I am all ears.
And no, it has not to be exactly this device, can be any jfet, bjt or mosfet.
I doubt though that in the future I'll going to calculate distortion instead of measuring it....
Sure Ed, but it's just conjecture of your part that he didn't mention the Rk as feedback because he thought it wasn't.
Here's another conjecture: he was so preoccupied with getting his head around his new invention that he didn't realise that the Rk was also a form of nfb.
OK, one more: he realised late at night that the Rk was also a form of nfb but he thought, what the heck, I just wont mention it and see if anyone catches it.
My interest is in better understanding of circuits and if you realize degeneration is a form of nfb it clarifies several things at once: the gain of an ef being 0.9xxx and asymptotically going to 1.000 with the 100% nfb and a infinite transistor beta; or the fact that an ef can oscillate, as a result of phase shifted feedback with still present loop gain.
Occam's razor and all that; but maybe I'm just lazy
jan
But
Jan,
I don't want to argue what Black thought. It is a matter of record what he wrote,
The problem I have with cathode resistor type of gain reduction being classed completely the same as a distinctly added feedback path is that there are other changes that can occur.
In the deliberately added feedback resistor you must also consider the driving source impedance, stray capacitance and some even smaller considerations. These however still pretty much follow the basic feedback principle.
In the cathode resistor it changes the operating point of the tube. That may change the input resistor and capacitor. It raises the output impedance of the circuit along with other changes. So it is feedback with side effects that are non-trivial. (I can't quickly locate my Radiotron designers hand book so I'll limit my comment of parametric changes.) Those secondary effects are different than the other forms of feedback so I think it needs a warning label.
Pavel,
On the classic emitter follower you also have the collector current to deal with so your 100% feedback drops to 99% or maybe 99.6%. With some older transistors this could actually be as low as 90%! (OK, really old ones.) I do have my semiconductor book here so I could go into more details about the non-linear elements but I am pretty sure you know at least some of them.
Scott,
From my long ago days of playing with semiconductor design NPN's and PNP's were really just that. My understanding is that to prevent base spreading things have changed a bit. These changes if I understand it correctly provide a more linear Beta over the current range. Are there other issues that have changed?
Guys, (And Gals if one ever shows up here!)
For 99.9% of applications the deviation of reality from theoretical or mathematical models is trivial. It seems for high end audio it is not.
ES
Scott,
On this one we are going to disagree to a point. Back in 1967 folks were noticing that with triangle waveforms they got distorted results out of op amp circuits. So the explanation of slew rate limits needed to be broadcast to the masses. I believe that was the intent of this piece. The fact that it is applicable to more advanced topics is correct because the piece was correct on all the basics. But to say that it brought to light the same issues applied in audio power amplifiers is I think stretching it a bit far.
I do think once I get my library back in shape I might go through the old classic circuit books to see when triangle wave generators became popular. I think it was the use of IC op amps that made it so. I do know sawtooth waveforms were much used from the earliest days of scanning and television. But to me triangle waves and slew rate go hand in hand. Since there is little practical use of triangle waves other than to check linearity with a scope, the issue almost seems circular.
ES
Negative Feedback theory doesn't start or end with Harold Black
Nyquist, Bode among others had a bit to say to understand the stability issues – Bode’s book is often cited as a foundational feedback theory document
the impedance transformation relations characteristic of feedback were articulated as late as 1943 by Blackman
“Classical Feedback/Control Theory” today is taken as meaning extensions, refinements of Bode, Nyquist frequency response methods for stability and sensitivity analysis
it is the formulation 1st taught in EE courses – was fairly settled as a body of theory and applications by ~1960s
and it is a continuing field of exploration with occasional new result - although strongly nonlinear systems attract the most attention today
Middlebrook in particular was an advocate of “Classical” presentation, tools – with his own contributions intended to “make the math work for you”
“Modern Feedback/Control” is formulated in Linear Algebra State Space formulation and more abstract tools like Lyapunov stability criteria are applied in a “theorem proving” style that is hard on this engineer’s brain
Nyquist, Bode among others had a bit to say to understand the stability issues – Bode’s book is often cited as a foundational feedback theory document
the impedance transformation relations characteristic of feedback were articulated as late as 1943 by Blackman
“Classical Feedback/Control Theory” today is taken as meaning extensions, refinements of Bode, Nyquist frequency response methods for stability and sensitivity analysis
it is the formulation 1st taught in EE courses – was fairly settled as a body of theory and applications by ~1960s
and it is a continuing field of exploration with occasional new result - although strongly nonlinear systems attract the most attention today
Middlebrook in particular was an advocate of “Classical” presentation, tools – with his own contributions intended to “make the math work for you”
“Modern Feedback/Control” is formulated in Linear Algebra State Space formulation and more abstract tools like Lyapunov stability criteria are applied in a “theorem proving” style that is hard on this engineer’s brain
Several popular early audio power amps were just that, big power op-amps. The app note discussed slew rate in the context of full power BW and distortion. The amount of stretch is a matter of opinion.
