I think Bobken is John Linsley Hood's friend. From his saying, JLH concerns about what happened in Mhz figures (far above 20khz)
http://www.diyaudio.com/forums/showthread.php?postid=841752#post841752
Really now..........
A lot of "experts" should learn Bode, Nyquist, and Root Locus analysis. And then learn how it has been mis-understood, mis-applied, and mis-sold, and maligned. There are plenty of good articles explaining the mysterious and horrible effects of feedback, when done WRONG. And it is not so mysterious... at least for the last 50 years or so.
An amp with 200 - 300 kHz open loop response, carefully closed with judicious amounts of local and global feedback is a beautiful thing, and the reasons why, are NOT a mystery. Tube amps can get by with less feedback and bandwidth (say 70 - 100 kHz) because they are fundamentally more linear in the first place.
Without feedback... 95% percent of our world would not operate. Case in point, drive your car, on your most favorite and well known route... with your eyes closed...................
A lot of "experts" should learn Bode, Nyquist, and Root Locus analysis. And then learn how it has been mis-understood, mis-applied, and mis-sold, and maligned. There are plenty of good articles explaining the mysterious and horrible effects of feedback, when done WRONG. And it is not so mysterious... at least for the last 50 years or so.
An amp with 200 - 300 kHz open loop response, carefully closed with judicious amounts of local and global feedback is a beautiful thing, and the reasons why, are NOT a mystery. Tube amps can get by with less feedback and bandwidth (say 70 - 100 kHz) because they are fundamentally more linear in the first place.
Without feedback... 95% percent of our world would not operate. Case in point, drive your car, on your most favorite and well known route... with your eyes closed...................
I understand now that feedback amplifier with
1. low open loop gain (50 db)
2. low open loop distortion (0.5%)
3. wide open loop fr. band (20k)
4. properly compensated (Nyquist)
sounds better that one with 1.100db/ 2. 20%/ 3. 1k/ 4. the same.
Logically no feedback amp with
1. 28 db
2. 0.05%
3. 20k
4. not compensated
will sound much better !?
1. low open loop gain (50 db)
2. low open loop distortion (0.5%)
3. wide open loop fr. band (20k)
4. properly compensated (Nyquist)
sounds better that one with 1.100db/ 2. 20%/ 3. 1k/ 4. the same.
Logically no feedback amp with
1. 28 db
2. 0.05%
3. 20k
4. not compensated
will sound much better !?
john curl said:
It's clear.
Poor designer never can do good feedback or no-feedback amplifier. When top-class designer construct AB feed amp
it may sounds better than many (include my) class A no feed amp.
But this is prove nothing.
When top-class designer will construct no feed amp also and
compar it to his feed amp then we talk.
Rozak... I guess you are more concerned with the method and not the end result... the designer's journey and not the destination. I would imagine many of todays designers are very frustrated because they must adhere first to today's fashion and second to sound engineering principles in order to satisfy the public.
It is a good thing that engineers building MRI machines, GPS systems, airplanes, etc... do not have this problem.
I think I get something from MikeB's posts.
High feedback design (=high OL gain) is fine IF the OL distortion is already low.
What is bad : an amp that have high OL distortion, trying to lowering it by tons of feedback. I mean in "sonics"/audible terms.
So, not all high-feedback (ie, both have final distortion of 0.005%)will sound the same. Is this right, MikeB?
What is the guidelines to make low OL distortion design?
lumanauw,
Two feedback amps with identical distortion could sound very different. The open loop performance of each amp will again be the determining factor.
I would propose that bandwidth rather than linearity is the more important factor. It is very desireable to have the open loop response at least 10 times that of the desired closed loop response (this is rarely possible in mechanical systems or high speed electrical systems... but, it is easy enough in audio systems). This minimizes phase shift and its effects with frequency multiplication (spectrum spreading due to FM).
Our current methods of measureming THD do not detect this spectrum spreading. Brian Beck is conducting interesting research and describes some of this in the "blowtorch amplifier" thread.
Two feedback amps with identical distortion could sound very different. The open loop performance of each amp will again be the determining factor.
I would propose that bandwidth rather than linearity is the more important factor. It is very desireable to have the open loop response at least 10 times that of the desired closed loop response (this is rarely possible in mechanical systems or high speed electrical systems... but, it is easy enough in audio systems). This minimizes phase shift and its effects with frequency multiplication (spectrum spreading due to FM).
Our current methods of measureming THD do not detect this spectrum spreading. Brian Beck is conducting interesting research and describes some of this in the "blowtorch amplifier" thread.
Greetings from Norfolk
When a good designer designs a feedback amplifier he starts off by designing a good (minimal distortion) forward path amplifier. He then assesses this and adds the required feedback, either global or local. In escence the basic amplification 'elements' are designed for quality and linear operation.
