Of course, there are always special cases...
Like I was saying: of course, there are always special cases...
For example, when too much negative feedback is used and wrapped back into a diff amp at an early stage, there is the potential that the feedback signal will suck all the juice out of the ["constant"] current source biasing that diff amp. In some cases, this can cause something nasty to happen in the signal path. As I recall, this is considered an advanced form of slewing, and it is revealed by measuring transient intermodulation distorion using a sine wave riding synchronized on top of a square wave.
This is one of the many little design details that may have caused some early solid state amps not so sound as good as they could have with less feedback. As engineers, we learn diff amp as small signal problems and as biasing problems -- but often with little care about how the two problems can interact. On the other hand, what do you say to a smartypants know-it-all expert who refuses to believe that electrons can get back there fast enough to have a beneficial effect on the signal?
I'm a million years old, and I've heard a lot of arguments in the audio world. Such passions! At the end of the day (or at the end of one of the days, eventually), we realize a number of important things:
(1) Some people hear what they want to hear,
and they may delve into buffoonery when they take an aggressive technical position based on their imaginary evidence.
(2) Some people can hear real differences, but for whatever reasons, attribute these differences to ridiculous theories (often overlooking the fact their test system is a piece of, or broken, etc.); we need to remember there are points taken off for wrong answers.
(3) Certain technical people dive into buffoonery when they assert the impossibility of sound qualities
reported by truly excellent and perhaps even gifted listeners.
(4) Many audio nuts (like I am) sometimes enjoy taking sides in a way that polarizes forums
including the once popular neighborhood stereo equipment store. Thank God for the Internet!
(5) There are seven more items on this list -- unless I am entirely mistaken in that belief.
Do you know what I'm saying?
Cheers, Dude. Welcome to the Free World. And be gentle.
Like I was saying: of course, there are always special cases...
For example, when too much negative feedback is used and wrapped back into a diff amp at an early stage, there is the potential that the feedback signal will suck all the juice out of the ["constant"] current source biasing that diff amp. In some cases, this can cause something nasty to happen in the signal path. As I recall, this is considered an advanced form of slewing, and it is revealed by measuring transient intermodulation distorion using a sine wave riding synchronized on top of a square wave.
This is one of the many little design details that may have caused some early solid state amps not so sound as good as they could have with less feedback. As engineers, we learn diff amp as small signal problems and as biasing problems -- but often with little care about how the two problems can interact. On the other hand, what do you say to a smartypants know-it-all expert who refuses to believe that electrons can get back there fast enough to have a beneficial effect on the signal?
I'm a million years old, and I've heard a lot of arguments in the audio world. Such passions! At the end of the day (or at the end of one of the days, eventually), we realize a number of important things:
(1) Some people hear what they want to hear,
and they may delve into buffoonery when they take an aggressive technical position based on their imaginary evidence. (2) Some people can hear real differences, but for whatever reasons, attribute these differences to ridiculous theories (often overlooking the fact their test system is a piece of, or broken, etc.); we need to remember there are points taken off for wrong answers.
(3) Certain technical people dive into buffoonery when they assert the impossibility of sound qualities
reported by truly excellent and perhaps even gifted listeners. (4) Many audio nuts (like I am) sometimes enjoy taking sides in a way that polarizes forums
including the once popular neighborhood stereo equipment store. Thank God for the Internet! (5) There are seven more items on this list -- unless I am entirely mistaken in that belief.
Do you know what I'm saying?Cheers, Dude. Welcome to the Free World. And be gentle.
Re: Of course, there are always special cases...
Konnichiwa,
I would not know about that, considering that "Electrons" are imaginary and have no proven reality doews it matter what these inaginary particles do? BTW, I agree that there is something, but the "Electron Particle" model is by far to poorely correlated with reality to be taken as the gospel (eg. IF the electron where a traditional newtonian particle as the model aledges Heisen would have never become famous).
