you really shouldn't be thanking people for egregiously wrong information
global negative feedback does change important parameters of even “perfectly linear” amplifiers in a desirable direction
there is a feedback “error” signal even without nonlinear distortion in the amplifier gain
negative feedback trades excess open loop gain of a amplifier for reduced variation of closed loop parameters such as gain, frequency response and output impedance at the price of a lower closed loop gain defined by the much more linear and stable feedback network
I can think of only 3 ways to linearize a gain stage (which have already been mentioned here and Ed Cherry has spelled out several times)
Increase the bias to signal ratio - use less of the nonlinear gain curve
Cancellation – really only useful with diff pair and push-pull emitter followers, and even order distortions
Negative Feedback – local feedback as in degeneration, emiiter/source followers or larger feedback loops including more gain stages
The differential pair input with degeneration can be linear enough for any practical application – a subtlety of global negative feedback is that forward loop gain directly reduces the signal at the diff pair input, increasing the bias to signal ratio by decreasing the signal and further “linearizing” the diff pair as a consequence of large negative feedback
you aren’t likely to learn the subject matter of 4+ (including prereqs) college level math/engineering courses in a discussion forum – especially in a audio related forum where many misguided amateurs are repeating really wrong but pervasive "theories" of audio pseudoscience and even some professionals may be disingenuous to support their "no feedback" marketing positions
global negative feedback does change important parameters of even “perfectly linear” amplifiers in a desirable direction
there is a feedback “error” signal even without nonlinear distortion in the amplifier gain
negative feedback trades excess open loop gain of a amplifier for reduced variation of closed loop parameters such as gain, frequency response and output impedance at the price of a lower closed loop gain defined by the much more linear and stable feedback network
I can think of only 3 ways to linearize a gain stage (which have already been mentioned here and Ed Cherry has spelled out several times)
Increase the bias to signal ratio - use less of the nonlinear gain curve
Cancellation – really only useful with diff pair and push-pull emitter followers, and even order distortions
Negative Feedback – local feedback as in degeneration, emiiter/source followers or larger feedback loops including more gain stages
The differential pair input with degeneration can be linear enough for any practical application – a subtlety of global negative feedback is that forward loop gain directly reduces the signal at the diff pair input, increasing the bias to signal ratio by decreasing the signal and further “linearizing” the diff pair as a consequence of large negative feedback
you aren’t likely to learn the subject matter of 4+ (including prereqs) college level math/engineering courses in a discussion forum – especially in a audio related forum where many misguided amateurs are repeating really wrong but pervasive "theories" of audio pseudoscience and even some professionals may be disingenuous to support their "no feedback" marketing positions
I think this is impossible. The differential pair will force the input signal and feedback signal have the same timing. None has "delay". Differential pair is 2 transistors tied together in the emitor. Is it possible that 1 base has delay than the other base?
jcx said:
Many of the hifi magazines I read push some of the above. I suppose ideally you need to know how these magazines are funded, and who writes for them. I haven't studied this so it's difficult for me to critically analyse the arguments, and as such I rely on the goodwill of others combined with peer review.
Thank you for your comments.
quote:
Also I understand that it takes a finite time for the feedback loop to operate so corrections occur after the initial event but never before.
Sorry please disregard my comment.
Also I understand that it takes a finite time for the feedback loop to operate so corrections occur after the initial event but never before.
lumanauw said:
Sorry please disregard my comment.
Nobody suggested otherwise.
All that was pointed out was that a feedback loop's behaviour is non-ideal and could introduce problems . That's not pseudoscience.
Hi, Ashdac,
I used to think exactly like you do, that feedback suffer from delay so the feedback signal comes too late. But later I found out that some high-feedback design sounds not so good because of other cause, that is how the feedback is implemented in the cct. How you design the whole cct. The feedback itself is a GOOD thing.
I'm not as clever as JCX is, but I found out (from experiments, not from math calculations) that many things are misunderstood about feedback. 😀
I used to think exactly like you do, that feedback suffer from delay so the feedback signal comes too late. But later I found out that some high-feedback design sounds not so good because of other cause, that is how the feedback is implemented in the cct. How you design the whole cct. The feedback itself is a GOOD thing.
I'm not as clever as JCX is, but I found out (from experiments, not from math calculations) that many things are misunderstood about feedback. 😀
Actually I must apologise for being very negative about feedback. It's an important tool, it's just caused me alot of pain due to being misused in musical instrument electronics. ( Usually being used when there is a danger of clipping, in eq stages and the like, or else to disguise some really cheap electronics and give a reasonable thd rating )
This really isn't relevant in good hifi design.
