Jan, if that was a criterion, in my case, I would think all threads are stimulating
I find this thread has a quality I have missed lately. Participant talk technical terms and stay there, no personal attacks.
They respect each other.
In the course of time, a public forum participant's quality that surfaces is not the IQ but the blend of hard skills (here technical knowledge) together with his/her soft (social) skills.
And you are all old timers here
George
You can't fight human knowledge at its peak...it's like saying chimps are smarter than humans 🙂 What that guy in the tree at the end said is what I had in mind : while we removed most of the chimp's natural predators discovering all the nature's physics laws, electricity, nuclear power and advanced mathematics, chimps seemed to have had all the fun for themselves... Maybe we should enjoy our GNFB amps more and doubt them less... let mathematicians spend their days arguing in favour or against it 😉
Last edited:
I wonder if the Chimps already know what Dark Matter is. All the Kings physicists and all the Kings mathematicians can't figure out what it is, except that it's everywhere around us. When all the candidates have been dismissed (getting close to that now), the last standing one has to be it, no matter how objectionable.
They probably don't care...they just try to enjoy the Earth while their stupider relatives allow it as for a relativistic being as chimps the time flows way slower than for those who can't perceive flow time... By the way...I digress here, but realized that the working memory is what makes the time slowing down.Starting with this late Pandemic I got to play chess online and realised game after game how frustrated I was when the clock stopped with me loosing on time and saw the opponent who I thought was as fast or even slower than me having more than a minute left at the end of the game .
Now I get why guys with good working memory enjoy the most of it:
https://www.celebritynetworth.com/richest-celebrities/authors/magnus-carlsen-net-worth/
As R.Penrose said once, from the point of view of a photon the time is infinite and there's essentially no end of universe or time at his speed.A photon is immortal and for a photon there's no time passing, no end or beginning of the universe while there's nothing faster than a photon ....The fastest guys in the universe have no ideea of anything happening...the smartest ones finish all the tests while those like me never have enough time for evolution 🙂 Boredom is the true sign of inteligence...
Now I get why guys with good working memory enjoy the most of it:
https://www.celebritynetworth.com/richest-celebrities/authors/magnus-carlsen-net-worth/
As R.Penrose said once, from the point of view of a photon the time is infinite and there's essentially no end of universe or time at his speed.A photon is immortal and for a photon there's no time passing, no end or beginning of the universe while there's nothing faster than a photon ....The fastest guys in the universe have no ideea of anything happening...the smartest ones finish all the tests while those like me never have enough time for evolution 🙂 Boredom is the true sign of inteligence...
Last edited:
You can also use Laplace transforms to calculate step or impulse responses of linear time-invariant systems, especially when they are lumped (have no pure delays). The e-sT factors that you get when there is a delay T complicate things.
My 5 cents, in my experience a small amount of NF can improve a good amp, and for sure a strong NF will spoil a mediocre amp.
In any case NF will not do magic in a tube amp. Operationals amps are a different story, they by design have to be used with NF.
Regarding the first question, NF signal should never be too strong to overcome the input.
In any case NF will not do magic in a tube amp. Operationals amps are a different story, they by design have to be used with NF.
Regarding the first question, NF signal should never be too strong to overcome the input.
smoking-amp,
Yes.
Start with no global negative feedback.
Use a delay line with a moderate time delay, and measure the results.
Start with no global negative feedback.
Put the delay line between the signal and the amplifier input, or put the delay line after the amplifier output, that gives exactly the same results.
Now, (re)-apply the negative feedback around the outside of:
1. delay line, amp input, amp output
2. amp input, amp output, delay line
Replace the delay line with several other delay lines, each one having increasing delay times, until the negative feedback does not work properly for signals from 20Hz to 20kHz.
How much delay time can a specific amplifier take, and not have a problem with the global all-inclusive negative feedback.
Now consider the delay line's delay time, plus the amplifier's intrinsic internal delay time.
Then consider taking the delay line out, and designing the amplifier to have more intrinsic internal delay time.
With the same total delay time, does the global negative feedback work?
It depends on the total delay time, regardless of where the delay(s) are, and does not matter if all of the delay time is in the delay line, or the amplifier, or is in both of them.
Yes.
Start with no global negative feedback.
Use a delay line with a moderate time delay, and measure the results.
Start with no global negative feedback.
Put the delay line between the signal and the amplifier input, or put the delay line after the amplifier output, that gives exactly the same results.
Now, (re)-apply the negative feedback around the outside of:
1. delay line, amp input, amp output
2. amp input, amp output, delay line
Replace the delay line with several other delay lines, each one having increasing delay times, until the negative feedback does not work properly for signals from 20Hz to 20kHz.
How much delay time can a specific amplifier take, and not have a problem with the global all-inclusive negative feedback.
Now consider the delay line's delay time, plus the amplifier's intrinsic internal delay time.
Then consider taking the delay line out, and designing the amplifier to have more intrinsic internal delay time.
With the same total delay time, does the global negative feedback work?
It depends on the total delay time, regardless of where the delay(s) are, and does not matter if all of the delay time is in the delay line, or the amplifier, or is in both of them.
Last edited:
Just a reminder of what I said in Post # 17:
"First of all, I am not trying to support/criticize: negative feedback, or non-negative feedback.
