Dado, you just demonstarted some childish response there yourself and so you are not in a position to complain about that yourself.
It is a fact that CFA topology exists. It also a fact that Esperado is a fan of CFA. Obviously, D. Self does not believe in CFA, and I don't either. My reasons are very simplle - of all the amps which use it, I have not heard one which offeredanything new or extra over what I heard from others in the same price bracket. The scientific thing from D. Self would have been to say simply that he doean't believe in it, show how the principles work and say if you want it, do more research, I will not because it hasn't convinced me that it's worth the time.
Instead, as Esperato reports, D. Self bluntly says no I won't. That to me seems like evading the issue. You do the same, in effect - you learnt a lot from his work, so you like him and support his rather feeble response, thus repeating the mistake. It's only in mathematics that two minuses make a plus, in real life, two minuses make a compound thrid BIG minus.
Not even to mention that several people here suggested you should trust Bob Cordell more rather thab D. Self, a note I also agree with. Bob also has some strong personal views (don't we all?), but has wisely kept them out of his book to the point that if I hadn't read his published papers, I wouldn't be able to make them out from reading his book.
http://www.diyaudio.com/forums/soli...-self-wants-your-opinions-10.html#post3409004
That was the first comment from DS about CFA I can find. I guess by the 9th time he was asked he got a bit more blunt. A well enough reasoned comment.
That was the first comment from DS about CFA I can find. I guess by the 9th time he was asked he got a bit more blunt. A well enough reasoned comment.
You should read more carefully what I wrote.
In my opinion, to technically criticize some one just because of an arrogant answer not reading his book is not very polite. I have the books from Self as from Cordell too. Both books contains some errors, but that is no reason to say they are rubbish, they are excellent book, specially for a hobbyist as I am. I think that Self explanation about non inverting DC servo, one cap(Deboo) and two caps is technically sound, and I don't have long DC servo experience to say myself which one is better.
Regarding CFA - VFA debate I can't say that any of those technology is better, I am trying to design both, why Self don't want to write about CFA he knows better.
Damir
Hi dvv,
I hold Douglas Self in high regard since his is correct about most things, and he is very approachable when it comes to his work. He does share his knowledge about the finer points of audio design without hesitation. He does not treat the subject as a black art that he alone has mastered as some other folks have.
Your comment on his alleged work with Cambridge Audio might be unfair if it turns out that his involvement occurred much earlier in time. In fact we find out that it was a different person who you are unhappy and underwhelmed by. You also have to accept that working for a company involves designing products in their way to retain "their sound". Most engineers are not free to work on products in their own style. Lastly, Mr Self is just a human who isn't always perfect - and he will admit to that. I have had the opportunity to exchange ideas with him and talk about his work. He is a very nice individual who makes time to pass on the knowledge. We can not ask for anything more than that.
Both Douglass Self and Bob Cordell are very knowledgeable. Each is approachable and willing to share what they know. You may disagree with one, but they are both very decent human beings, as are most accomplished designers here. They too understand that they cannot judge their own work, so your comments in that direction are in agreement with just about everyone.
-Chris
I hold Douglas Self in high regard since his is correct about most things, and he is very approachable when it comes to his work. He does share his knowledge about the finer points of audio design without hesitation. He does not treat the subject as a black art that he alone has mastered as some other folks have.
Your comment on his alleged work with Cambridge Audio might be unfair if it turns out that his involvement occurred much earlier in time. In fact we find out that it was a different person who you are unhappy and underwhelmed by. You also have to accept that working for a company involves designing products in their way to retain "their sound". Most engineers are not free to work on products in their own style. Lastly, Mr Self is just a human who isn't always perfect - and he will admit to that. I have had the opportunity to exchange ideas with him and talk about his work. He is a very nice individual who makes time to pass on the knowledge. We can not ask for anything more than that.
Both Douglass Self and Bob Cordell are very knowledgeable. Each is approachable and willing to share what they know. You may disagree with one, but they are both very decent human beings, as are most accomplished designers here. They too understand that they cannot judge their own work, so your comments in that direction are in agreement with just about everyone.
