Maybe this'll clear it up: Nyquist Stability Criterion. The math gets a little deep, but that's because you can't assume resistance only. (Be great if you could because there'd be no possible instability from DC to daylight.)
It's not a case of two end points with no feedback. It's a function of phase shifts at extremes of frequency where the open loop phase shift approaches 180 deg thereby turning negative feedback into positive feedback. The usual suspects are coupling/decoupling capacitors, and OPTs and ISTs that introduce phase shifts as they lose inductive reactance at the low end. At the high end: device internal and stray capacitance, and leakage inductance and stray capacitance in OPTs and ISTs.
If that happens, and there is enough open loop gain remaining, then you get a number of nasty side effects: a rising amplitude with frequency response that can excessively emphasize bass or treble response. If sufficient gain remains: out right oscillation. At the low end, that can mean "motorboating" or "breathing speakers". At the high end, RF interference, an amp that seems to go into overdrive at unusually low volumes, excessively hot finals, or in the case of SS amps, poofed finals.
The usual "fixes" are phase compensation capacitors across the feedback resistor and/or Zobel networks at the output of the pre that also contains the NFB summing node. These help by increasing the phase margin.
The "no feedback" condition occurs during a hard clip. The correction signal disappears, gain increases, deepening the clip.
Wish I did, but that was back in Ought Five or Six, and it definitely didn't occur here. I do recall coming to Lynn Olson's defense regarding a topology called "quasi-complimentary" that allows the use of the same finals for both phases, as opposed to fully complimentary pairs.
It's one of those cases that are extremely difficult to argue. Yes, Doug Self was 110% correct: full complimentary emitter follower drivers and fully complimentary emitter follower finals do measure better. It is a more convenient topology that I have used when designing long and short wave QRP rigs. However, they don't sound as good: upon that, Olson and I agreed. I've tried it both ways, and Q-Comp just sounds better, as do Sziklai pairs, and I just don't like the sound of fully complimentary finals. Those designs sound "off", and I suppose a lot would agree: I haven't seen anyone praising Self's "Blameless" design as being particularly desirable to listen to.
Subjects like these go without a mention in EE school. Is it any wonder why so much of production SS designs sound mediocre at best? They were designed by folks who literally don't know what they're doing.
Is that a personal preference? You bet it is! That's one thing Self tends to forget: the amp is just one subsystem of a complete audio system. You can account for a lot of variables, measure and characterize a lot of it from studio mic to loudspeaker. However, there remains that most variable of variables, that you simply can't measure reliably: the end user. If it were that easy, the whole matter could be settled once and for all: no more arguing over solid state v. hollow state, SE v. PP, pents v. triodes, etc, and ad infinitum. We could come up with that perfect amp everyone likes regardless of preferences in music and shut down these forums forever.
It ain't gonna happen. Self can sneer all he likes at "Subjectivists", but that will never change the fact that listening is Subjective.
Miles - not to repeat old hat (or perhaps yes, but not in a disparaging tone then)!
Your first statement: It sounds like a somewhat thick generalisation ... but you would probably desire me to read carefully, and I must then agree - although it does give the impression that most of us FC'ers don't know what we are doing! Still, let my defence be less of an impression only.
By the very recognition of subjectivity, does one not then argue for the fact that judgment of what is inherently good/poor by hearing, is questionable regarding definition of quality? Perhaps to clarify my meaning: If one shops for the equipment of one's choice, then by all means listen! It would in fact be dishonest to choose otherwise. But as soon as subjectivity/taste/whatyoucallit is recognised, I as a designer have a difficulty: For whose taste must I design? And this is going to be repetition so I will try not to bore: How do we define 'high-fidelity' if not 'truth to the original'? If subjectivity/taste is recognised, it follows that whatever a designer does, there will be folks out there liking his design and others disliking it. The amplifier then becomes a musical instrument in however small degree.
We know that QC started because of absence in early days of suitable complementary pairs. And similarly (and sadly
) even nowadays, FC pairs are not ideally matched (though I understand specials come close). Again analyses (measurements, Spice-wise ...) show that FC simply yields better linearity. After that .... what does scientific integrity require me to do? Back to hearing results, ad infinitum?Hopefully you will at least understand my predicament even if you do not agree with me. How do I sort out whose hearing is more justified than others - and even if that was possible, can I then blame 'mediocre' sounding equipment on the output topology only?