Note, that includes those who want to hear "NO DIFFERENCE" and who then fo on to endlessly wave "no result" DB Tests around which are easily dismissed purely on statistical grounds, never mind the validity of the actual test-setup.
The problem is that people want certainties. They (the pee-pull) rather live with false certainties (as in 'we know what is going on') than to face the real uncertainty (as in 'we have no footbaling clue why it does that') and are always ready to laugh at and ridicule the fool who sits on his hill with his cap and his bells, completely failing to note that the real fools is they....
Sayonara
Konnichiwa,
JimRodgers said:
I would not know about that, considering that "Electrons" are imaginary and have no proven reality doews it matter what these inaginary particles do? BTW, I agree that there is something, but the "Electron Particle" model is by far to poorely correlated with reality to be taken as the gospel (eg. IF the electron where a traditional newtonian particle as the model aledges Heisen would have never become famous).
JimRodgers said:(1) Some people hear what they want to hear,
Note, that includes those who want to hear "NO DIFFERENCE" and who then fo on to endlessly wave "no result" DB Tests around which are easily dismissed purely on statistical grounds, never mind the validity of the actual test-setup.
The problem is that people want certainties. They (the pee-pull) rather live with false certainties (as in 'we know what is going on') than to face the real uncertainty (as in 'we have no footbaling clue why it does that') and are always ready to laugh at and ridicule the fool who sits on his hill with his cap and his bells, completely failing to note that the real fools is they....
Sayonara
Re: Of course, there are always special cases...
Hi Jim,
I gather you refer to the fact that feedback can't work because the electrons come back too late to prevent the damage?
I must say you make a convincing case, and I really would like to believe you, but there I have a problem. You see, all those millions of years I have been building equipment with feedback, it worked everytime. All the time. Exactly as predicted. Exactly as calculated. So why does reality refuse to follow your theory, or am I missing something?
Edit: Jim, if I misread you, that is, if you DO believe that electrons CAN get there fast enough, scratch my above post, in that case you, I and reality DO agree. Semantics...
Jan Didden
JimRodgers said:
Hi Jim,
I gather you refer to the fact that feedback can't work because the electrons come back too late to prevent the damage?
I must say you make a convincing case, and I really would like to believe you, but there I have a problem. You see, all those millions of years I have been building equipment with feedback, it worked everytime. All the time. Exactly as predicted. Exactly as calculated. So why does reality refuse to follow your theory, or am I missing something?
Edit: Jim, if I misread you, that is, if you DO believe that electrons CAN get there fast enough, scratch my above post, in that case you, I and reality DO agree. Semantics...
Jan Didden
What debate? Did I start a thread? This is cool.
Thanks, SY, for running to my side. I'm new here, so I appreciate the third dimension of your remarks.
Satonara, I welcome your remarks on my other points as well. like numbers 4, 8, and 11.
And keep in mind that while electrons are a theoretical abstraction, so also is electricity: the theoretical flow of theoretical charge in a over a supposed period of so-called time and in an approximate place. I remember those days -- in the dorm at Emory: what if the whole universe is just a spec under the thumbnail of some giant... Wait a minute! Electricity is the flow of positive charges!!! I KNOW there is no such thing as THAT in a wire? OR what do I REALLY know?
Seriously, I did not mean to trip over any previous positions you may have expressed about electrons or positive feedback, etc. I was just trying to make the point feedback theory very often gets repeatable results we can hear in the sound machines we build using that theory. Good results and bad results. Actually, I thought I was posting a response to welcome a new guy who was as odds with another forum for expressing himself about the theory feedback in amplifer circuit design. So I was just trying to be nice ot him.
By the way, I just saw a certain lodspeaker website was selling audiophile duplex power receptacles for $150! For electrons! Any takers? 
Just kidding. It nice to have such a rich place to bring up Audio topics. I have many questions to ask, and I hope I can offer some information back from time to time.