-Chris
Last edited:
Something along the lines of EUVL's "I-V converter", possibly enhanced with an error amp wrapped around the input device, and with a resistor as input, could work, although cumbersome due to the floating supply. But Cordell's two-amp approach is straightforward.
One detail that can be mentioned as well is that the voltage noise contribution of an inverting integrator never is less than unity noise gain. If an integrator is realized with a transconductance or current conveyor, with an output cap termination, the capacitor keeps rolling off at higher frequencies---although you get noise again from the buffer amp. Again this should be a small effect on overall noise, but one for which to account.
I'll have to get hold of Self's account, I'm sure it is not unreasonable. As to cap mismatch, the basic effect of mismatch of RC taus in the differential integrator is to slow setting time in response to a change in the main path's offset voltage, in the same way that some forms of mismatched compensation effect much faster settling times in op amps, and it's probably not very audible in most systems. In fact we work hard to make these d.c. average voltages close to zero, yet I doubt that there are many power amps out there that don't have a d.c. block (or their own servo) somewhere.
Anatech,
I never denied or even implied otherwise. No matter what I or anyone thinks, it is fairly obvious that these gentlemen put much work into their publications, and having done that myself, I appreciate the very effort.
No-one is right all of the time. No matter how hard you work, in the end, mistakes will always creep in somehow, despite many proofreadings. They are different only in their topic nature, are they related to a principle, or are they simple calc mistakes like 2 and 2 equalling something other than 4. Both types can happen.
I do not wish to now engage in a shootout who's better, D. Self or B. Cordell, because that's a matter of taste. I made it clear that I prefer Bob Cordell, and it appears others here feel in a similar way. That fact in no way puts D. Self's work down, and as Dado said, there's much to be learnt from his book.
However, I can't help noting that every time anything other than total agreement is said of D. Self's work, a heated debate is guranteed. Anytime, anywhere, which only proves that far too many people take his work as Gospel.
I never denied or even implied otherwise. No matter what I or anyone thinks, it is fairly obvious that these gentlemen put much work into their publications, and having done that myself, I appreciate the very effort.
No-one is right all of the time. No matter how hard you work, in the end, mistakes will always creep in somehow, despite many proofreadings. They are different only in their topic nature, are they related to a principle, or are they simple calc mistakes like 2 and 2 equalling something other than 4. Both types can happen.
I do not wish to now engage in a shootout who's better, D. Self or B. Cordell, because that's a matter of taste. I made it clear that I prefer Bob Cordell, and it appears others here feel in a similar way. That fact in no way puts D. Self's work down, and as Dado said, there's much to be learnt from his book.
However, I can't help noting that every time anything other than total agreement is said of D. Self's work, a heated debate is guranteed. Anytime, anywhere, which only proves that far too many people take his work as Gospel.
Hi dvv,
Yes, I was attempting to settle things down before an eruption occurred. Both men are intelligent and forthcoming. I agree with both on varying topics. The idea of a shootout would not please either author. I have no real preference between either man, and I own both author's books. Both are well worth reading. I am also building projects from both of these sharp minds.
-Chris
Yes, I was attempting to settle things down before an eruption occurred. Both men are intelligent and forthcoming. I agree with both on varying topics. The idea of a shootout would not please either author. I have no real preference between either man, and I own both author's books. Both are well worth reading. I am also building projects from both of these sharp minds.
-Chris
I recall that discussion distantly, and I did some experiments on samples, but I don't think I got as far as incorporating such a driven substrate into a design. I do recall that the isolation of the gates of the 389 was more complicated than a simple reverse-biased diode, and also that the breakdown voltages gate-to-gate or gate-to-substrate were surprisingly high. I'm not sure how the Linear Integrated Systems equivalent parts compare.
Without any additional information, I suppose you could try biasing the substrate with a voltage that tracks the follower output, perhaps with a battery at least for the experiment. With a resistor in series you can preclude damage from excessive voltages.