This really isn't relevant in good hifi design.
wildswan said:
Greetings from Norfolk (UK)
Not true 🙁. Amplifiers with or without feedback can have the same 'operating' gain. It is just that an amplifier with feedback has a greater than required gain (open loop) which is then reduced by the application of feedback. The input signal amplitude does not need to change to produce the same output signal with either type of amplifier designed to the same gain specification.
Richard
lumanauw said:
Hi,
I think that the main problem of "delay" lies in the delay of amplifier reaction to input signal change.None transistors are perfect ; also there are many parasitic capacitances and other things wich makes amplifier speed limited.
However , a good amplifier with fast circuit should react fast enough that distortion is low in the audio range.
I would not like to agree with widely stated propositions about transient intermodulation distortion(TIMD).
For this type of distortion to happen , we need very high input slew rates , so that the amplifier's feedback does not work.
Normally , music content should not have such a high slew rates.
Most of music is a composition of sine waves , ranging from 30 to 20kHz. Recording / playback systems can not process much higher frequency signals(and slew rates as well).And the speed at which the air pressure can change is limited by force that makes the sound.
Regards,
Lukas.
I agree with Buzukas about TIMD. A poor sounding circuit will sound poor at very low listening levels as well as extremes. The reason for NFB killing the sound is not normally TIMD, although it could be in theory.
Note that the spectral content of the feedback signal will be wider in frequency than the original signal due to the non-linearity of the amp. So although the highest music frequency may be 20kHz the highest feedback signal frequency may be much higher. Of course, you cannot hear anything much above 20kHz so the impact of this or otherwise needs careful thought.
Note that the spectral content of the feedback signal will be wider in frequency than the original signal due to the non-linearity of the amp. So although the highest music frequency may be 20kHz the highest feedback signal frequency may be much higher. Of course, you cannot hear anything much above 20kHz so the impact of this or otherwise needs careful thought.
Bazukaz said:
Input signal not, but return signal applied to inverting input, and therefore 'after' forward path can have higher slew. The class B distortion is wide spectrum from its nature. Also there may be some radio picked up signal or transient ringing of picked up 'supply rubbish' at frequency where amplifier has very low or no feedback. These signals cannot be corrected by overal feedback loop but they ARE amplified by early stages and can overdrive them.
regards
Adam
Debates about feedback are nearly as fruitless as whether choice of speaker and interconnect wire is important. Nonetheless, for any who may be interested in the last week or so Rod Elliot has added a new article to his website on (http://sound.westhost.com) on the topic.
I'm not offereing an opinion about it other than it's a typical Rod article.
I'm not offereing an opinion about it other than it's a typical Rod article.
Yes the feedback debate keeps going around in circles. Funny though, 'cause that's how feedback works. I think the discussion of tough topics is fun and useful for people trying to learn.
Thanks for the link. I like the cute diagram of the toilet cistern.😀
I've installed a few of these in my time and I'd say it is a rather good illustration of how they work.
Regarding feedback, though, Rod makes a few claims that simply don't stand up to engineering scrutiny and seem to be a rehash of common folklore. The article does cover a lot of stuff and it is mostly quite informative, but a few statements drive a stake through the articles' heart as far as I'm concerned.
"Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied."
Whether one considers the signal within the feedback loop to be circulating iteratively or not is not of much consequence, but to imply that applying feedback to a system cannot create new harmonics is quite wrong. Sorry.
The pinnacle occurs later on when he goes on to say:
"There is no possibility that the use of feedback will make a good amplifier sound bad."
Good grief! That is just naive. I have read many observations that say just that and I have witnessed it myself.
"Read any articles about distortion you may come across (including this!) with care"
Indeed.
Thanks for the link. I like the cute diagram of the toilet cistern.😀
I've installed a few of these in my time and I'd say it is a rather good illustration of how they work. Regarding feedback, though, Rod makes a few claims that simply don't stand up to engineering scrutiny and seem to be a rehash of common folklore. The article does cover a lot of stuff and it is mostly quite informative, but a few statements drive a stake through the articles' heart as far as I'm concerned.
"Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied."
Whether one considers the signal within the feedback loop to be circulating iteratively or not is not of much consequence, but to imply that applying feedback to a system cannot create new harmonics is quite wrong. Sorry.
The pinnacle occurs later on when he goes on to say:
"There is no possibility that the use of feedback will make a good amplifier sound bad."
