a Zobel network can help when driving capacitive loads.
I am now at 45 deg and 11 dB. Pure cap load is still a problem, but not with say 1 ohms in series, which I think is more realistic.
Jan
The Mohr patent does describe something I have remarked on before, which is using a low-Z input structure which feedback enhances, rather than starting with a high-Z input as in a traditional VFB op amp and reducing the impedance with feedback. Some good discussions of the general approaches can be found in a book by an Israeli nuclear scientist Arbel, Analog Signal Processing and Instrumentation. For example for an analog differentiator, he prefaces an op amp with a common-base stage.
There are a lot of other good things in the book, despite not being directly audio-related, and last I looked it was fairly cheap used (unlike some other recommendations of mine, like Cherry and Hooper!).
Settling Time
As this has been mentioned in connection with phase margin, it's worth perusing Bob Demrow's classic article from Analog Dialogue:
http://www.analog.com/media/en/tech...on-notes/466359863287538299597392756AN359.pdf
As this has been mentioned in connection with phase margin, it's worth perusing Bob Demrow's classic article from Analog Dialogue:
http://www.analog.com/media/en/tech...on-notes/466359863287538299597392756AN359.pdf
I think you will like it. It's somewhat dated but still has a lot of stuff I'd not seen elsewhere.
One of the poignant and tragic aspects of that book is the influence of the culture of nuclear science, and the plight of that country then (and now). One circuit example that Arbel cites, contributed by one of his students, is described as The Mosquito on the Tank 🙁
the known Psychoacoustics don't point to any particular human audio/music listening need for fast in electronic terms settling
temporal masking is very strong, includes substantial pre masking as well as even more post masking - with envelope time constant near 1 millisecond
judging feedback amplifier stability by eyeball/scope step settling/ringing can be "too simple" for perfectly useful, recommended higher order compensation
even worse when using TMC - can look good in step response when tuned to disastrously low phase margin
temporal masking is very strong, includes substantial pre masking as well as even more post masking - with envelope time constant near 1 millisecond
judging feedback amplifier stability by eyeball/scope step settling/ringing can be "too simple" for perfectly useful, recommended higher order compensation
even worse when using TMC - can look good in step response when tuned to disastrously low phase margin
Last edited:
I agree, although some people I respect (like Keith Johnson) seemed to value it, at least for a while. I don't know what KOJ's thinking is on the subject today, but having mentioned it to him many years ago in connection with a charge preamp I designed (where it was very much needed as part of fast data acquisition) when he visited my office at UCLA, and his dismissing it as anything important for audio, some years later he cited the settling time behavior of a Spectral power amp 😀 In fact he mentioned the choice of an input FET as assisting in the achievement of fast settling, despite it being suboptimal for noise.
So I think it is nice to have, not necessarily in power amps but if it can be achieved without sacrifice of other criteria, one may as well do it. Of course the higher order compensation schemes have benefits as well.
It has been put in the public domain and can be downloaded free of charge.
https://archive.org/details/CherryHooperAmplifyingDevicesAndLowPassAmplifierDesign1968RR
That's nice. I suspect the ~700 dollar copies on the web will be of interest to but a few 🙂 Although I vastly prefer having a physical book.
An anecdote: Floyd Toole had intended to become a circuit designer and study for his PhD with Ed Cherry. But when he got to the institution he discovered that the Cherry was Colin Cherry, a very different person with a very different focus. FT decided to stay on anyway, and although I don't doubt that he would have become a fine designer, I'm happy for the direction he pursued.
- Status
- Not open for further replies.