I am getting annoyed at Misrepresentation tantamount to Lying
I am suspicious of “naïve subjectivist” claims of "clearly audible" differences in great measuring amplifiers (I r a engineer, I get to define what constitutes "great measurements" to me - it isn't just THD, take a trip through today's AP menus, manuals, literature for a start)
but in this post I’m trying to make clear my objection to supporters of “the other side” bringing simply wrong or discredited “theories” of how (high, global) “feedback” “doesn’t work” to the table – to a degree that amounts to bald faced Lying about the engineering state of the art understanding of the issues
For people serious about intellectual debate they should learn the theory, read the literature – engineers have changed practice, adopted new measures when shown objectively demonstrable errors, including data from perceptual studies
For those ignorant of the engineering principles, feedback, signal theory math – quit posting in “authorative voice” pseudo-technical comments deriding engineering knowledge, feedback theory that you don’t understand, are just aping from the “audiophile” press
the following are objective objections to PMA’s comments – you can trace the ideas through the literature, show in sim or in measured hardware – I have done all three over the years for the issues/comments below
amazingly wrong for a successful amp designer - you have let prejudice blind you to feedback principles and made several statements that are objectively wrong, reveal your ignorance – what is the motivation that causes people to distort engineering reality this way – no need to misrepresent engineering theory to “support” a “subjective sound” hypothesis – being technically wrong detracts from my confidence in you on the validity of the subjective opinion
I think I understand feedback fairly well - I have fixed industrial instrumentation analog output instability into long cable C load by replacing the output op amps with Faster devices - without changing the existing decoupling impedance - so my understanding isn't just "theoretical"
likewise in audio power amps a "faster" circuits - with higher speed output devices that allow loop gain intercept to be pushed up enables Lower decoupling L values while maintaining immunity to C load by keeping the "decoupling frequency" a constant fraction of the gain intercept
I shouldn't have to explain the mechanism to competent engineers qualified to comment on feedback amplifier principles and tradeoffs
For those with less competence you could at least scan titles, abstracts from web search with key words like: “op amp” “stabilize” it will give hits for free, online papers by Feucht, Green, Graeme, Mancini, manufacturer’s app notes – its not difficult to find a number of useful sources - the principles are similar but if you want audio power amp specfic understanding read Cherry, Self, Cordell (the later 2 have up to date books)
slew rate and high global feedback can be considered "orthogonal" in amplifier "design space" - you can vary either independent of the other within large ranges if you choose the right techniques - saying Only "Miller" VAS dominant pole compensation with fixed diff pair tail current can be used just reveals ignorance of more advanced techniques, alternative topologies - that are near 30 years old – even in the “audio” design literature (JAES)
Cordell showed this in circuit analysis, by building a high feedback amplifier with 300 V/us slew rate, by building special hardware for measuring Otala's "TIM" - in Otala's terms - by 1984
CordellAudio.com - A MOSFET Power Amplifier with Error Correction
for the personality cult types John Curl refuses to respond Cordell's technical points, analysis, measurements (I have a copy of the infamous "withheld from publication" "rebuttal" - it is technically vacuous)
there is no reason to say "Higher harmonics above 7th would probably rise" with higher global feedback - at any frequency that the total feedback is higher than in an alternative Cdom compensated amp with similar capability/linearity output stage the distortion will be lower with proper application of higher feedback with appropriate choices of circuit changes for high linearity added gain – such as VAS “Beta enhancement”, bootstrapped Cascodes, other linearization “tricks” shown in Self, Cordell’s books, Cherry, Hawksford's JAES papers
none of the above involves subjective opinion – just genuine knowledge of the technology – expected to trivially accessible to a EE specializing in analog design, control theory
I am suspicious of “naïve subjectivist” claims of "clearly audible" differences in great measuring amplifiers (I r a engineer, I get to define what constitutes "great measurements" to me - it isn't just THD, take a trip through today's AP menus, manuals, literature for a start)
but in this post I’m trying to make clear my objection to supporters of “the other side” bringing simply wrong or discredited “theories” of how (high, global) “feedback” “doesn’t work” to the table – to a degree that amounts to bald faced Lying about the engineering state of the art understanding of the issues
For people serious about intellectual debate they should learn the theory, read the literature – engineers have changed practice, adopted new measures when shown objectively demonstrable errors, including data from perceptual studies