Good grief! That is just naive. I have read many observations that say just that and I have witnessed it myself.
"Read any articles about distortion you may come across (including this!) with care"
Indeed.
Rod’s article isn’t too bad, but he doesn’t quite get it right
One error is his over statement that negative feedback never increases the level of a higher order distortion harmonic component, his use of exponential diode characteristic distortion sources with significant harmonics of every order in the power series expansion masks the real situation which can be seen in my sim
http://www.diyaudio.com/forums/showthread.php?postid=912012#post912012
Baxandall and Cherry have shown the distortion harmonic series math of negative feedback in audio amplifiers, I’m sure ½ century old analysis of tube based negative feedback amplifier distortion could be found
Saying cathode/source/emitter degeneration is not local negative feedback is simply a inexcusable error – degeneration as a circuit technique may have preceded Black’s negative feedback patent and subsequent development of negative feedback theory, but that theory fully explains the properties of degeneration as a negative feedback system
Other annoyances (to me at least) include any reference to Cheever’s thesis – which I consider an embarrassment to the issuing institution.
One error is his over statement that negative feedback never increases the level of a higher order distortion harmonic component, his use of exponential diode characteristic distortion sources with significant harmonics of every order in the power series expansion masks the real situation which can be seen in my sim
http://www.diyaudio.com/forums/showthread.php?postid=912012#post912012
Baxandall and Cherry have shown the distortion harmonic series math of negative feedback in audio amplifiers, I’m sure ½ century old analysis of tube based negative feedback amplifier distortion could be found
Saying cathode/source/emitter degeneration is not local negative feedback is simply a inexcusable error – degeneration as a circuit technique may have preceded Black’s negative feedback patent and subsequent development of negative feedback theory, but that theory fully explains the properties of degeneration as a negative feedback system
Other annoyances (to me at least) include any reference to Cheever’s thesis – which I consider an embarrassment to the issuing institution.
My toilet seems to be oscillating, they must have better ones down under. Sorry Mr Elliot, i flushed it.
Is JCX here? He could be angry with what I wrote down here 😀
There are things that people not realized when thinking about feedback.
First : In a feedback amplifier there is the main cct and feedback itself. When arguing about feedback, people usually think (not realized) that the cct (that is subjected to feedback) is a perfect cct. That it will take forward gain and feedback flawlessly.
If they (not realized) thinking that the cct is already "perfect" then the one who makes the bad sounding must be the feedback.
In reality, a real cct is not acting like this ideal condition. Feedback will work at any condition, wheter the cct is smooth or have many bottlenecks, feedback will work it's duty.
In an amplifier, signal goes to base of transistorA, then from it's emitor/collector it goes to base of transistorB, then from transistorB's emitor/collector it goes to base of transistorC, and so on.
Sometimes there are "bottleneck", example : signal after transistorB's collector is not having the properties (impedance) of what transistorC's base is wanting. The more transistorB's forcing, the more "bottleneck" happening at transistorC's base. TransistorB cannot force transistorC's base to accept B's output if the demands of transitorC's base is not fullfilled.
A good example is a stage after VBE multiplier (this VBEmultiplier lies in VAS' collector). From here to the next stage there should be base stoppers. This is just one example, I just want to show that it is not the feedback to blame for poor sonics, but there are more important factor. The cct have to be as "blameless" as possible, for the audio signal to travel, before blaming feedback for anything.
I think that TIM is exactly this "bottleneck". The signal will stuck at one point because the previous stage cannot make the next stage accept what the previous is forcing.
About frequencies. We looked so much graphs that has the horizontal axis of f (frequency), like drawing A. In drawing A we see, from left to right, frequency is rising, from 10hz, 1khz, 100khz, 1Mhz, etc. Looking too much of this graph, again, people (not realized) think that frequencies above 20khz is located in far area in right direction. 10Mhz is located quite remote from 20-20khz so it will not matter, because it is very distanced from our hearing range.
But actually, when we see drawingB (this is osciloscope), if we have 1khz sinusoidal, where is 10Mhz signal? It is NOT located away from the 1khz, the 10Mhz signal lies WITHIN the 1khz sinusoidal. So I think behavior for frequencies above 20khz (and below 20hz) will certainly affect the spectrum of 20hz-20khz, because those frequencies (>20khz and <20hz) are not located distanced away from 20-20khz range, they are right there at the same place, right between this range. They can intermodulate each other, so it is not only the 20-20khz that has to be perfect, but we have to look also of <20hz and >20khz to make good sounding amp 😀
There are things that people not realized when thinking about feedback.
