The good, the bad and the ugly
Gentlemen:
I guess we are educated enough to acknowledge there is no reason to take camp in "GNF is all" or in "GNF is anathema".
That negative feedback is an extremely powerfull concept without which I cannot imagine analog(e) electronics as it is today, even in the genuine top audio segment should not be debatable.
At the same time, to try to conceal in G=A/(1+AB) a poorly designed A as long as making sure it is large, is equally certain to elicit strong objections from golden eared subjects, and they should not be flatly ignored just because measurements read 0.1% THD or even 0.01%.
Signals will certainly go through the A stage, and if it is bad, they will generate lots of correlated low level garbage. This is physics and cannot be avoided.
Yet this may be OK if we can make sure this garbage goes way down, below a frequency dependent threshold as much as a transient aberration threshold, something whose definition and quantification is proving to be a holly grail of high performance, definitive, objective, figure of merit.
Rodolfo
Gentlemen:
I guess we are educated enough to acknowledge there is no reason to take camp in "GNF is all" or in "GNF is anathema".
That negative feedback is an extremely powerfull concept without which I cannot imagine analog(e) electronics as it is today, even in the genuine top audio segment should not be debatable.
At the same time, to try to conceal in G=A/(1+AB) a poorly designed A as long as making sure it is large, is equally certain to elicit strong objections from golden eared subjects, and they should not be flatly ignored just because measurements read 0.1% THD or even 0.01%.
Signals will certainly go through the A stage, and if it is bad, they will generate lots of correlated low level garbage. This is physics and cannot be avoided.
Yet this may be OK if we can make sure this garbage goes way down, below a frequency dependent threshold as much as a transient aberration threshold, something whose definition and quantification is proving to be a holly grail of high performance, definitive, objective, figure of merit.
Rodolfo
Rodolfo,
My feeling is that the answers need to be sought not in the feedback concept itself, but in those physical events that are left out of the simple A.b/1+A.b feedback equation, so to speak.
Things like the complex impedance of the speaker, the fact that intermodulation and HD products keep going 'around' through the closed loop. Also, in case of large error feedback, that the amp internally starts to be very non-linear and no longer satisfies the condition that A should be very large. I think it is those dynamic events that are not captured in the simple concept, that separate the men from the boys in amplifier country, so to speak.
I do believe that there is nothing inherently wrong in neg feedback. It permeates our lives to the extend that we all would die a sudden, horrible death if the feedback mechanisms in our bodies would stop. And Mother Nature is known to develop solid systems, over the millions of years she had available to do R&D.
Jan Didden
My feeling is that the answers need to be sought not in the feedback concept itself, but in those physical events that are left out of the simple A.b/1+A.b feedback equation, so to speak.
Things like the complex impedance of the speaker, the fact that intermodulation and HD products keep going 'around' through the closed loop. Also, in case of large error feedback, that the amp internally starts to be very non-linear and no longer satisfies the condition that A should be very large. I think it is those dynamic events that are not captured in the simple concept, that separate the men from the boys in amplifier country, so to speak.
I do believe that there is nothing inherently wrong in neg feedback. It permeates our lives to the extend that we all would die a sudden, horrible death if the feedback mechanisms in our bodies would stop. And Mother Nature is known to develop solid systems, over the millions of years she had available to do R&D.
Jan Didden
Jan:
Absolutely.
And we should also be aware of something named causality, that usually translates mostly as an afterthought in phase margin for stability sake.
But Black's 1927 equation assumes zero delay, so in the real world the error term is dealing with different - albeit slightly - beasts.
Rodolfo
Absolutely.
And we should also be aware of something named causality, that usually translates mostly as an afterthought in phase margin for stability sake.
But Black's 1927 equation assumes zero delay, so in the real world the error term is dealing with different - albeit slightly - beasts.
Rodolfo
ingrast said:
Rodolfo,
I am not sure. Let's speak like people graduated from Institutes of Technology, that we are.
A and beta have always been considered as functions of complex variable, so they do assume delay.
ingrast said:
That's what I wanted to add as well. We have the instruments of non-linear analysis, which is also one of the commonly lectured subjects.
Pavel:
What I wanted to stress, is that when we derive the textbook FB gain we write:
Vout=A*(Vin-B*Vout)
And merrily factor the leftmost and rightmost Vout as if one and the same. By large this is irrelevant but not to be forgotten.
Rodolfo
What I wanted to stress, is that when we derive the textbook FB gain we write:
Vout=A*(Vin-B*Vout)
And merrily factor the leftmost and rightmost Vout as if one and the same. By large this is irrelevant but not to be forgotten.
Rodolfo
john curl said:
It is of interest that this view was expressed only after Otala claimed it to be so.......
An elementary psychological error methinks...
Not at all. My first experience with open loop bandwidth involved a comparison of my own Comp Diff Class AB1 design (.5A bias) compared to a vacuum tube triode amp in 1969. Both amps were rated about 10W and drove a Klipshorn with 104dB/W sensitivity.
Both amps measured essentially the same closed loop bandwidth, IM distortion at various levels, ie less than .005% at 1W, slowly rising until clipping. Both had 2'nd and 3'rd order distortion, almost exclusively, both were stable with any capacitive load, and both had a damping factor of about 30.
We still heard a difference. Why? How?
Well the only major difference between the two amps, as I saw it then, was that the triode used 20dB of negative feedback and my amp used 40dB to get the same results. I can presume from this that the open loop bandwidth of my solid state design was about 10 times less, since the closed loop bandwith was the same in both amps of about 100KHz.
Otala's paper, gave me a better understanding of why the amount of negative feedback mattered, more than 1/3 of a century ago.