For those ignorant of the engineering principles, feedback, signal theory math – quit posting in “authorative voice” pseudo-technical comments deriding engineering knowledge, feedback theory that you don’t understand, are just aping from the “audiophile” press
the following are objective objections to PMA’s comments – you can trace the ideas through the literature, show in sim or in measured hardware – I have done all three over the years for the issues/comments below
amazingly wrong for a successful amp designer - you have let prejudice blind you to feedback principles and made several statements that are objectively wrong, reveal your ignorance – what is the motivation that causes people to distort engineering reality this way – no need to misrepresent engineering theory to “support” a “subjective sound” hypothesis – being technically wrong detracts from my confidence in you on the validity of the subjective opinion
I think I understand feedback fairly well - I have fixed industrial instrumentation analog output instability into long cable C load by replacing the output op amps with Faster devices - without changing the existing decoupling impedance - so my understanding isn't just "theoretical"
likewise in audio power amps a "faster" circuits - with higher speed output devices that allow loop gain intercept to be pushed up enables Lower decoupling L values while maintaining immunity to C load by keeping the "decoupling frequency" a constant fraction of the gain intercept
I shouldn't have to explain the mechanism to competent engineers qualified to comment on feedback amplifier principles and tradeoffs
For those with less competence you could at least scan titles, abstracts from web search with key words like: “op amp” “stabilize” it will give hits for free, online papers by Feucht, Green, Graeme, Mancini, manufacturer’s app notes – its not difficult to find a number of useful sources - the principles are similar but if you want audio power amp specfic understanding read Cherry, Self, Cordell (the later 2 have up to date books)
slew rate and high global feedback can be considered "orthogonal" in amplifier "design space" - you can vary either independent of the other within large ranges if you choose the right techniques - saying Only "Miller" VAS dominant pole compensation with fixed diff pair tail current can be used just reveals ignorance of more advanced techniques, alternative topologies - that are near 30 years old – even in the “audio” design literature (JAES)
Cordell showed this in circuit analysis, by building a high feedback amplifier with 300 V/us slew rate, by building special hardware for measuring Otala's "TIM" - in Otala's terms - by 1984
CordellAudio.com - A MOSFET Power Amplifier with Error Correction
for the personality cult types John Curl refuses to respond Cordell's technical points, analysis, measurements (I have a copy of the infamous "withheld from publication" "rebuttal" - it is technically vacuous)
there is no reason to say "Higher harmonics above 7th would probably rise" with higher global feedback - at any frequency that the total feedback is higher than in an alternative Cdom compensated amp with similar capability/linearity output stage the distortion will be lower with proper application of higher feedback with appropriate choices of circuit changes for high linearity added gain – such as VAS “Beta enhancement”, bootstrapped Cascodes, other linearization “tricks” shown in Self, Cordell’s books, Cherry, Hawksford's JAES papers
none of the above involves subjective opinion – just genuine knowledge of the technology – expected to trivially accessible to a EE specializing in analog design, control theory
Last edited:
@JCX
In my view, a solid engineer realizes he is a midget sitting on the shoulders of giants. When I need to do a design, I beg, steel and borrow from others that have solved the same problem before me, as long as I can, before coming up with my own solutions. Self, Cordell, others too, they have amassed a wide body of knowledge. Use it, I would say, and only when you find a problem with their solutions, start questioning their theories. So, I am squarely on your side with much of what you say, putting me on the side of the 'objectivists'.
Having said that, there is also an observation I want to make. A number of 'subjectivists' say things I can't quite figure out on the basis of my acid engineering beliefs. At the same time, they are succesful audio design professionals with proven track records. So, they might be on to things that escape the present design paradigms. And they may not themselves be able to formulate exactly what it is, yet. I want to give them leeway for that. I think feedback in conventional amplifiers is the wrong tree to bark against, but it may point into directions we have to take seriously.
In my view, a solid engineer realizes he is a midget sitting on the shoulders of giants. When I need to do a design, I beg, steel and borrow from others that have solved the same problem before me, as long as I can, before coming up with my own solutions. Self, Cordell, others too, they have amassed a wide body of knowledge. Use it, I would say, and only when you find a problem with their solutions, start questioning their theories. So, I am squarely on your side with much of what you say, putting me on the side of the 'objectivists'.
