You may not be convinced, but would you accept the opinion of Edward M. Cherry, who I should think would not need further introduction. In the paper quoted below he proceeds to get to the bottom of the circuit analysis of all possible feedback configurations, and concludes:
"Stripped of its input level shifters and complementary first stage, and with the current mirrors and voltage followers replaced by the simplest inverting amplifier (a single common-emitter transistor), the current feedback amplifier reduces to [Fig 6]. A generation ago this circuit was called a voltage-feedback pair. It differs from [Fig 2a] only in that the current flowing into the left-hand side of the feedback network is the input current multiplied by the gain of the first transistor, as distinct from the input current itself. The feedback connection is voltage-sensing at the output and voltage-subtracting at the input, the circuit is an ordinary voltage feedback amplifier. The term ‘current-feedback amplifier’ seems a complete misnomer".
Edward M Cherry, ‘Feedback amplifier configurations’, IEE Proceedings on Circuits, Devices and Systems, Vol 147, No 6, Dec 2000.
I rest my case.
Jan Didden
"Stripped of its input level shifters and complementary first stage, and with the current mirrors and voltage followers replaced by the simplest inverting amplifier (a single common-emitter transistor), the current feedback amplifier reduces to [Fig 6]. A generation ago this circuit was called a voltage-feedback pair. It differs from [Fig 2a] only in that the current flowing into the left-hand side of the feedback network is the input current multiplied by the gain of the first transistor, as distinct from the input current itself. The feedback connection is voltage-sensing at the output and voltage-subtracting at the input, the circuit is an ordinary voltage feedback amplifier. The term ‘current-feedback amplifier’ seems a complete misnomer".
Edward M Cherry, ‘Feedback amplifier configurations’, IEE Proceedings on Circuits, Devices and Systems, Vol 147, No 6, Dec 2000.
I rest my case.
Jan Didden
Remember this?
Current feedback is applied when the feedback signal contains information on the outgoing current.
Voltage feedback is applied when the feedback signal contains information on the outgoing voltage.
The way the feedback signal is represented does not really matter, since the negative input must always be the same kind as the positive (voltage or current). The reason is the same as the reason for trouble subtracting tomatoes and carrots
they are not the same.
Current feedback is applied when the feedback signal contains information on the outgoing current.
Voltage feedback is applied when the feedback signal contains information on the outgoing voltage.
The way the feedback signal is represented does not really matter, since the negative input must always be the same kind as the positive (voltage or current). The reason is the same as the reason for trouble subtracting tomatoes and carrots
they are not the same.Attachments
Yes. I even remember the thread! It's a way to classify the topologies depending on whether they modify the input signal by putting the feedback signal in series or in parallel with the input signal, and whether the return signal is a sample of the output voltage or the output current. Cherry addresses this also in his paper, but he goes much further and dissects the topologies to the bone.
Jan Didden
It's a way to classify the topologies depending on whether they modify the input signal by putting the feedback signal in series or in parallel with the input signal, and whether the return signal is a sample of the output voltage or the output current. Cherry addresses this also in his paper, but he goes much further and dissects the topologies to the bone.Jan Didden
Since zero or infinite impedances do not exist in the real world, I think that 'current feedback' and 'voltage feedback' are theoretical concepts that don't happen in real circuits
Real-world impedances are finite and non-zero, so every feedback mechanism may be succesfully analysed both in terms of voltage or in terms of current
Voltage and current are allways related one to the other in real circuits
Real-world impedances are finite and non-zero, so every feedback mechanism may be succesfully analysed both in terms of voltage or in terms of current
Voltage and current are allways related one to the other in real circuits
Example:
It is VERY EASY to distinguish between voltage-sensing feedback or current-sensing feedback. If you short the amp output, and the feedback signal is zero, that amp has voltage-sensing fb. If you disconnect the load, and the feedback signal is zero, that amp has current-sensing feedback.
I know that V and I go together via Ohms law, but that's not the point here.
Jan Didden
It is VERY EASY to distinguish between voltage-sensing feedback or current-sensing feedback. If you short the amp output, and the feedback signal is zero, that amp has voltage-sensing fb. If you disconnect the load, and the feedback signal is zero, that amp has current-sensing feedback.
I know that V and I go together via Ohms law, but that's not the point here.