I was exploring the possibility of using leakage to that substrate pin as a probe of temperature, but it was not promising iirc.
Brad
capacitance versus voltage
Another quick note about the gate-substrate capacitance effects: it is a natural tendency to qualify them as "nonlinear", and indeed they are, with Caldwell showing some equations. But note that Variation of C with V is sufficient to generate distortion, even if that dependency were linear. So what is wrong with saying "nonlinear"? Merely that it is misleading for the student attempting to understand distortion mechanisms.
Another quick note about the gate-substrate capacitance effects: it is a natural tendency to qualify them as "nonlinear", and indeed they are, with Caldwell showing some equations. But note that Variation of C with V is sufficient to generate distortion, even if that dependency were linear. So what is wrong with saying "nonlinear"? Merely that it is misleading for the student attempting to understand distortion mechanisms.
Last edited:
That is quite the statement. Someone should send the man a good 200W
CFA - let him listen for a day !
Then he goes on to say that a VFA's oscillation is "tamer" than a CFA's.
Oscillated both - quite the opposite.
Asymmetrical slew on a CFA ? - also , quite the opposite.
then lazy cat get's into it - oh boy ...D. Self says a CFA is not "blameless" grade ... both my CFA's
are @10ppm @ >90db psrr (below 2).
I like both type amps , personally - but a CFA can easily match
(most) VFA attributes. But , instruments are better "placed"
with a CFA.
Performance wise - I get the same with either VFA/CFA (below 1/2). I
can't imagine why a CFA would get a "not valid for audio" (assertion).
PS -
Cambridge audio looks like a bunch of overpriced junk !
Cheap PS's and "wimpy" output stages .....
I WANT my Parasound's ...
OS
Attachments
That's funny, as i asked Doug Self why he did not use a jfet as the first stage in his elektor 2012 pre-amp. I showed him the amber 3501 front end ckt as an example of such a hybrid arrangement. the elektor 2012 pre uses the old ne5534a. he said a j310? would offer a 1.7 dB theoretical lower noise but he brought up the point about the jfet c/v non-linearity issue causing imd =
I also asked why new opamps have higher current noise than a old ne5534? he said it was a trade off, semi mfg's want lower voltage noise over current noise, so i infer you can not have it both ways.
Last edited:
For a bipolar with a given beta, and for a given impedance in the source and/or feedback divider, you will have an optimal bias for the combination of voltage and current noise (aka series and parallel noise). Barrie Gilbert mentioned some experimental devices that had beta around 10^5, but I guess they were not that straightforward to make. Good low rbb' parts are usually more like a beta of 350 or so.That's funny, as i asked Doug Self why he did not use a jfet as the first stage in his elektor 2012 pre-amp. I showed him the amber 3501 front end ckt as an example of such a hybrid arrangement. the elektor 2012 pre uses the old ne5534a. he said a j310? would offer a 1.7 dB theoretical lower noise but he brought up the point about the jfet c/v non-linearity issue causing imd =
I also asked why new opamps have higher current noise than a old ne5534? he said it was a trade off, semi mfg's want lower voltage noise over current noise, so i infer you can not have it both ways.
The effects of the voltage-dependence of discrete JFET capacitances can be drastically reduced with series feedback to the sources and cascoding of the drains. If the cascode device has sufficient transconductance and if the voltage swing at its drain is not large, the gain from the lower JFET's gate to its own drain will approach unity, hence the Miller effect multiplier will approach unity. If you use control-electrode-charge recycling a la Aldridge with such a cascode part, and tie the upper JFET or DMOS device's gate (suitably biased) to the input device source, the Miller multiplier will get less than unity. The drawback is a tendency to a net negative input impedance, which will require a little lumped C to avoid oscillations.
Such structures can be cascaded further, with adequate power supplies, or versions of folded cascodes pursued.
The nice thing about JFETs is that they can be paralleled readily for a square-root-of-N reduction in series noise. Provided the source impedance is sufficiently low, the additional capacitance is handled readily. Note a different story for very-high-impedance sources like photodiodes and mic capsules: gate leakage becomes important and as well the best broadband signal-to-noise at audio frequencies is usually where the JFET C is about the same as the source C. Thus it makes no sense to make the JFET a lot bigger, as discussed by Wurcer in his Linear Audio articles on condenser microphone preamps.