Thanks, SY, for running to my side. I'm new here, so I appreciate the third dimension of your remarks.
Satonara, I welcome your remarks on my other points as well. like numbers 4, 8, and 11.
And keep in mind that while electrons are a theoretical abstraction, so also is electricity: the theoretical flow of theoretical charge in a over a supposed period of so-called time and in an approximate place. I remember those days -- in the dorm at Emory: what if the whole universe is just a spec under the thumbnail of some giant...
Wait a minute! Electricity is the flow of positive charges!!! I KNOW there is no such thing as THAT in a wire? OR what do I REALLY know?Seriously, I did not mean to trip over any previous positions you may have expressed about electrons or positive feedback, etc. I was just trying to make the point feedback theory very often gets repeatable results we can hear in the sound machines we build using that theory. Good results and bad results. Actually, I thought I was posting a response to welcome a new guy who was as odds with another forum for expressing himself about the theory feedback in amplifer circuit design. So I was just trying to be nice ot him.
By the way, I just saw a certain lodspeaker website was selling audiophile duplex power receptacles for $150! For electrons!
Any takers? Just kidding. It nice to have such a rich place to bring up Audio topics. I have many questions to ask, and I hope I can offer some information back from time to time.
Re: Re: Of course, there are always special cases...
That's right, Jan, I DO believe feedback works.
It's really a matter of bandwidth, I believe. Maybe it's really true that some "damage" must occur if the feedback is created that that can be helpful in preventing further "damage." But this delay is irrelevant in our frame of reference. If I'm not mistaken, a detailed treatment of this issue involves the Theory of Relativity.
Perhaps it is even more useful to make the approximation that charge carriers move close enough to the speed of light most of the time that this is VIRTUALLY instantaneous as far as most human audio engineers and listeners are concerned. Given, then, that the technical problem involves no delays due to charge carrier transit times, we are dealing with a set of simultaneous equations in the context of control theory. The solution does not imply anything about whether the damage ever really occurred.
However, there also is the concept of position error (and/or velocity error and/or acceleration error, etc.). This may seem to address the idea that some damage is necessary in order to correct it. Personally, I’m not convinced that this is what it means. Then you have the issue of zero order, first order, second order control systems where integrators may/may not be present in the feedback loop.
One must be careful not to forget that there are BIG delays, however, associated with reactive elements in the system. Thus, negative feedback can become positive feedback at certain frequencies, etc. Also, it is interesting to solve these equations when discrete delays are present in the system. Formulated with Laplace variables, as I recall, a discrete delay of time=T is described by
f(s) = exp(-sT)
Perhaps this is useful for modeling processing delay in computer controlled systems. Anybody done this? I wonder: what are the stability criteria of a control system (feedback network) where such a delay is in the feedback path? Any takers? (How about you carputer guys?)
Bringing this full circle (in infinite-dimensional Hilbert Space, perhaps!), couldn't charge carrier transit delays be treated mathematically as discrete delays? I assume the extra terms drop out as too tiny to calculate.
Signal and System Theory stands on the shoulders of wonderful mathematicians like Laplace and Fourier who lived long before Einstein, Bohr, or Heisenberg got us wondering about everything.
But if it were not for the occasional reminder that electrons are somewhat nebulous in many ways, we might not want to stay sharp enough to remember where our canned engineering formulas came from.
janneman said:
That's right, Jan, I DO believe feedback works.
It's really a matter of bandwidth, I believe. Maybe it's really true that some "damage" must occur if the feedback is created that that can be helpful in preventing further "damage." But this delay is irrelevant in our frame of reference. If I'm not mistaken, a detailed treatment of this issue involves the Theory of Relativity.
Perhaps it is even more useful to make the approximation that charge carriers move close enough to the speed of light most of the time that this is VIRTUALLY instantaneous as far as most human audio engineers and listeners are concerned. Given, then, that the technical problem involves no delays due to charge carrier transit times, we are dealing with a set of simultaneous equations in the context of control theory. The solution does not imply anything about whether the damage ever really occurred.