Having said that, there is also an observation I want to make. A number of 'subjectivists' say things I can't quite figure out on the basis of my acid engineering beliefs. At the same time, they are succesful audio design professionals with proven track records. So, they might be on to things that escape the present design paradigms. And they may not themselves be able to formulate exactly what it is, yet. I want to give them leeway for that. I think feedback in conventional amplifiers is the wrong tree to bark against, but it may point into directions we have to take seriously.
I like this your interpretation, and, without shunt reg, one could trace the situation even further than till the rail caps. What goes on, when the rectifying bridge diodes are conducting and, at the same time, signal experiences fast transient? Can one find a hint why power cables and power connectors play a role? My intuitive vision here is, that microlevel distortions produced by electro-magnetic environment and isolation material effects of power cable, and micro-current-spikes at power connectors, produce around -100dB (or less) specific distortions, that are nevetheless quite listenable. This kind of effects, influence the sound stage perception, in addition to NFB specific effects. at this, we should clearly understand, that subjectivists always speak about hardly measurable distortions, special super low noise instruments are needed for objective investigations of such effects. And, thanks to Ed Simon, subjectivists have got at least first minor objective support to their claims.
Last edited:
Well, I have done a lot of sims on shunt regulators, and what I like about is the ease with which they can be stabilized. They can also handle load and line transients much faster and with lower overshoot than series linear topologies. However, do they improve the sound stage or the 'sonics'?
I am afraid if you say yes to this question, your answer can only be based on pure conjecture. There are no published papers or reports or even a decent DBT to prove it.
I am afraid if you say yes to this question, your answer can only be based on pure conjecture. There are no published papers or reports or even a decent DBT to prove it.
Last edited:
I did not check it by explicit measurements, but I am grateful to you for the sims confirmations also. Even small differences in rail voltage overshoots play a role. And in real equipment these differences can be higher, because we speak about very short time periods, when NFB and PSRR do not have their action completed, and with serial PS current passes through longer path and produce more electro-magnetic interference.
From my listening experience, I did believe that the shunt-like stabs better treat transients, and this ensures better reproduction of signal micro-details.
I am almost sure, that shunt PS can be frequently found in high-end equipment. Some of them speak explicitely about using shunt PS (VITUS AUDIO), other prefer to keep it as know-how.
Serial PS, even with zero Zout, transfer transient current consumption along the PS chain, to the electrolytics after diode bridge. On the contrary, shunt PS cut them (electrolytics) out of play, because of high dynamic resistance of the current source, at the input of shunt PS.
Last edited:
Shunt VS series regulator.
Not sure the shunt regulator is that much better than the series regulator.
First, most circuits using regulators, preamp or amp low power stages only have a slight ratio of varying current (AC) to the overall drawn DC current.
Secondly, the rejection of the upstream power supply may be better with the series regulator.
Finally, stability. Let's compare series and shunt positive regulators built around a same perfect voltage reference. In the most simple form, the output of the series regulator would consist of an NPN emitter follower and the output of the shunt regulator of a PNP emitter follower. More elaborated circuits would add NFB around these emitter followers. Would not them be submitted to the same frequency compensation requirements and then behave in the almost same manner towards transients from the load ? From a few experiences I made years ago, I retain that shunt circuits do not behave better than series circuits from this point of view.
.
Not sure the shunt regulator is that much better than the series regulator.
First, most circuits using regulators, preamp or amp low power stages only have a slight ratio of varying current (AC) to the overall drawn DC current.
Secondly, the rejection of the upstream power supply may be better with the series regulator.
Finally, stability. Let's compare series and shunt positive regulators built around a same perfect voltage reference. In the most simple form, the output of the series regulator would consist of an NPN emitter follower and the output of the shunt regulator of a PNP emitter follower. More elaborated circuits would add NFB around these emitter followers. Would not them be submitted to the same frequency compensation requirements and then behave in the almost same manner towards transients from the load ? From a few experiences I made years ago, I retain that shunt circuits do not behave better than series circuits from this point of view.
.
Last edited:
It has not to be necessarily the feedback that affects the soundstage and imaging. dadod has given the hint that wider stereo image can be created with negative crosstalk.
The NJM4565 has a wide image because it has obviously a bad layout design.
But the NJM4565 has also a front in the front and at the same time a distant back !
How could this be?
The NJM4565 has a wide image because it has obviously a bad layout design.
But the NJM4565 has also a front in the front and at the same time a distant back !