Last edited:
I mentioned in other forums that most all the high-end products of a line of products by various manufacturers went from VFA to CFA topology. Why was that? japanese mfr and American and others. Their top of the line was a CFA topology. So that lead to a forum on What is a CFA and what are the differences between them (VFA). Its all very good reading now as much wasnt known before nor clear (see Solid State).
I think DS has the right to say his opinions in his books he sells but in the end, he is shooting himself in the foot marketing/sales-wise. CFA is here in many forms well beyond audio and is a bad move to not include it. So, we here at DIYAudio did our own investigations and one by one the limitations of each (VFB and CFB) ways were found to get them more and more alike in performance. Limitations at first were later over come.
But, in the end, there are engineering differences and trade-offs. if nothing else, some things are easier to attain with one or the other in practice. High freq stability and distortion are a couple areas of such differences for audio.
There are many good sources recently (10 years) to learn about CFB Amps and the forum gave several.... another one is from the publishing company Springer -- Current Feedback Operational Amplifiers and Their Applications. [Analog Circuits and signal processing]
THx-RNMarsh
I think DS has the right to say his opinions in his books he sells but in the end, he is shooting himself in the foot marketing/sales-wise. CFA is here in many forms well beyond audio and is a bad move to not include it. So, we here at DIYAudio did our own investigations and one by one the limitations of each (VFB and CFB) ways were found to get them more and more alike in performance. Limitations at first were later over come.
But, in the end, there are engineering differences and trade-offs. if nothing else, some things are easier to attain with one or the other in practice. High freq stability and distortion are a couple areas of such differences for audio.
There are many good sources recently (10 years) to learn about CFB Amps and the forum gave several.... another one is from the publishing company Springer -- Current Feedback Operational Amplifiers and Their Applications. [Analog Circuits and signal processing]
THx-RNMarsh
Last edited:
To make things clear, i do not want to oppose one author against an other.
Apart laws of physics, that had been verified so many times than it would be a waste of time to verify ourselves, everything need to be verified in order to get a based opinion.
When i was young, half a century ago, there were no name for transistors arrangements, like Diamonds, super-pairs etc... We used them as and as we had the idea of one of them. It was not surprising we used them all ;-)
I don't think it is a good idea to refer to authority, well known topologies, simulators etc...
We all have to learn to imagine, calculate, measure how behave our voltages and currents in each and every point of a complex assembly, like a finished amp. learning the evils of the components, parasitic capacitance, non linearity, thermal behavior, parasitic inductance everywhere etc...
That is the first step. Objectivism.
The second one is to become to feel instead of referring only to numbers. That comes with experience. And, on this point, because we are never exactly interested in the same things, we will search in a personal direction.
By example, i was very interested by the searches of OS on thermal behaviors and their compensations, assembling stages with an opposite temp behavior. Never had this idea.
Apart laws of physics, that had been verified so many times than it would be a waste of time to verify ourselves, everything need to be verified in order to get a based opinion.
When i was young, half a century ago, there were no name for transistors arrangements, like Diamonds, super-pairs etc... We used them as and as we had the idea of one of them. It was not surprising we used them all ;-)
I don't think it is a good idea to refer to authority, well known topologies, simulators etc...
We all have to learn to imagine, calculate, measure how behave our voltages and currents in each and every point of a complex assembly, like a finished amp. learning the evils of the components, parasitic capacitance, non linearity, thermal behavior, parasitic inductance everywhere etc...
That is the first step. Objectivism.
The second one is to become to feel instead of referring only to numbers. That comes with experience. And, on this point, because we are never exactly interested in the same things, we will search in a personal direction.
By example, i was very interested by the searches of OS on thermal behaviors and their compensations, assembling stages with an opposite temp behavior. Never had this idea.
- Status
- Not open for further replies.