However, there also is the concept of position error (and/or velocity error and/or acceleration error, etc.). This may seem to address the idea that some damage is necessary in order to correct it. Personally, I’m not convinced that this is what it means. Then you have the issue of zero order, first order, second order control systems where integrators may/may not be present in the feedback loop.
One must be careful not to forget that there are BIG delays, however, associated with reactive elements in the system. Thus, negative feedback can become positive feedback at certain frequencies, etc. Also, it is interesting to solve these equations when discrete delays are present in the system. Formulated with Laplace variables, as I recall, a discrete delay of time=T is described by
f(s) = exp(-sT)
Perhaps this is useful for modeling processing delay in computer controlled systems. Anybody done this? I wonder: what are the stability criteria of a control system (feedback network) where such a delay is in the feedback path? Any takers? (How about you carputer guys?)
Bringing this full circle (in infinite-dimensional Hilbert Space, perhaps!), couldn't charge carrier transit delays be treated mathematically as discrete delays? I assume the extra terms drop out as too tiny to calculate.
Signal and System Theory stands on the shoulders of wonderful mathematicians like Laplace and Fourier who lived long before Einstein, Bohr, or Heisenberg got us wondering about everything.
But if it were not for the occasional reminder that electrons are somewhat nebulous in many ways,
we might not want to stay sharp enough to remember where our canned engineering formulas came from.
Jim,
Never had to incorporate any fixed time delays into control loop design... LaPlace style at least. They can be simply modeled in BODE analysis (NOT perfectly by any means). Flat gain and a curious phase function.
I generally try to keep control loops smaller than a state or country.
Got me thinking though... now I want to dig out some books. I remember vaguely that time delay is on some level related to backlash and freeplay (equivalent or reciprocal in s domain?)
As far as numerical loops though, and again these are systems running much slower than the computer; I have found that the math gets you close, but compensation coefficients still had to be tweaked in by hand. Done by careful measurement and trial and error over all extremes of operation. Before I get the "ear" people all riled up; let me say that our inability to achieve perfect results through math was a function of our inability to precisely model the system (usually electro-mech or electro therm) under control... or that we had deliberately simplified the model to ease the math knowing that we would tweak it in anyway.
In the case of purely electronic systems, I have had the same luck as Jan... math works.
Hmmmm....... your turn,
Oh, BTW, the site is really about violent writers & authors that are interested in audio & electronics. Read your posts at least 400 times before pushing button... the vultures are always nearby.
Never had to incorporate any fixed time delays into control loop design... LaPlace style at least. They can be simply modeled in BODE analysis (NOT perfectly by any means). Flat gain and a curious phase function.
I generally try to keep control loops smaller than a state or country.
Got me thinking though... now I want to dig out some books. I remember vaguely that time delay is on some level related to backlash and freeplay (equivalent or reciprocal in s domain?)
As far as numerical loops though, and again these are systems running much slower than the computer; I have found that the math gets you close, but compensation coefficients still had to be tweaked in by hand. Done by careful measurement and trial and error over all extremes of operation. Before I get the "ear" people all riled up; let me say that our inability to achieve perfect results through math was a function of our inability to precisely model the system (usually electro-mech or electro therm) under control... or that we had deliberately simplified the model to ease the math knowing that we would tweak it in anyway.
In the case of purely electronic systems, I have had the same luck as Jan... math works.
Hmmmm....... your turn,
Oh, BTW, the site is really about violent writers & authors that are interested in audio & electronics. Read your posts at least 400 times before pushing button... the vultures are always nearby.
Re: Re: Re: Of course, there are always special cases...
Jim,
Thank ohm! Another guy who has his facts straight! I fully agree with you.