It isn't. Now isn't that a relief
jan
Gosh Jan, that's swell of you. But there's a dozen other guys on this thread that are willing to hit you over the head with gold-standard textbooks and meter-long equations who will disagree with you to the death.
But fear not Jan - those textbooks hurt a whole lot less than the monster 10,000-hour listening log books. Those are the real killers.
Funny thing is that I never have the fortune to see these books and equations. I just have to *assume* they do exist.
Maybe I should show good faith and start with some 'gold standards' myself.
The name Peter Baxandall ring a bell?
(This is a small excerpt from "Baxandall and Self on Audio Power", Linear Audio Classic/collected papers, to be published).
jan.
Attachments
Last edited:
Gee Scott,
Like I said we disagree. But to be clear I assume we are talking about http://www.analog.com/library/analogdialogue/cd/vol1n3.pdf#page=1 This is a very good article about op amps in general.
The discussion of full power response is very good and applicable to the op amps of the time. It discusses slew rate limiting primarily due to the output transistors' limits.
That certainly was a problem in very early solid state audio amplifiers. But the issue with earlier stages being the limit was to my way of thinking an evolutionary development and credit for broadcasting that belongs to others.
So we will just disagree.
ES
Jan,
Then the issue is strictly over wording. Today it is more common to have multiple discrete feedback on stages of the over all system. If we call the cathode resistor style local feedback, then what do we call the stage by stage discrete feedback?
I like Nelsons term of degenerative feedback. It really is more accurate. The reduction in gain is done by changing the way that stage behaves in more than just gain.
But you are free to use what ever labels you like. Of course my pending article on feedback will get confusing without clear labels.
ES
Ed we don't necessarily disagree. I just get a bit weary of all those gold standards and equations and whatnot that are presented as showing that I and others have it all wrong, that never materialize. I thought I'd take the first step 
Err... your article? Why don't I have that yet??
jan
Err... your article? Why don't I have that yet??
jan
Except that you did. You stated words to the effect that, because Black did not write that Rk provided nfb, that therefore he thought that it wasn't.
I love you too.
jan
Ed please allow for the sematics of these early papers. It is clear to me what they are talking about, the output of the Vas. If you read on distortion at the input summing junction is directly mentioned which is the key idea in input stage non-linearity and the ideas that followed over the years. Since the op-amp is a first order integrator the error signal is a square wave to a triangle wave excitation, this is a direct measure of the + and - voltage the input stage is stressed at for a given slew rate. BTW with tiangle wave excitation the crossover blivit shows up right in the middle of the square wave at the input. Makes a lot clear.
Attachments
Last edited:
Actually in a memoir of Black's that I can't find, he discussed the issues and delays on getting a patent. One of the biggest was that everyone KNEW that feedback caused oscillation! The evolution of what we call feedback is still an ongoing process.
I am not sure what math you mentioned. I haven't seen a complete treatment of, for example the cathode resistor. It ain't simple!
You ain't seen nut'n yet as the work is still in progress. I am sure you will be surprised to learn some of the experimental results do not agree with some presentations of the math.
ES
Scott,
We really do see things differently. I read that same material as meaning the inverting input of the op amp is the summing point as shown in figure 8!
But I think it is a really good paper... are you going to argue with that?