Jan Didden
Hi,
Per-Anders,
From a technical POV Dimitri's quote of Pavel is correct as is of course Maestro Pass'.
The whole issue evolves around marketing parlance where high global NFB amplifier have been bashed, often without any insight of what exactly it is that it does.
Hence you'll see such statements as "no feedback" amplifier of brandname xyz.
What they usually imply is "no global feedback" loop and even that is sometimes abused.
In order to avoid global feedback and still have low enough output impedance and acceptable distortion levels you can resort to all kinds of parallel techniques such as local, nested loops, degenerative feedback, distortion cancelation gain blocks...ad nauseam.
As for you question about tubes, fets:
If you consider a triode tube with reasonable linearity you'll see the feedback loop already "built in" due to capacitive coupling between the various elements.
If you consider a pentode/tetrode even a FET as a triode with an added grid you'll have far high distortion and reduced linearity as the extra grid reduces the feedback path between the active elements thanks to less interelectrode coupling.
The side effect for a small signal penthode is much higher gain which, when all is said and done, will be greatly reduced anyway as this gainstage will have to be included in the global feedback loop especially when the output stage is also employing powerpentode/kinkless tetrodes.
Inbetween the extremes there are literally hundreds, if not thousands of variations on a theme thinkable all of which will likely have their own subset of pros and cons.
Moral of the story: there's no such thing as a free lunch...
So, to paraphrase Nelson Pass: Yawn.
Cheers,
Per-Anders,
From a technical POV Dimitri's quote of Pavel is correct as is of course Maestro Pass'.
The whole issue evolves around marketing parlance where high global NFB amplifier have been bashed, often without any insight of what exactly it is that it does.
Hence you'll see such statements as "no feedback" amplifier of brandname xyz.
What they usually imply is "no global feedback" loop and even that is sometimes abused.
In order to avoid global feedback and still have low enough output impedance and acceptable distortion levels you can resort to all kinds of parallel techniques such as local, nested loops, degenerative feedback, distortion cancelation gain blocks...ad nauseam.
As for you question about tubes, fets:
If you consider a triode tube with reasonable linearity you'll see the feedback loop already "built in" due to capacitive coupling between the various elements.
If you consider a pentode/tetrode even a FET as a triode with an added grid you'll have far high distortion and reduced linearity as the extra grid reduces the feedback path between the active elements thanks to less interelectrode coupling.
The side effect for a small signal penthode is much higher gain which, when all is said and done, will be greatly reduced anyway as this gainstage will have to be included in the global feedback loop especially when the output stage is also employing powerpentode/kinkless tetrodes.
Inbetween the extremes there are literally hundreds, if not thousands of variations on a theme thinkable all of which will likely have their own subset of pros and cons.
Moral of the story: there's no such thing as a free lunch...
So, to paraphrase Nelson Pass: Yawn.
Cheers,
Eva said:
You don't have to say that you think it is so, you can safely
claim that there is no pure voltage- or current feedback in
real circuits. However, there are at least two observations
to make.
1) Theoretical models are usually not accurate models of the
real world, but are intended to highlight certain
aspects of real behaviour. We all know physical resistors are
not true resistances and that Ohms law really doesn't apply
to them. Yet we persist in using Ohms law for a lot of
calculations. Why is that? Because Ohms law, although a
simplification of real resistors, highlights a very important
property of physical resistors, the property "to resist". In
most cases this is our primary concern and this property is
the one we are mainly interested in. Other physical aspects
of real resistors enter as ornaments, or necessary evils, that
we sometimes must take into account and sometimes can
ignore.
2) When two competing theoretical models exist and neither
is accurate, there can still make a difference which one we
use. One model can give better predictions in a certain case,
or it can be easier to calculate with. It is true that there are
not pure voltage or current feedback amplifiers. We can write
down a lot of mathematical equations describing the circuit,
and the feedback is gone in a sense, because all we have are
mathematical equations. I think we must consider feedback
as a philosophical concept we humans use to understand
circuits. There is no such concept in the mathematical
equations themselves, only in our interpretation of them.
Hence, whether we should try to analyze an amplifier as
a voltage feedback or a current feedback one is best decided
by trying to understand which of these concepts the designer
intended the circuit to approximate. By doing so, we can
often make certain simplifications in our analysis without
sacrificing too much precision, but also without getting the
full complexity of treating the circuit as a black box and just
write down all the equations we can come up with.