As promised, not to belabour. But I do have a difficulty with using taste as the main yardstick of judgment of the quality per se of my amplifier. That also so often followed by the "there are audible things that science cannot yet measure" routine. Sorry, not enough for me! The worthy John Curl once had as slogan; "Condemnation without examination is prejudice". And recently I noticed another pronouncement: "An argument without justification can be dismissed without justification" (exact words may be different - Christopher Hitchins I think.)
The problem with "truth to original" is, different symptoms of "false to original" are perceived differently.
Even the same person at different times, or listening to different records, pays more attention on different kinds of distortions. That's why we still have so many solutions, and no one can satisfy everybody. I myself have many designs. All of them are nice, but I can't name the one that is the best.
May be Tower design sounds the closest to the reality, but consumes enormous power and generates lots of heat. And it is heavy. Pyramid is powerful, generates less heat, but I would still prefer Tower for home music listening when possible. May be it is subjective because I believe that "no compromise" class A that uses active components in as linear as possibly regimes must sound better, I don't know.
My Gubernator sounds very real, however it can't drive such speakers like Pyramid and Tower can do, due to more limited output power. How to compare different amps that drive different speakers and tell which amp is better?
What would you prefer, more flat frequency response, or less of dynamic distortions? It is hard to tell. Some speakers have flatter response, but require higher power. If to equalize them in the amplification chain, before power amplifiers, they are used more effectively, but the equalizer adds own distortions.
Which evil is less evil?
It's still all about equilibrium, according to personal tastes.
Even the same person at different times, or listening to different records, pays more attention on different kinds of distortions. That's why we still have so many solutions, and no one can satisfy everybody. I myself have many designs. All of them are nice, but I can't name the one that is the best.
May be Tower design sounds the closest to the reality, but consumes enormous power and generates lots of heat. And it is heavy. Pyramid is powerful, generates less heat, but I would still prefer Tower for home music listening when possible. May be it is subjective because I believe that "no compromise" class A that uses active components in as linear as possibly regimes must sound better, I don't know.
My Gubernator sounds very real, however it can't drive such speakers like Pyramid and Tower can do, due to more limited output power. How to compare different amps that drive different speakers and tell which amp is better?
What would you prefer, more flat frequency response, or less of dynamic distortions? It is hard to tell. Some speakers have flatter response, but require higher power. If to equalize them in the amplification chain, before power amplifiers, they are used more effectively, but the equalizer adds own distortions.
Which evil is less evil?
It's still all about equilibrium, according to personal tastes.
Thanks DF96
My state of mind is normally confused; perhaps well into the state of delusional and with a strong lack of sense to know what direction to blow at the wind. My object herein is to bring into as stark a light as possible the questions I have as particularly related to the significance of near resonant systems and why such systems by their nature could present sonic improvements by my experiences.
In consideration of your presentation let’s consider a more extreme theoretical case. In this case consider that the network is sitting on the threshold of a loop oscillation, this time of infinite frequency. This creates a theoretical situation that the AC audio input signals are of such low frequency in relation to the infinite frequency that the changes can be considered as all DC (infinitely low AC) and that the amplitude changes in the DC are of miniscule amount before being corrected in the infinite high frequency near oscillation loop. This can be compared to an identical circuit having no tendency toward loop oscillation. Under such conditions it is thought that both circuits would appear behaving identically and in accordance to any phase, frequencies, etc., as you suggest. My question is specific to how and why they should and if any significance exists in the variances.
Intuitively there seems must exist two independent mechanisms to cause the same result. One is the normal condition of feedback and the other seems can only be a function of the creation and theoretical shift from the quadrature of the near resonant infinite frequency pole. Certainly it seems an exceptional spectrum analyzer would peak at the infinite frequency. This is to suggest that the gain is dominant at the quadrature of the infinite frequency and that disturbances only result in the creation of that frequency with a phase shift from the quadrature.