I think part of the feedback-comes-too-late confusion comes from failing to understand two separate things:
- the fact that the electrons or charge or whatever needs a finite time to travel trough the wires to get back to the input. That is as you say for practical purposed neglectable (although it is not zero of course);
- the phase shift within the (forward and) feedback loop due to reactive elements, which eventually turns neg feedback into pos feedback with all the known problems. That is sometimes also seen as delay but it isn't.
A good example is the phaseshift between current into a cap and the voltage across it. It seems superficially that the voltage is delayed w.r.t. the current, but it isn't! At the very instance that the current starts to flow, the voltage starts to rise. There is NO 'dead-time' between start of the current and start of the voltage (other then the propagation effects mentioned above). So, in this sense, there is NO delay in feedback, only phase shift. Not always easy to keep apart.
Neg feedback works by subtracting the input from the feedback signal so that the (inverted) error remains that is then amplified to correct the output. Because of the phase shift, the subtraction gets less and less complete with rising freq so the feedback gets less and less effective. If the phase shift is large enough, the subtraction changes into addition and the output signal is sustained without an input: we've got ourselves an oscillator.
Jan Didden
JimRodgers said:
Jim,
Thank ohm! Another guy who has his facts straight! I fully agree with you.
I think part of the feedback-comes-too-late confusion comes from failing to understand two separate things:
- the fact that the electrons or charge or whatever needs a finite time to travel trough the wires to get back to the input. That is as you say for practical purposed neglectable (although it is not zero of course);
- the phase shift within the (forward and) feedback loop due to reactive elements, which eventually turns neg feedback into pos feedback with all the known problems. That is sometimes also seen as delay but it isn't.
A good example is the phaseshift between current into a cap and the voltage across it. It seems superficially that the voltage is delayed w.r.t. the current, but it isn't! At the very instance that the current starts to flow, the voltage starts to rise. There is NO 'dead-time' between start of the current and start of the voltage (other then the propagation effects mentioned above). So, in this sense, there is NO delay in feedback, only phase shift. Not always easy to keep apart.
Neg feedback works by subtracting the input from the feedback signal so that the (inverted) error remains that is then amplified to correct the output. Because of the phase shift, the subtraction gets less and less complete with rising freq so the feedback gets less and less effective. If the phase shift is large enough, the subtraction changes into addition and the output signal is sustained without an input: we've got ourselves an oscillator.
Jan Didden
Re: Re: Re: Re: Of course, there are always special cases...
Konnichiwa,
And circuits with negative feedback invariably display an "error band" (which is in other words the resut of the neccesity for an error to be present first before feedback operates), meaning a previously straight line signal becomes rather fuzzy. The error band becomes the wider the slower the circuit is.
As large open loop gains and wide open loop bandwidth are generally mutually exclusive you allways have your rocks between a hard place and a hard wall.
What has also been only very incompletely researched is the subjective effect of this "fuzziness" with music.
Another way of viewing this is to say that a Feedback amplifier circuit is allways a little wrong by about the same absolute amount in an entierly unpredictable fashion while a non-feedback Amplifier is progressively less wrong with falling signal and is wrong in an entierly predictable fashion.
Yet another view of feedback is as distortion multiplier, in other words distortion energy as such is not actually reduced, merely high levels of low order distrotion are converted into low levels of high order distortion which is hoped will be below the error band and other noise.
On another note, electrons do not travel through wire or semiconductors at lightspeed (or even at an appreciable fraction of it), what travels are EM Waves, electrons as such travel through solids at speeds in the region of a few meters per second....
Sayonara
PS, I am not saying "feedback does not work", on the contrary, merely that it's proponents rarely explain the sideeffects and drawbacks in detail.
Konnichiwa,
janneman said:
And circuits with negative feedback invariably display an "error band" (which is in other words the resut of the neccesity for an error to be present first before feedback operates), meaning a previously straight line signal becomes rather fuzzy. The error band becomes the wider the slower the circuit is.