As for the discussion whether there are amplifiers without
feedback, I think Nelson were getting close to an answer in
another thread. Isn't it perhaps better to discuss this in terms
of intention? Is the circuit intentionally designed to have
feedback, or does it just have what is unavoidable and that
some call feedback and others not, like emitter degeneration?
This is getting a bit philosphical, but I think the error made
in most of these discussions and which tend to make them
endless is that feedback is treated as a physical or mathematical
phenomenon, while I think it is more appropriate to say that
feedback is just a model we use to understand certain physical
systems in a better way. It is about manipulating the
mathematical equations into a certain form that makes it
possible for us to force our feedback model onto them and
understand them in terms of feedback, but
the feedback isn't really there in the equations themselves.
Frank, I'm not sure I go along with your capacitance = feedback conceptual triode. After all, the gain through the passband is more a function of mu, rp, and gm (pick any two). The capacitance does become a feedback issue at very high frequencies (e.g., Miller effect), but still, the basic gain structure is a function of the physical configuration of the elements, the fundamental constants of the tube, not internal feedback.
Jan, you still are one behind. Don't start playing dirty until you've scored at least once.
Jan, you still are one behind. Don't start playing dirty until you've scored at least once.
janneman said:
Hi Jan,
That is too easy
Whatever it is called is a moot point. What is important is that the so-called “current mode op-amps” and “voltage mode op-amps” behave basically different. The inverting input node (the node where the feedback signal is acting on) is sensing current, not voltage. The output of the amp is either a current times -I_in or voltage times -I_in. Consider the inverting “current mode op-amp” topology with the amp as a black box (leaving out the non-inverting input because it is grounded) …I don’t believe those “current mode” op-amps were developed for just marketing purposes but because there was a need for very fast op-amps in the video arena where impedance is usually low.
Whether or not the nomer “current mode” is correct is not that important, it is more important how these amps behave. It is at least a handy way to distinguish between topologies IMHO.
Cheers
Hi,
Yes and no, of course.
Yet the distance/spacing between the elements determines the animal, hence also the interelectrode capacitances and therefore the f of the internal feedback mechanism.
This is mostly only relevant for HF unless we take very high mu triodes where the Miller effect will come into play as it will multiply this capacitance and form a virtual capacitive filter element.
(I'm grossly simplifying.)
So, it's that chicken/egg enigma if you prefer and entirely dependent on whether want to view this as an intrinsic FB mechanism or not.
To honour the great Harold S.J. Black_ there's an idea for an avatar_ I do.
Cheers,
Yes and no, of course.
Yet the distance/spacing between the elements determines the animal, hence also the interelectrode capacitances and therefore the f of the internal feedback mechanism.
This is mostly only relevant for HF unless we take very high mu triodes where the Miller effect will come into play as it will multiply this capacitance and form a virtual capacitive filter element.
(I'm grossly simplifying.)
So, it's that chicken/egg enigma if you prefer and entirely dependent on whether want to view this as an intrinsic FB mechanism or not.
To honour the great Harold S.J. Black_ there's an idea for an avatar_ I do.
Cheers,
At first may I repeat my opinion regarding the stages which have comparatively low internal/local feedback at audio frequencies?
http://www.diyaudio.com/forums/showthread.php?postid=339207#post339207
Triode stage (common cathode), working on low impedance load (RL<Ri), with cathode resistor shunted by bypassing capacitor (or without cathode resistor at all).
b) Triode cascode stage (common cathode-common grid) under the same conditions
c) Triode-based long-tailed pair under the same conditions
d) Pentode stage (common cathode), working on any load, with cathode resistor shunted by bypassing capacitor (or without cathode resistor at all) and without feedback to the second grid.
I'm sure that one can design amplifier with inaudible levels of distortion using these stages
OK, John Curl wrote: "Actually, I would prefer not to use feedback", Charles Hansen manufactures low-feedback products, so as Nelson Pass offers us differential power FET designs. AKSA wrote that the addition of some second-order nonlinearity to the input differential pair is good for sonic. I could site numerous posts from other respectful members about this or that feedback stage/design/amp which sounds unsatisfactory. Should we ignore their opinion?