In normal systems wherein the differential gain between open loop and closed loop is increased the difference between the input and feedback signal diminishes and moves into the noise floor of the system. This leads to a condition that increasing open to closed loop gain can result in degradation as a function of dealing with more noise. This is not to suggest that the noise floor increases rather that the difference signals are decreasing. In the theoretical system having a near resonant pole the loop corrective mechanism seems would be dominated by the generation of a signal being off the quadrature.
This leads to me wondering if the mechanism under such a theoretical condition can be considered as being caused by disturbances to the equal of a carrier that indirectly causes the variances in the signal. In other words the normal random noise, gain, phase etc., as normal responsible for changes is overpowered and replaced by the dominant singularity of variations to a carrier. From here the thought moves to theoretically shifting this carrier toward the top of the audio band whereupon it may be that the more critical noise is being filtered out or blocked above the pole frequency.
I agree with you about Self and his sneering at "subjectivists" by doing so he ignore the basic question What is the purpose of the "Hi-Fi" system . To my mind it is the enjoyment of the end user ! Take for example watches some costing 10s of thousands of dollars that when tuned up are almost as good a 15 dollar digital watch. Keeping time is not the only thing . Thus as wavebourn said many solutions for many end users.
i agree that when it comes to listening to music we are all subjectivists, but when discussing technical matters like circuits, we can not just say that global negative feedback ruins the sound or that sand in tube circuits ruin the sound unless the poster clearly states that such was his opinion......
Well, Sir, you could bluff me!
Yes, this is true. That is why one cannot view this outside of the practical optimal case, i.e. in regions into which nfb should not progress.
I think one should keep these 'limitations' in mind. In practice one has a lower and upper limit of operation; let's use the ubiquitous 20Hz and 20kHz. To the degree that the open-loop design within these limits is kosher, i.e. constant gain and very little if any phase shift, the end result will not venture into the regions discussed above.
It is possible with both tube and semiconductor design to keep open-loop gain within 1 dB and phase shift within some <5 deg. of 180. One then has the condition where more than enough nfb to reduce all distortion products to well below audibility without danger of instability, can be applied and no more. I.e., the circuit should be capable of rolling off outside these limits, within the requirements for stability of the phase-amplitude relationship. There should be no need for considering what might happen under the previous post's conditions; those should not be allowed to exist in the first place. This is also basic for attaining equal reduction of all harmonic orders as mentioned before. It is when going outside these conditions that harmonic multiplication and other nasties start rearing their heads, as described previously.
[And sadly in practice they do (Miles mentioned existence of mediocre designs) .... one reason that I am unfond of the internally compensated op-amps with their 90 degree phase shift over most of the audio range in order to guarantee stability under all conditions, including other topologies with similar tendencies. It is possible to keep C.dom or its equivalent from beginning to act until above 20kHz - but that is a another story.]
P.S: New posts appeared while I was typing. By the "above post" I was referring to Hierfi's previous post. #170
Last edited:
True here we are dealing with two different thing 1. Purpose of the system and 2. How we build the system . So is it God or the devil that we find in the detail ? For me the use of global feedback is a big bandaid on a broken leg. `Historically it use was to stabilize amps that did not have power supplies up to the task and not enough stage isolation from each other. Given the time it came into use with what they had to work with it was a very good idea. Very bright men made it work for control that stability problem and better bandwidth . I would rather make each stage as stable as possible with isolation in power supply from each other and not use global feedback . The production of high odd order distortion if you use less say about 21db of feedback and first stage overload are some of the reason I try to stay away form global . This a field with more grey areas than just black and white .i agree that when it comes to listening to music we are all subjectivists, but when discussing technical matters like circuits, we can not just say that global negative feedback ruins the sound or that sand in tube circuits ruin the sound unless the poster clearly states that such was his opinion......
I been to EE school myself, and know lots of others who have as well. I never had so much as one lecture on sonic performance, and neither did they. If I haven't, then neither have a lot of others.
I don't have that problem because I "eat my own dog food". I listen to what I design and build, and have not even had a production system since I retired the last one I had in 2005 after stripping it of anything that looked interesting. My preferences run towards Black Metal, Punk and Techno. Would I expect one who liked choral music to like Le Renard or the Vixen? Certainly not.
