Re: Re: Re: Re: Re: Neutralization
Wow! This is the kind of stuff I love!
Take a problem that *everyone* runs into and come up with a completely unique solution. (I'm assuming that you've actually tried this and that it works.)
In my particular case, yes it would. However, the feedback around the unity-gain second order Sallen&Key filter only serves to sharpen up the corner at the rolloff point. So if the idea works, you could leave out the feedback and have a unity-gain second order lowpass filter with no feedback.
It's a very interesting idea, and just thinking about it gives a springboard for a half-dozen other ideas that might also work.
Like I said, this is the kind of stuff I love. Instead of just doing the same old junk that everyone has been doing for the last 40 or 50 years, come up with something new and exciting and different. It's the only way that improvements can be made. If you just use the same old circuits that everybody else uses, it's just going to sound the same as the same old circuits everybody else uses. And what's the point of that?
Terry Demol said:
Wow! This is the kind of stuff I love!
Take a problem that *everyone* runs into and come up with a completely unique solution. (I'm assuming that you've actually tried this and that it works.)
Terry Demol said:
In my particular case, yes it would. However, the feedback around the unity-gain second order Sallen&Key filter only serves to sharpen up the corner at the rolloff point. So if the idea works, you could leave out the feedback and have a unity-gain second order lowpass filter with no feedback.
It's a very interesting idea, and just thinking about it gives a springboard for a half-dozen other ideas that might also work.
Like I said, this is the kind of stuff I love. Instead of just doing the same old junk that everyone has been doing for the last 40 or 50 years, come up with something new and exciting and different. It's the only way that improvements can be made. If you just use the same old circuits that everybody else uses, it's just going to sound the same as the same old circuits everybody else uses. And what's the point of that?
Justcallmedad said:
Sorry, I think there is a problem with the language barrier.
I understood your previous posts to say that you had chosen to use a cascoded current mirror instead of the much more commonly used Wilson current mirror because the Wilson mirror had worse high-order distortion.
Another poster correctly pointed out that the Wilson current mirror employs a feedback loop as its basis of operation.
My comment was only intended to highlight the fact that feedback (among other things) tends to increase the higher order harmonics, which is why I prefer to not use feedback and also prefer not to use the Wilson current mirror.
I think the schematics you posted are very interesting and will sound very good. I would guess that the first one would sound better (as well as offering less DC drift), as the current mirror has a much better PSRR than the folded cascode.
Your worry about "extra" junctions in the signal path is unfounded, in my experience. All of that myth started with an article that Bongiorno wrote in Audio Amateur about 30 years ago. He examined the behavior of various arrangement of output stage triples and made some completely erroneous conclusions based on a total misunderstanding of what he was measuring and how the circuits actually operated.
As I have noted many time before, in my experience the best output stage is the "T-Circuit" (basically a triple emitter follower) developed by Bart Locanthi at JBL in 1966. Leach used it in his amplifiers (and has a copy of one of Locanthi's original articles on his website). I have used that output stage in all of my power amplifier designs for the last eight years. The only improvement I have been able to make is to stabilize it so that no output inductor or Zobel network is required for stability with a capacitive load.
Edit: Just to add fuel to the fire, the way I determined that the emitter follower triple was the best output stage was via simulations. I tried error-correction circuits (such as the one that Levinson stole from Sansui), double emitter followers, and several other variations before settling on the emitter follower triple. It has many advantages over other output stage topologies. However, I have to admit that I would be tempted to use a double with a MOSFET driver if I could find a MOSFET that was worth a dang. But the audio market is too small for the semiconductor manufacturers to bother with.
Charles Hansen said:
That must be the article where he urged us to use massive film caps as power supply caps because he 'proved' that they were in the signal path. He was really out of luck then because just after AA published the amp article, they found out that his output devices were discontinued... I had some sympathy for him at that point...
Jan Didden
janneman said:
That was the article, but you are mis-remembering what he actually said.
What he really said was that capacitor quality makes no difference whatsoever. His argument was that the output signal in a (non-bridged) power amp went through the electrolytic caps in the power supply. Therefore, even if caps *did* make a difference (which he didn't admit), it wouldn't matter because the electrolytic caps in the PS would still create a performance limitation.
A very outspoken guy, with very outspoken opinions. It was that article that caused me to lose a lot of respect for him. On the other hand, you can learn something from almost everybody. I learned a couple of things from that article, despite the errors and false claims he made in it.
I agree, we need better mosfets to design with.
For the record, the Blowtorch used Fairchild mosfets (thanks for the advice on this) and my new output stage is for the Vendetta phono preamp is using the Hitachi K214, j76.
I sure wish both Linear Systems and Supertex would get their act together.
For the record, the Blowtorch used Fairchild mosfets (thanks for the advice on this) and my new output stage is for the Vendetta phono preamp is using the Hitachi K214, j76.
I sure wish both Linear Systems and Supertex would get their act together.