In this way the good designer produces an amplifier, which when the feedback is added, is excelent, the feedback not being required to correct basic errors in the amplifier forward path, but only to further reduce the remenant errors.
Richard
rozak said:
When a good designer designs a feedback amplifier he starts off by designing a good (minimal distortion) forward path amplifier. He then assesses this and adds the required feedback, either global or local. In escence the basic amplification 'elements' are designed for quality and linear operation.
In this way the good designer produces an amplifier, which when the feedback is added, is excelent, the feedback not being required to correct basic errors in the amplifier forward path, but only to further reduce the remenant errors.
Richard
Excellent point Gandalph,
A proper forward function (open-loop behaviour) is paramount to good performance in all things. Often in thermal systems using resistive heating, a "square rooter" is employed so that the input voltage (or error signal) is converted to power at the output which is what the resistive heater requires to proportionately create temperature.
While this is not strictly required, I have been able to make temperature regulation systems that can hold 0.005 degree C.
Feedback done right is a beautiful thing...
A proper forward function (open-loop behaviour) is paramount to good performance in all things. Often in thermal systems using resistive heating, a "square rooter" is employed so that the input voltage (or error signal) is converted to power at the output which is what the resistive heater requires to proportionately create temperature.
While this is not strictly required, I have been able to make temperature regulation systems that can hold 0.005 degree C.
Feedback done right is a beautiful thing...
About the high order harmonics generated by nfb...
Some interesting tests about the effects of nfb can be done in sims with a virtual amp. No phasehifts/delays, only plain gains and feedbacks. This shows the naked face of nfb... This circuit gives openloop a nice and plain 2nd harmonic.
The simulationresults are: (outputvoltage 1v)
Green: OL-gain of 1000, OL-distortion 10%
Red: OL-gain of 100, OL-distortion 1%
Blue: OL-gain of 1000, OL-distortion 1%
Yellow: OL-gain of 32000, OL-distortion 1%
Observe that the nfb added high order harmonics that are not existant in the OL-dist.
The degree of the harmonics falling with frequency seem to be set by openloop distortion.
This shows just one thing: no matter if no feedback, low feedback or high feedback, OL-distortion must be kept at a minimum. (not a surprise)
This does not only apply to global nfb, local feedbacks have the same problem. This means, a single miller cap can create high order harmonics for higher freqs.
Unity gain feedback shows the same symptoms.
Lumanauw, to your question, simple r-loads to the vas do a great job, they reduce distortions without local feedback AND flatten out the OL-bandwidth. They simply reduce distortions the same amount they reduce gain, at first sight this does not make sense, but through the nature of nfb it does reduce high order harmonics.
Symasym for example has <1% OL-distortion (checked with 3V 20khz into 4 ohms) and 89db OL-gain.
With low enough OL-distortion and high feedback the high order harmonics created by nfb will be below noise.
Observe the yellow distortion, the first high order harmonic created by nfb is already at -140db, below a typical noisefloor of -120db. Here, the negative effects from nfb become negligible.
These attached result shows, that if nfb is misused, it does VERY bad things.
This is no mystique voodoo, just plain physics/math.
The real problem with nfb is stability.
Mike
Some interesting tests about the effects of nfb can be done in sims with a virtual amp. No phasehifts/delays, only plain gains and feedbacks. This shows the naked face of nfb... This circuit gives openloop a nice and plain 2nd harmonic.
The simulationresults are: (outputvoltage 1v)
Green: OL-gain of 1000, OL-distortion 10%
Red: OL-gain of 100, OL-distortion 1%
Blue: OL-gain of 1000, OL-distortion 1%
Yellow: OL-gain of 32000, OL-distortion 1%
Observe that the nfb added high order harmonics that are not existant in the OL-dist.
The degree of the harmonics falling with frequency seem to be set by openloop distortion.
This shows just one thing: no matter if no feedback, low feedback or high feedback, OL-distortion must be kept at a minimum. (not a surprise)
This does not only apply to global nfb, local feedbacks have the same problem. This means, a single miller cap can create high order harmonics for higher freqs.
Unity gain feedback shows the same symptoms.
Lumanauw, to your question, simple r-loads to the vas do a great job, they reduce distortions without local feedback AND flatten out the OL-bandwidth. They simply reduce distortions the same amount they reduce gain, at first sight this does not make sense, but through the nature of nfb it does reduce high order harmonics.
Symasym for example has <1% OL-distortion (checked with 3V 20khz into 4 ohms) and 89db OL-gain.
With low enough OL-distortion and high feedback the high order harmonics created by nfb will be below noise.
Observe the yellow distortion, the first high order harmonic created by nfb is already at -140db, below a typical noisefloor of -120db. Here, the negative effects from nfb become negligible.
These attached result shows, that if nfb is misused, it does VERY bad things.
This is no mystique voodoo, just plain physics/math.
The real problem with nfb is stability.
Mike
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.