ES
Right Ed, especially emitter followers and source followers may be pretty tough, oscillating on higher higher frequencies than an average scope would reveal
I think it should be clear where the problem starts. The term, feedback, can be used both in a general sense and in a specific sense. Audiophile marketing has succeeded in training non-engineer consumers to believe that 'feedback' is 'bad' and that they should pay more money for a better amplifier with No Feedback. It seems obvious to me that degenerative feedback is a kind of feedback, and there are many kinds of feedback. Just because each kind of feedback is not exactly like the others does not mean they are not, still, a kind of feedback. In other words, you can say that all feedback is alike, and you can say that all feedback is different, and both statements would be true. It does lead to a lot of arguing.
The more subtle point is when the mathematicians point out that each variety of feedback can be reduced to exactly the same equations. That takes a while to wrap your brain around. Some say the equations are wrong, some say that degeneration and global NFB are effectively the same, if not exactly the same in every minor detail.
All I know for sure is that the marketing focus on 'feedback' really frustrates me. I showed up here earlier this year after reading many of Nelson's articles, and I was really excited to share what I was learning with local friends. The problem is, the first conversation I had where I mentioned that I was reading and learning about how to design audio power amplifiers, particularly from the renowned Nelson Pass and Bob Cordell, the immediate response was "So, you're learning how to design amplifiers without feedback?" My friend is a very technical end user of audio electronics, but not at the circuit level by any stretch. For him, these are all black boxes with mysterious magic smoke inside. That I would have to explain how negative feedback is just as important as driving with your eyes open was, needless to say, quite frustrating. It actually rather pisses me off that people who haven't the slightest idea how to read a schematic are firm in their belief that they are making a compromise if they purchase an audio power amplifier that utilizes feedback.
I really do not understand the distinction between a circuit that has an obvious 'loop' and a circuit which does not. It's well known that an amplifying stage can have its gain affected by the load, and if the load is reactive then the gain changes are frequency dependent. If the load can affect the gain without a specific loop back to the input, then why is it so important to distinguish non-loop feedback? Just because the schematic does not depict a visual 'loop' does not mean it does not involve feedback.They might put an explanatory note (maybe they did?) into user manual, like:
The term "Feedback" herein means global, loop negative feedback. The statement "zero feedback amplifier" means we do not use the global loop negative feedback, but we may use many kinds of feedback in more common meaning. We use non-loop negative feedbacks like followers, emitter degenerations, nested feedbacks and many other kinds of this circuit design techniques
In other words, I like your idea of putting some fine print in the User Manual (that nobody will read after listening to their salesman and spending vast amounts of cash), but I think that the wording needs a few editing stages before publication.
Class D amplifiers often make use of triangle waves, as do switching power supplies and PWM LED driver circuits. I hope I am not trying to understand your comments out of context.
Last edited:
Scott,
We really do see things differently. I read that same material as meaning the inverting input of the op amp is the summing point as shown in figure 8!
But I think it is a really good paper... are you going to argue with that?
ES
Jeez Ed what do think I mean?? You're losing me again, that's exactly what I meant. You make the op-amp reproduce a triangle wave and you see a square wave at the inverting input, etc., etc. This is a direct observation of what you are making the input stage do i.e. for slew constant current into Cdom, so you tilt the input with a "DC" voltage.
EDIT - I did edit for clarity before you replied. Now that you made the same observation, I don't see how you don't see it. Just put a divider to the error voltage and you get Bob Pease's proof the the low THD of Nationals latest op-amps. Old ideas never die.
Last edited:
Nobody said that it is completely the same. It is completely feedback, though.simon7000 said:
Cathode degeneration does not change the operating point, as it will already have been included in any cathode bias calculation. It raises the output impedance. Some other types of feedback lower the output impedance. So what? Still feedback. Why does one type of feedback need a special warning while others do not? Now if someone has told you that feedback always reduces output impedance then I can understand why you might say this, but whoever told you that was mistaken. Negative feedback tries to maintain the aspect of output that is sampled, so voltage sampling maintains output voltage (i.e. reduces o/p impedance) while current sampling maintains output current (e.g. raises o/p impedance). Similarly at the input end: series nfb raises input impedance, shunt nfb reduces it.
Degeneration is current sampling at the output, and applied in series with the input, so it raises both impedances. Just as feedback theory says it should!
- Status
- Not open for further replies.