First : In a feedback amplifier there is the main cct and feedback itself. When arguing about feedback, people usually think (not realized) that the cct (that is subjected to feedback) is a perfect cct. That it will take forward gain and feedback flawlessly.
If they (not realized) thinking that the cct is already "perfect" then the one who makes the bad sounding must be the feedback.
In reality, a real cct is not acting like this ideal condition. Feedback will work at any condition, wheter the cct is smooth or have many bottlenecks, feedback will work it's duty.
In an amplifier, signal goes to base of transistorA, then from it's emitor/collector it goes to base of transistorB, then from transistorB's emitor/collector it goes to base of transistorC, and so on.
Sometimes there are "bottleneck", example : signal after transistorB's collector is not having the properties (impedance) of what transistorC's base is wanting. The more transistorB's forcing, the more "bottleneck" happening at transistorC's base. TransistorB cannot force transistorC's base to accept B's output if the demands of transitorC's base is not fullfilled.
A good example is a stage after VBE multiplier (this VBEmultiplier lies in VAS' collector). From here to the next stage there should be base stoppers. This is just one example, I just want to show that it is not the feedback to blame for poor sonics, but there are more important factor. The cct have to be as "blameless" as possible, for the audio signal to travel, before blaming feedback for anything.
I think that TIM is exactly this "bottleneck". The signal will stuck at one point because the previous stage cannot make the next stage accept what the previous is forcing.
About frequencies. We looked so much graphs that has the horizontal axis of f (frequency), like drawing A. In drawing A we see, from left to right, frequency is rising, from 10hz, 1khz, 100khz, 1Mhz, etc. Looking too much of this graph, again, people (not realized) think that frequencies above 20khz is located in far area in right direction. 10Mhz is located quite remote from 20-20khz so it will not matter, because it is very distanced from our hearing range.
But actually, when we see drawingB (this is osciloscope), if we have 1khz sinusoidal, where is 10Mhz signal? It is NOT located away from the 1khz, the 10Mhz signal lies WITHIN the 1khz sinusoidal. So I think behavior for frequencies above 20khz (and below 20hz) will certainly affect the spectrum of 20hz-20khz, because those frequencies (>20khz and <20hz) are not located distanced away from 20-20khz range, they are right there at the same place, right between this range. They can intermodulate each other, so it is not only the 20-20khz that has to be perfect, but we have to look also of <20hz and >20khz to make good sounding amp 😀
Attachments
David,
your post makes think of the parallel with operations research, finding the bottleneck in economic processes is the best guarantee for the best process flow, the feedback regulation system optimises it. Feedback in a system with one or several bottlenecks remaining lowers the stagnation but inherently lowers the flow rate.
your post makes think of the parallel with operations research, finding the bottleneck in economic processes is the best guarantee for the best process flow, the feedback regulation system optimises it. Feedback in a system with one or several bottlenecks remaining lowers the stagnation but inherently lowers the flow rate.
I'm glad someone else has some concerns about Cheever's conclusions. I read this a few nights ago and although I sort of liked some of his methods I think the thesis is untennable...he ignores too many alternatives to the one he sets out to prove. But hey, it just another thesis to get a degree and it was good enough for that.
lumanauw:
You're on the right lines. Feedback can be shown to work perfectly in the rarified environment of mathmatics. But the actual circuit is not so perfect. Feedback ends up doing what you tell it to: so garbage in, garbage out. The spectral content of the amp's signal path is much richer when the loop is closed.
jcx said:
What aspects of it? The higher masking of lower order distortion components would appear to be firmly established in the scientific literature. Is it where Cheever goes from there?
Traderbam said:
"Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied."
Whether one considers the signal within the feedback loop to be circulating iteratively or not is not of much consequence, but to imply that applying feedback to a system cannot create new harmonics is quite wrong. Sorry.
Rod Elliot said:
Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied.
Feel free to re-read the last statement as many times as you need to. This is a claim that has been made on numerous occasions, and it is simply false.
Rod seems to agree.
"Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied."
Whether one considers the signal within the feedback loop to be circulating iteratively or not is not of much consequence, but to imply that applying feedback to a system cannot create new harmonics is quite wrong. Sorry.
Rod Elliot said:
Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied.
Feel free to re-read the last statement as many times as you need to. This is a claim that has been made on numerous occasions, and it is simply false.
Rod seems to agree.
- Status
- Not open for further replies.