what, no comment on the Putzeys paper? http://www.linearaudio.net/userfiles/file/letters/Volume_1_BP.pdf
curious minds want to know...
Barrie is quite interested, you might not be happy. When folks just say out of the blue that feedback has a "sound" I just wince. They reciprocate by having no interest in any quantifying of that stand at all. I have no problem at all in who they are, I don't agree.
For me, it is not the 'objective' evaluation of negative feedback that I most value, as much as the 'subjective' evaluation of negative feedback that I have acquired. It is my EXPERIENCE with negative feedback, coupled with advanced classes in nonlinear analog design that I took at UC Berkeley, almost 40 years ago, that convinced me that feedback 'could' be more problematic than most engineers realized. Some have come to learn this, along with me, and others stick with the op amp model as being ideal. My job in life is to make the best amps, preamps, possible, not to make the best 'measuring' amps and preamps. My designs still measure pretty well, perhaps better than they need to to sound good.
As I said The attack on feedback is a misplaced battlefield, choose your sides. If you set up every test of your hypothosys as a no win situation why bother?
My concern over feedback, today, is mostly OPEN LOOP frequency response. The higher the open loop frequency response, the better. After TIM is taken care of, and that took awhile, then open loop bandwidth was also associated with improved TIM, which started with the so called Otala amplifier, back in 1973.
Hi,
NOT SO FAST.
Negative feedback as such is a simple mechanism that produces certain results.
And it CAN be illustrated that while negative feedback can improve overall THD, it can lead to increases in higher order (and thus subjectively more ojectionable) harmonics.
The problem is that some brainlessly promote NFB as the cure all panacea.
To the point that they remove well considered degeneration from circuits (aka localised negative feedback linearising individual stages) in favour of more loop gain and lower measured THD at 1KHz and full unclipped power.
It is understandable that some oppose this kind of thing as stupid and non-conductive to high sound quality.
Another funny thing.
If we actually have done our homework we can make an open loop amp with ~ 0.1% THD near clipping dominated by the output stage (CFB Pair in heavy biased Class AB) with near monotonic decay of HD and with the contribution from the VAS/Input Stages much lower than this.
Such an amplifier can easily have several 100KHz bandwidth, it does of course use significant levels of localised feedback in each stage.
An interesting question would be if the addition a global negative feedback loop and the reduction of the localised feedback to give loop gain would make such an amplifier "better" in a measured sense, especially if consider distortion audibility (say using GedLee metric, or D.E.L. Shorter's distortion weighting) and to then compare this amplifier by listening to the two states (which could be made switchable) under blind conditions with a listening panel.
The point of NFB is not if it's good or bad, but if the application is good or bad. In industrial control applications I have had very graphic (and very scary) illustrations what happens when control loops suddenly open up (slew rate exceeded) or start to ring (error signal and loop design cause damped or even un-damped oscillation). Un-damped oscillation in a system that attempts to pour a steady stream of liquid steel from a 140 Ton "pot" of liquid steel but fails and breaks into oscillation is a sight to behold - thanks the creator of that particular system for including manual overrides and the operator for engaging them quickly).
Another interesting thing to do is to listen to error signal in a global loop feedback amp. In many such amplifiers interestingly the major distortion source is not the output stage, but stages earlier in the chain.
In the end we come to the conclusion that we want to maximise open loop linearity and bandwidth open loop as much as possible, so that the only non-linearities that our feedback loop addresses are really unavoidable (e.g. Class AB Output stage crossover and other distortion), however lack of linearity in other stages is better dealt with locally.
And once we are sufficiently far down that path, we readily come to a point where global feedback loops become optional tools, that may safely omitted if we accept that some parameters are degraded, in trade we may find other parameters, possibly ones not characterised fully buy the limited set of traditional amplifier measurements in common use, improved.
From this end I agree for example with Robert Cordell that making amplifiers that use no looped feedback and perform acceptably teaches us much that is useful also in looped feedback applications.
Ciao T
NOT SO FAST.
Negative feedback as such is a simple mechanism that produces certain results.
And it CAN be illustrated that while negative feedback can improve overall THD, it can lead to increases in higher order (and thus subjectively more ojectionable) harmonics.
The problem is that some brainlessly promote NFB as the cure all panacea.
To the point that they remove well considered degeneration from circuits (aka localised negative feedback linearising individual stages) in favour of more loop gain and lower measured THD at 1KHz and full unclipped power.
It is understandable that some oppose this kind of thing as stupid and non-conductive to high sound quality.
Another funny thing.
If we actually have done our homework we can make an open loop amp with ~ 0.1% THD near clipping dominated by the output stage (CFB Pair in heavy biased Class AB) with near monotonic decay of HD and with the contribution from the VAS/Input Stages much lower than this.
Such an amplifier can easily have several 100KHz bandwidth, it does of course use significant levels of localised feedback in each stage.
An interesting question would be if the addition a global negative feedback loop and the reduction of the localised feedback to give loop gain would make such an amplifier "better" in a measured sense, especially if consider distortion audibility (say using GedLee metric, or D.E.L. Shorter's distortion weighting) and to then compare this amplifier by listening to the two states (which could be made switchable) under blind conditions with a listening panel.
The point of NFB is not if it's good or bad, but if the application is good or bad. In industrial control applications I have had very graphic (and very scary) illustrations what happens when control loops suddenly open up (slew rate exceeded) or start to ring (error signal and loop design cause damped or even un-damped oscillation). Un-damped oscillation in a system that attempts to pour a steady stream of liquid steel from a 140 Ton "pot" of liquid steel but fails and breaks into oscillation is a sight to behold - thanks the creator of that particular system for including manual overrides and the operator for engaging them quickly).
Another interesting thing to do is to listen to error signal in a global loop feedback amp. In many such amplifiers interestingly the major distortion source is not the output stage, but stages earlier in the chain.
In the end we come to the conclusion that we want to maximise open loop linearity and bandwidth open loop as much as possible, so that the only non-linearities that our feedback loop addresses are really unavoidable (e.g. Class AB Output stage crossover and other distortion), however lack of linearity in other stages is better dealt with locally.
And once we are sufficiently far down that path, we readily come to a point where global feedback loops become optional tools, that may safely omitted if we accept that some parameters are degraded, in trade we may find other parameters, possibly ones not characterised fully buy the limited set of traditional amplifier measurements in common use, improved.
From this end I agree for example with Robert Cordell that making amplifiers that use no looped feedback and perform acceptably teaches us much that is useful also in looped feedback applications.
Ciao T
Hi,
Yes, there is the cryptically named myrefc which combined a mediocre op-amp and a chipamp inside it's feedback loop. Some swear by it, seeing what it takes to be stable and what the main op-amp is I am not as certain.
Ciao T
Yes, there is the cryptically named myrefc which combined a mediocre op-amp and a chipamp inside it's feedback loop. Some swear by it, seeing what it takes to be stable and what the main op-amp is I am not as certain.
Ciao T
As the Putzeys article explains, "nested feedback is functionally equivalent to global feedback." What you describe seems twice as hard as designing a single amplifier with proper feedback.
I have looked several times at the JC3 circuit before. Now i realize that there are two feedback loops and how innovative it was at that time. Anyway, i bring this up because i thought it satisfies both the high feedback camp plus the ones that claim high open loop bandwidth is necessary.
- Status
- Not open for further replies.