I don't always agree with Dr. Otala, but his input has been more useful to me, over the decades, along with Dr. Hawksford, than almost any other designer.
Both amps measured essentially the same closed loop bandwidth, IM distortion at various levels, ie less than .005% at 1W, slowly rising until clipping. Both had 2'nd and 3'rd order distortion, almost exclusively, both were stable with any capacitive load, and both had a damping factor of about 30.
We still heard a difference. Why? How?
Well the only major difference between the two amps, as I saw it then, was that the triode used 20dB of negative feedback and my amp used 40dB to get the same results. I can presume from this that the open loop bandwidth of my solid state design was about 10 times less, since the closed loop bandwith was the same in both amps of about 100KHz.
Otala's paper, gave me a better understanding of why the amount of negative feedback mattered, more than 1/3 of a century ago.
I don't always agree with Dr. Otala, but his input has been more useful to me, over the decades, along with Dr. Hawksford, than almost any other designer.
unconventional feedback
I was just looking thru some of the postings a ways back and noticed some thoughts on unconventional feedback correction. (around # 737)
Just wanted to add another crazy idea. Suppose one adds a small high frequency pilot signal to the actual input signal. This would be above the audio range so as to be possible to filter it out of the output easily. By using a small constant amplitude input pilot signal, one can measure the instantaneous forward gain function of the amplifier (synchronous detection at output)
Now the trick would be to modulate the tail current in the input diff. pair so as to change the amplifier gain so as to maintain a constant output pilot signal. Effectively, this would linearize the amplifier in real time. This has an advantage in not requiring excess forward gain in the amplifier, much like error correction, but not having a critical max gain limit like error correction. So some stability restrictions should be alleviated. One could also contemplate a PID (proportional - integral - derivative) like control loop to maintain constant pilot signal gain. Just an idea, no doubt some complexities to implement.
Don
I was just looking thru some of the postings a ways back and noticed some thoughts on unconventional feedback correction. (around # 737)
Just wanted to add another crazy idea. Suppose one adds a small high frequency pilot signal to the actual input signal. This would be above the audio range so as to be possible to filter it out of the output easily. By using a small constant amplitude input pilot signal, one can measure the instantaneous forward gain function of the amplifier (synchronous detection at output)
Now the trick would be to modulate the tail current in the input diff. pair so as to change the amplifier gain so as to maintain a constant output pilot signal. Effectively, this would linearize the amplifier in real time. This has an advantage in not requiring excess forward gain in the amplifier, much like error correction, but not having a critical max gain limit like error correction. So some stability restrictions should be alleviated. One could also contemplate a PID (proportional - integral - derivative) like control loop to maintain constant pilot signal gain. Just an idea, no doubt some complexities to implement.
Don
Re: unconventional feedback
This pretty much sums up the subjectivist position!
smoking-amp said:
This pretty much sums up the subjectivist position!
Hey! No subjectivists here.
I'm no subjectivist! That smoke and mirror line came from soldering smoke, burning components, and magnifying current mirrors circuitry with H. error correction. I'm going to change my signature now. I'm insulted. Boo Hoo
😀
Don
I'm no subjectivist! That smoke and mirror line came from soldering smoke, burning components, and magnifying current mirrors circuitry with H. error correction. I'm going to change my signature now. I'm insulted. Boo Hoo
😀
Don
Six weeks you've got...
Why don't you offer it as "pay if you like ware"
Why don't you deliver and we'll see if it worth any? The current limit thing you wrote was pretty good I think so obviously you have the ability.mikeks said:
Why don't you offer it as "pay if you like ware"
Mike, I ment that you could publish it here or at your homepage if you have any. If you behave I can even give you some space for a while. Let us see what you can achieve and you'll see how much money you can get. Set 5 dollars to be the price.
I think you can forget about the EW if you want any money so let's try other publishing methods.
I think you can forget about the EW if you want any money so let's try other publishing methods.
Make it 10 Euros to UNICEF, have the confirmation in paper (photocopy) sent to you, and ship a hard copy of your article. Some would prefer PDF, but at least no one's gonna send PDF files across the globe this way. And while you don't have to build up a credit-card infrastructure, you're not giving it away for free either.
Or try ebay fixed-price via paypal, say, 5 EUR, and post e.g. here (and /or there) an abstract of the paper. Make sure the price covers both S&H and ebay/paypal fees. And that there are no typos or other goofs in the article (you won't have access to EW's editing and proof-reading).
Welcome to the world of globalized markets. 😎
Or try ebay fixed-price via paypal, say, 5 EUR, and post e.g. here (and /or there) an abstract of the paper. Make sure the price covers both S&H and ebay/paypal fees. And that there are no typos or other goofs in the article (you won't have access to EW's editing and proof-reading).
Welcome to the world of globalized markets. 😎
Re: unconventional feedback
Interesting idea. Maybe it can be done with a DC component on the signal? Filter out the AC and use the recovered DC to directly modulate the LTP tail current; I think DC can be processed more accurately then detecting an hf component. Problem to be solved would be to filter DC from output. But let's see, maybe we can set it up so that the target condition is to get zero volts DC at the output? Hmm.
Jan Didden
smoking-amp said:
Interesting idea. Maybe it can be done with a DC component on the signal? Filter out the AC and use the recovered DC to directly modulate the LTP tail current; I think DC can be processed more accurately then detecting an hf component. Problem to be solved would be to filter DC from output. But let's see, maybe we can set it up so that the target condition is to get zero volts DC at the output? Hmm.
Jan Didden
- Status
- Not open for further replies.