It is truly a shame that there remains friction between John and me. I suspect it is largely the fault of each of our personalities, which tend to clash as a result of different ways of dealing with technical controversies. John doesn’t like to be questioned and I tend to back people into a corner with tough, sometimes blunt, questions. A recipe for volatility, to say the least.
For those interested, here is what happened about 30 years ago.
In February of 1980 I published “Another View of TIM” in Audio magazine. The original working title was “Understanding TIM”, but the Editor changed the title and that may have given the article a more confrontational appearance that was never intended. The original article is on my web site at www.cordellaudio.com. I think most who read it will agree that what was said was reasonable and not overly attacking. The paper did, however, assert that Otala was wrong in saying that high NFB and/or low open-loop bandwidth led to the production of TIM. I was not the first to point out that TIM was largely a symptom arising from slew rate limiting or its onset. Otala, John, Walt Jung and Marshall Leach were upset with my article (needlessly so, in my view) and wrote a long letter of protest to the Editor that never got published. I have not seen it since. Subsequently a much shorter letter did get published. I think this was the beginning of John’s anger toward me.
In September of 1981 I published a paper in the JAES entitled “A Fully In-Band Test for Transient Intermodulation Distortion”. This paper disclosed a three-tone test that folded both even and odd distortion products down in frequency so that they could be measured by inexpensive instrumentation. Otala’s test was fine, but it required an expensive spectrum analyzer. The paper also showed comparative tests for TIM using Otala’s DIM test and THD-20. I think the paper again asserted that Otala was wrong in blaming large values of NFB and/or small open-loop bandwidth for the production of TIM. This rubbed salt in the wound.
A common theme emerged; Otala would come up with new bad things to blame on large amounts of negative feedback and I would investigate the claims and show them to be less than fully correct. What a persistent PIA I must have been to Matti. John probably felt that I was too ruthless in calling Otala to task; maybe I was.
In 1982 I presented a paper at the 64th AES Convention titled “Open-Loop Output Impedance and Interface Intermodulation Distortion in Audio Power Amplifiers”. Otala had come up with another reason why we should not use large amounts of negative feedback, IIM. Once again, he rightly put the spotlight on an important issue (largely the one of importance of high output current capability) and also, to his credit, proposed a way to measure this new distortion. He was wrong in blaming it on large amounts of NFB, however. I built real amplifiers and made real IIM measurements on amplifiers with low and high open loop output impedance and proved him wrong. More salt in the would.
In March of 1983 I published a paper in the JAES titled “Phase Intermodulation Distortion – Instrumentation and Measurements”. Matti had again come up with yet another reason why NFB is bad, and, to his credit, proposed a way to measure it. Being the PIA that I am, I set about to build an instrument to measure it and once again compared amplifiers with different amounts of feedback, and once again showed that Otala’s generalities were wrong. More salt in the wound. Otala was a very smart guy, but I think he was obsessed with his intuitive perception of the badness of negative feedback. I was a persistent and curious guy, and was a real PIA in challenging his assertions at every turn.
I then published “A MOSFET Power Amplifier with Error Correction” in the JAES in January of 1984. It sported A LOT of negative feedback in addition to the error correction. It had 40 dB of NFB at 20 kHz and probably had an open loop corner down around 100 Hz. I measured DIM, PIM, IIM and other distortions. They were vanishingly small. More evidence that, when properly used, large amounts of NFB and small open loop bandwidth did not in any way lead to these readily measurable distortions. This again probably did not endear me to those critical of NFB.
All of these papers are on my web site. I think they are still a good read. The fact that I put them up on my web site about 18 months ago probably further angered John.
I have to emphasize that I was not the only one to take Otala to task, but I think I became the lightning rod for John’s anger.
The negative feedback controversy has always fascinated me. For all I know, there may be something to the claims that NFB is not as good as we would like it to be. I don’t think that is the case, but if it were the case, it would come under the category of things we just do not yet understand. I do know that the signals inside the loop take a different form. But I have generally stuck to showing that the objective theories for the supposed badness of NFB simply don’t hold water. If there is one out there that does hold water, I’d love to prove it correct and embrace it.
Cheers,
Bob
For those interested, here is what happened about 30 years ago.
In February of 1980 I published “Another View of TIM” in Audio magazine. The original working title was “Understanding TIM”, but the Editor changed the title and that may have given the article a more confrontational appearance that was never intended. The original article is on my web site at www.cordellaudio.com. I think most who read it will agree that what was said was reasonable and not overly attacking. The paper did, however, assert that Otala was wrong in saying that high NFB and/or low open-loop bandwidth led to the production of TIM. I was not the first to point out that TIM was largely a symptom arising from slew rate limiting or its onset. Otala, John, Walt Jung and Marshall Leach were upset with my article (needlessly so, in my view) and wrote a long letter of protest to the Editor that never got published. I have not seen it since. Subsequently a much shorter letter did get published. I think this was the beginning of John’s anger toward me.