I like local negative feedback, global negative feedback, and non-negative feedback (when they are implemented properly)."
That means I am neither arguing for negative feedback, nor arguing against negative negative feedback.
Here we are, 173 posts later. I do not expect a final conclusion.
But, just a little advice, to take or reject:
Enjoy the Music.
"First of all, I am not trying to support/criticize: negative feedback, or non-negative feedback.
I like local negative feedback, global negative feedback, and non-negative feedback (when they are implemented properly)."
That means I am neither arguing for negative feedback, nor arguing against negative negative feedback.
Here we are, 173 posts later. I do not expect a final conclusion.
But, just a little advice, to take or reject:
Enjoy the Music.
I was just suggesting the delay line to demo the proposed multi-looping taking place, which would be visible initially with a step function input.. I would expect the amplifier to work best without any delay. It does bring to mind an idea I've seen where each side of a P-P amp has a separate feedback loop except the feedbacks are crossed somehow (must be some inversions). That way the Fdbk does a figure 8 around the amp. Might influence the Feldtkeller/Farren/Baxandall/Putzeys.... (FFBP distortion ?), but I'd be surprised if it does much. If you could just get the FFBP dist. to cancel out in the output P-P summing. But it probably sums to twice.
Hmm, how about feeding some pure input signal into the (inverted) N Fdbk loop signal to leave just the error, then cancel the full forward error in a single pass? That might just be equivalent to EC, but I'm not readily seeing positive Fdbk then.
Hmm, how about feeding some pure input signal into the (inverted) N Fdbk loop signal to leave just the error, then cancel the full forward error in a single pass? That might just be equivalent to EC, but I'm not readily seeing positive Fdbk then.
Last edited:
The original question was answered by Jan in post #2. In post #25 I raised the point that Black's 1934 model is an approximation, as explained by Baxandall and more recently Putzeys. MarcelvdG found the even earlier work of Farren and then Feldtkeller (still working on the translation, but Marcel's explanation was better than I could hope to do). This led to the discussion about granular functioning of the recursive amplifier. smoking-amp developed a working model that includes the causal, arrow-of-time, mechanism of the amplifier's integrator calculating (here we include B. Putzeys poetry) the output from its instantaneous voltage and the voltage's first time derivative. Putzeys' analogies of feedback to any predictable function are very trippy, but I'm not smart enough to follow along.
Is that formal engineering terms? No, but this isn't engineering, it's learning. Could be worse ways to spend our lives.
All good fortune,
Chris
Surely you mean the Fourier transform? Not the Laplace transform?
The problem with the differentiation explanation is that not all NFB networks have lead capacitors: yet they all work, other than those that are unstable, which is a stability problem, not an NFB problem.
The problem with the differentiation explanation is that not all NFB networks have lead capacitors: yet they all work, other than those that are unstable, which is a stability problem, not an NFB problem.
I would imagine that leaving out the lead capacitor for phase corr. would just cause the error correction to be a tiny bit less accurate on 1st loop pass. For a hypothezed multi-loop Feldtkeller/Farren/Baxandall/Putzeys distortion it would cause the phase accuracy to deteriorate with each pass around the loop. Causing the extended FFBP harmonics to get worse with each higher order. Looking at Baxandall's graph (courtesy of Jan Didden), one does see that occurring in the measured curves. Maybe the cap was off value some or missing?
Last edited:
The lead cap/phase corr. got put in earlier simply to show how the N Fdbk could be correcting the signal time coherently. One of the arguments raised against N Fdbk was that it was correcting error after the fact. It really doesn't have much importance to the Feldtkeller/Farren/Baxandall/Putzeys model or discussion other than the higher order distortions possibly being worse than expected.
Fourier/Laplace? Its been 50 years since I took a course on those methods. I thought Laplace was just using complex numbers compared to reals. Easier boundary conditions. I'm familiar with solving Laplacians for electromagnetic fields, maybe not the same thing for signals.
Fourier/Laplace? Its been 50 years since I took a course on those methods. I thought Laplace was just using complex numbers compared to reals. Easier boundary conditions. I'm familiar with solving Laplacians for electromagnetic fields, maybe not the same thing for signals.
Last edited:
Feedback doesn't happen "after the fact", because it's wrapped into the integrator (an essential element, for stability, of all feedback amplifiers). For audio amplifiers, the integrator's time constant is very large compared to the arrival times of all of the (significant, not noise-limited) recursive feedbacks plural, and input. Feedback is not a single event, but a series of successive approximations, approaching a very complicated (difficult for us to calculate) result. In B. Putzeys' poetry, the integrator "calculates" the output voltage by successive approximation. It keeps on doing it forever (noise limited) and continuously. I think smoking-amp's conception of the integrator working in an arrow-of-time but directed-by-dV/dt is the most challenging result of the discussion (so far!)
Much thanks to all for a great birthday (72), very good vibes,
Chris
Much thanks to all for a great birthday (72), very good vibes,
Chris
But as that isn't true, the reliance on differentiation is unnecssary, and breaks down in the case where the lead capacitance isn't present.
Without the cap puts the correction maybe a few nSec or so behind the current incoming signal. Not surprising that works pretty good. You can't hear 1 nSec in audio.