Then there's the question of who's listening and why. The one who slaps on a CD for background noise while he does something else is gonna be a helluvalot easier to please than one who sits and listens while doing nothing else. Design to the lowest common denominator, make it cheap, sell cheap and make up the difference on volume is a viable business strategy. It has ever been thus, and there are no shortage of hideous designs even back in the "good ol' days".
If this were not so, there would be no tone controls and graphic equalizers. The perfect lens for a telescope, microscope, ort camera has yet to be ground, and the perfect amp -- a "wire with gain" -- has never been built. I doubt that they ever will.
"Again analyses (measurements, Spice-wise ...) show that FC simply yields better linearity".
I already conceded that.
What's your purpose? Are you designing for the masses, the vast majority of whom seldom listen? Those who just slap on a CD for background noise? Are you aiming for a more discriminating type of listener? Those who designed for car audio have different criteria than those for a home setting.
I have a choice: if I test listen and don't like what I'm hearing, I don't have to buy, do I?
How many times do I have to say it? I've tried it both ways, everything else being equal, and I know what I prefer, and that's how I'm gonna do it. If you don't like the results I'm getting, then choose something different. There's a lot of choice floating all over these forums, and on just ESSSSSS-loads of other forums and web sites, both professional and amateur. I'm sure you can find something that suits you, even if it doesn't suit everyone else.
... and that alone or even mainly determines the success of your engineering for the rest of your life? Neither did I have a single lecture on sonic performance or even hi-fi as such! But I was taught the basics and hopefully the ability to learn, discern and gain experience! What about the wealth of research out there - in modern times made accessable by the internet? I was also taught not to re-invent the wheel - heaven help us if that is lacking from especially a professional engineer's background and he tries to cope only with what he was taught at school.
I appreciate your reply - but we are talking of two different things. I have appartently failed to make it clear that I was speaking for the manufacturer of hi-fi equipment, or for that matter any engineering activity, and not simply of DIY-ing for myself. With my own modest amplifier designs I have never used hearing tests as an ultimate tool. In fact, the first time a customer listened to some of my products was also the first time I heard it. I must be either singularly gifted or lucky because my products were never not liked by clients (and neither am I psychic to foresee what they would like). I do not thereby say such folks do not exist (we have had that stated as often as ....) or that subjectivity does not exist. I believe I did state that when you design for yourself you can add all the funnies you like - but when doing a design for clients unknown, I follow the science proved through the years.
If a better way is ever discovered I will be the first to adopt it - but I am not holding my breath after a half-century.
Lynn Olson describes some of the engineering reasons why gNFB sho0uld be avoided, the second point is interesting and none obvious;
I am actually finding it quite difficult to find designers who advocate the use of gNFB. Call that fashion if you like, but then I would call that arrogance.
Shoog
I am actually finding it quite difficult to find designers who advocate the use of gNFB. Call that fashion if you like, but then I would call that arrogance.
Shoog
Last edited:
You need to get out more. It's 99.99% of the professional engineering community.
And stop reading marketing documents and thinking that they're technical treatises. Hint, for classification purposes: if there's no math and no experimental data, it's unsupported handwaving.
This is the sort of statement I don't understand; much of the rest of what you say is hewn from the same rock. You may be able to attach meaning to this collection of words, but my training in theoretical physics and electronic engineering is obviously inadequate for the task. I will leave it to others.Hierfi said:
Is this a reverse Sokal hoax?
Several of the 'gurus' you've mentioned are not engineers; they're overgrown hobbyists. They are very good at giving audiophile schpiel and lacing it with a few general engineering truths that make the whole thing seem convincing, but their knowledge of actual maths and electronic principles is often patchy or scant. Unfortunately this is only obvious to people who are engineers. Other hobbyists don't know any better, and will swallow it wholeheartedly.
I'm sure they don't mean to mislead -they're self-taught after all, and are acting with honest intentions- but you can't use their words as academic proof of something (blind leading the blind).
*sigh* Olsen needs to learn basic electrostatics. Inherent feedback in triodes is covered in chapter 1 of most general 1940s/50s textbooks on basic valve theory, it's not just some new-fangled 'theory'.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.