In September of 1981 I published a paper in the JAES entitled “A Fully In-Band Test for Transient Intermodulation Distortion”. This paper disclosed a three-tone test that folded both even and odd distortion products down in frequency so that they could be measured by inexpensive instrumentation. Otala’s test was fine, but it required an expensive spectrum analyzer. The paper also showed comparative tests for TIM using Otala’s DIM test and THD-20. I think the paper again asserted that Otala was wrong in blaming large values of NFB and/or small open-loop bandwidth for the production of TIM. This rubbed salt in the wound.
A common theme emerged; Otala would come up with new bad things to blame on large amounts of negative feedback and I would investigate the claims and show them to be less than fully correct. What a persistent PIA I must have been to Matti. John probably felt that I was too ruthless in calling Otala to task; maybe I was.
In 1982 I presented a paper at the 64th AES Convention titled “Open-Loop Output Impedance and Interface Intermodulation Distortion in Audio Power Amplifiers”. Otala had come up with another reason why we should not use large amounts of negative feedback, IIM. Once again, he rightly put the spotlight on an important issue (largely the one of importance of high output current capability) and also, to his credit, proposed a way to measure this new distortion. He was wrong in blaming it on large amounts of NFB, however. I built real amplifiers and made real IIM measurements on amplifiers with low and high open loop output impedance and proved him wrong. More salt in the would.
In March of 1983 I published a paper in the JAES titled “Phase Intermodulation Distortion – Instrumentation and Measurements”. Matti had again come up with yet another reason why NFB is bad, and, to his credit, proposed a way to measure it. Being the PIA that I am, I set about to build an instrument to measure it and once again compared amplifiers with different amounts of feedback, and once again showed that Otala’s generalities were wrong. More salt in the wound. Otala was a very smart guy, but I think he was obsessed with his intuitive perception of the badness of negative feedback. I was a persistent and curious guy, and was a real PIA in challenging his assertions at every turn.
I then published “A MOSFET Power Amplifier with Error Correction” in the JAES in January of 1984. It sported A LOT of negative feedback in addition to the error correction. It had 40 dB of NFB at 20 kHz and probably had an open loop corner down around 100 Hz. I measured DIM, PIM, IIM and other distortions. They were vanishingly small. More evidence that, when properly used, large amounts of NFB and small open loop bandwidth did not in any way lead to these readily measurable distortions. This again probably did not endear me to those critical of NFB.
All of these papers are on my web site. I think they are still a good read. The fact that I put them up on my web site about 18 months ago probably further angered John.
I have to emphasize that I was not the only one to take Otala to task, but I think I became the lightning rod for John’s anger.
The negative feedback controversy has always fascinated me. For all I know, there may be something to the claims that NFB is not as good as we would like it to be. I don’t think that is the case, but if it were the case, it would come under the category of things we just do not yet understand. I do know that the signals inside the loop take a different form. But I have generally stuck to showing that the objective theories for the supposed badness of NFB simply don’t hold water. If there is one out there that does hold water, I’d love to prove it correct and embrace it.
Cheers,
Bob
Bob, you have made a very good overview of the situation, except the the 14 page rebuttal to you that Matti, Walt Jung, Marshal Leach, and I made to your articles in 'Audio' and their tone which impugned our efforts. Remember when you and I saw each other at the 1980 ISSCC conference? You were surprised at my presence there (after all why would someone like me bother?) and you politely asked what I was up to, and I said 'Rebutting you!' in front of your boss, no less, or at least that's what I remember.
For the record, I also got into trouble with Matti because of editing by 'Audio'. The acknowledgment of Matti Otala and the Gvt Lab in Finland was deleted, in my article: 'Omitted Factors in Audio Design' published in 1979. Matti thought I had short-sided him as well. Gene Pitts got is both in trouble, how about that!
In any case, our rebuttal to you was adequate at the time, but it was never published to the public. Please remember it is not just Matti and me who felt this way about your articles, and Matti and I were hardly on speaking terms at the time.
For the record, I also got into trouble with Matti because of editing by 'Audio'. The acknowledgment of Matti Otala and the Gvt Lab in Finland was deleted, in my article: 'Omitted Factors in Audio Design' published in 1979. Matti thought I had short-sided him as well. Gene Pitts got is both in trouble, how about that!
In any case, our rebuttal to you was adequate at the time, but it was never published to the public. Please remember it is not just Matti and me who felt this way about your articles, and Matti and I were hardly on speaking terms at the time.
Following quietly off to the side and looking at some of the posted harmonic distortion simulations I was wondering (honestly don't know the answer) if there is any benefit to simulation of two or more tone "IM" distortion plots when seeking to know if one stage topology is better than the other??
Put another way, will the results of the simulation or real world tests be the same for either test method?
_-_-bear
Put another way, will the results of the simulation or real world tests be the same for either test method?
_-_-bear
- Status
- Not open for further replies.