As large open loop gains and wide open loop bandwidth are generally mutually exclusive you allways have your rocks between a hard place and a hard wall.
What has also been only very incompletely researched is the subjective effect of this "fuzziness" with music.
Another way of viewing this is to say that a Feedback amplifier circuit is allways a little wrong by about the same absolute amount in an entierly unpredictable fashion while a non-feedback Amplifier is progressively less wrong with falling signal and is wrong in an entierly predictable fashion.
Yet another view of feedback is as distortion multiplier, in other words distortion energy as such is not actually reduced, merely high levels of low order distrotion are converted into low levels of high order distortion which is hoped will be below the error band and other noise.
On another note, electrons do not travel through wire or semiconductors at lightspeed (or even at an appreciable fraction of it), what travels are EM Waves, electrons as such travel through solids at speeds in the region of a few meters per second....
Sayonara
PS, I am not saying "feedback does not work", on the contrary, merely that it's proponents rarely explain the sideeffects and drawbacks in detail.
Re: Re: Of course, there are always special cases...
What this is exactly my point and my experience too. I have at least 10 years designing and building audio amplifiers; and all the time this has worked perfectly. With negative feedback is possible with relative cheap components and basic topologies working altogether to build an accurate and stable Hi-Fi audio amplifier; with flat frequency response, ultra low distortion (<0.1%), high damping factor and excelent signal to noise ratio (90db to 120db typical).
Negative feedback has been proven and verified even the death to be very successful; and has been used as a essential tool of design by engineers by decades.
The vendors of snake oil (as the guy in SteveHoffman for example) that promotes the lie that negative feedback is not good cause a big damage; and, in the end, the honest engineer is who is harmed by this.
janneman said:
What this is exactly my point and my experience too. I have at least 10 years designing and building audio amplifiers; and all the time this has worked perfectly. With negative feedback is possible with relative cheap components and basic topologies working altogether to build an accurate and stable Hi-Fi audio amplifier; with flat frequency response, ultra low distortion (<0.1%), high damping factor and excelent signal to noise ratio (90db to 120db typical).
Negative feedback has been proven and verified even the death to be very successful; and has been used as a essential tool of design by engineers by decades.
The vendors of snake oil (as the guy in SteveHoffman for example) that promotes the lie that negative feedback is not good cause a big damage; and, in the end, the honest engineer is who is harmed by this.
Re: Re: Re: Re: Re: Of course, there are always special cases...
Thorsten,
Yes indeed there is always a cause before the effect! But I don't see the fuzziness? There is an error band, increasing with freq, or as you say, increasing the slower the circuit is (because of the increased phase shift, NOT because of some mythical 'delay').
But that error band is fully defined, theoretically, and can be measured in practise. It is not fuzzy, it is a continuous, smooth function showing the delta between what we would like and what we have.
The only case I can understand some 'raggedness' is when the system is on the verge of instability and the delta function becomes oscillatory, swinging around the 'average' value so to say.
Jan Didden
Kuei Yang Wang said:
Thorsten,
Yes indeed there is always a cause before the effect! But I don't see the fuzziness? There is an error band, increasing with freq, or as you say, increasing the slower the circuit is (because of the increased phase shift, NOT because of some mythical 'delay').
But that error band is fully defined, theoretically, and can be measured in practise. It is not fuzzy, it is a continuous, smooth function showing the delta between what we would like and what we have.
The only case I can understand some 'raggedness' is when the system is on the verge of instability and the delta function becomes oscillatory, swinging around the 'average' value so to say.
Jan Didden
You understand, of course, that light is NOT the presence of photons, but rather, the absence of darkons. I'm awaiting the call from Sweden.
Actually, if you stick two electrodes in an acidic solution, the net current actually is due to the motion of positive charges.
How about DC?
In reality, it's useless to talk about electron speed, just as it's useless for me to worry about how long it takes water to get to me from the reservoir/treatment plant when I open the tap in my kitchen.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.