How could this be?
For opamp based audio circuitry, I actually avoid tightly regulated (and by implication, low source impedance) supplies. A split winding transformer feeding two separate rectifier and filter banks then feeds a 'cheap' 317 (for both plus and minus supplies) stage. At each opamp, I filter through a 22 ohm and a 100uF cap for both rails. You have to take care with the return ground lines of the filter caps, and how you terminate any heavy loads like the headphone output for example. Most modern op amps have superb PSRR. If you couple this to the line and load regulation of a standard linear reg, and then the filter, you get really outstanding numbers and therefore reduction of line and load induced noise cross coupling between stages and channels- assuming your overall layout is good in this respect. The trace inductance on the ground is always a factor of course, but this applies to any design.
I always have a laugh when I read about people trying to design ultra low noise power supplies for VFA opamp based audio designs.
However, the point here was originally around whether or not the regulators affected sound stage. We have no formal proof.
I always have a laugh when I read about people trying to design ultra low noise power supplies for VFA opamp based audio designs.
However, the point here was originally around whether or not the regulators affected sound stage. We have no formal proof.
Last edited:
I have no idea what the subjective meaning of sound stage is. If sound stage be can improved by either NFB or power supply regulation does this mean that a mono amp can produce a wide sound stage and sound like stereo????
I hardly think that an amplifier is capable of producing any form of sound stage introducing selective enhancements of the signal applied.
I hardly think that an amplifier is capable of producing any form of sound stage introducing selective enhancements of the signal applied.
Last edited:
Hi Bonsai,
I would have to diagree with this. Whether op-amp or discrete circuitry, we must strive for very clean power supplies. For the highest fidelity we must never be complacent about the quality of the DC supplies. Indeed, we must take a two-pronged approach. First, we should always use circuits that are fundamentally resistant to power supply noise - i.e., circuits with good PSRR. Secondly, we should have a very clean supply. Whether "super regs" or something simpler is used is up to the individual designer. I believe that the rails should be regulated cleanly in the first place, then, if appropriate (especially for low-level circuits) use capacitance multipliers.
Cheers,
Bob
Bob,
That's exactly the point of my post. The op amp PSRR is superb at low frequencies (take a look at the 4562 and some of the other newer VFA's). Add a simple RC filter or a cap multiplier and you can easily add another 40dB at mid to high frequencies, just where the op amp PSRR begins to drop off. And, as an added benefit, especially if you use local filtering right at the op amp pins, you get rid of a lot of higher frequency radiated noise coupling from the rail to sensitive circuit nodes elsewhere on the board because you only have LF flowing in the supply lines.
So, in summary, no need for ppm level line and load regulation. 317 type regs with good post filtering are better in my view than a super reg driving op amp supply rails with 0.1uF local decoupling.
For CFA's and discrete circuits that are loop gain challenged, I of course would not go down this route. Ditto for clock circuits where you need to keep jitter down as well.
That's exactly the point of my post. The op amp PSRR is superb at low frequencies (take a look at the 4562 and some of the other newer VFA's). Add a simple RC filter or a cap multiplier and you can easily add another 40dB at mid to high frequencies, just where the op amp PSRR begins to drop off. And, as an added benefit, especially if you use local filtering right at the op amp pins, you get rid of a lot of higher frequency radiated noise coupling from the rail to sensitive circuit nodes elsewhere on the board because you only have LF flowing in the supply lines.
So, in summary, no need for ppm level line and load regulation. 317 type regs with good post filtering are better in my view than a super reg driving op amp supply rails with 0.1uF local decoupling.
For CFA's and discrete circuits that are loop gain challenged, I of course would not go down this route. Ditto for clock circuits where you need to keep jitter down as well.
Last edited:
Good point but every-time I rip the feedback sh*t out of an amp it sounds better (95% of all cases) or just by reducing it.
But what does "sounds better" mean, to you?
Maybe I can save you some time: If it involves "personal preference", or an opinion, then it's almost-certainly meaningless, to everyone else.
My personal preference is for the most-accurate reproduction of the source, even if I don't like or enjoy the sound as much as that of a less-accurate reproduction.
So, when you rip out the feedback, or reduce it, is the sound reproduction then more accurate? Or do you not actually know and just think it sounds better?
Tom
That might be nice. But isn't "Feedback" the primary word in the title of this thread?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.