What is the main drawback of the local/overall feedback? The distortion components return back, intermodulate with each other on amplifier nonlinearity, thus the higher order distortion components appear in the output signal. This was perfectly explained by Baxandall quarter century ago and recently by Boyk and Sussman. The individual high-order harmonic level increases with the increase in loop gain at first and decrease later. The careful observation shows than the higher is harmonic order the higher loop gain is required to suppress it. Terje Sandstrom wrote that they got a pleasant sound from Otala/Lohstroh when they increase the loop gain from 20 to 30 dB.
What is my point in the long run? The IMD audibility can be removed by two ways - by the low-feedback design with the relatively high harmonic content with a certain spectrum which mask IMD and by high-feedback designs, with loop gain more than 40dB in audio band, with special steps as error correction, with correspondent penalty in stability and PSRR, layout and component quality significance, etc.
http://www.diyaudio.com/forums/showthread.php?postid=339207#post339207
Triode stage (common cathode), working on low impedance load (RL<Ri), with cathode resistor shunted by bypassing capacitor (or without cathode resistor at all).
b) Triode cascode stage (common cathode-common grid) under the same conditions
c) Triode-based long-tailed pair under the same conditions
d) Pentode stage (common cathode), working on any load, with cathode resistor shunted by bypassing capacitor (or without cathode resistor at all) and without feedback to the second grid.
I'm sure that one can design amplifier with inaudible levels of distortion using these stages
OK, John Curl wrote: "Actually, I would prefer not to use feedback", Charles Hansen manufactures low-feedback products, so as Nelson Pass offers us differential power FET designs. AKSA wrote that the addition of some second-order nonlinearity to the input differential pair is good for sonic. I could site numerous posts from other respectful members about this or that feedback stage/design/amp which sounds unsatisfactory. Should we ignore their opinion?
What is the main drawback of the local/overall feedback? The distortion components return back, intermodulate with each other on amplifier nonlinearity, thus the higher order distortion components appear in the output signal. This was perfectly explained by Baxandall quarter century ago and recently by Boyk and Sussman. The individual high-order harmonic level increases with the increase in loop gain at first and decrease later. The careful observation shows than the higher is harmonic order the higher loop gain is required to suppress it. Terje Sandstrom wrote that they got a pleasant sound from Otala/Lohstroh when they increase the loop gain from 20 to 30 dB.
What is my point in the long run? The IMD audibility can be removed by two ways - by the low-feedback design with the relatively high harmonic content with a certain spectrum which mask IMD and by high-feedback designs, with loop gain more than 40dB in audio band, with special steps as error correction, with correspondent penalty in stability and PSRR, layout and component quality significance, etc.
Hi,
Maybe....however it's a little hard to define what's "audible" and what isn't.
Stages a) through d) would have no local FB as you describe the conditions yet the cascode will have the most distortion. Audible?
Very likely IMO.
OTOH, clever combination of a cascade of some of them may very well cancel out the distortion as will a PP stage that follows it downstream cancel some of the distortion too.
However, IMHO it's not so much the exact distortion figure as such that makes it audible or disturbing to the ear if you like but the its harmonic content.
To give an example: an amplifier with 5% of predominantly even order distortion will sound more pleasant, less obtrusive to the ear than one with a predominantly odd order distortion spectrum.
Also, local feedback deals with the linearisation of that gain block, global feedback deals with the entire amplifier from input to output and lastly, yes feedback CAN add it's own distortion artefacts wich are entirely unrelated to the input.
Still, whether pleasant sounding or not, at the end of the day distortion is, well, distortion.
Cheers,
Maybe....however it's a little hard to define what's "audible" and what isn't.
Stages a) through d) would have no local FB as you describe the conditions yet the cascode will have the most distortion. Audible?
Very likely IMO.
OTOH, clever combination of a cascade of some of them may very well cancel out the distortion as will a PP stage that follows it downstream cancel some of the distortion too.
However, IMHO it's not so much the exact distortion figure as such that makes it audible or disturbing to the ear if you like but the its harmonic content.
To give an example: an amplifier with 5% of predominantly even order distortion will sound more pleasant, less obtrusive to the ear than one with a predominantly odd order distortion spectrum.
Also, local feedback deals with the linearisation of that gain block, global feedback deals with the entire amplifier from input to output and lastly, yes feedback CAN add it's own distortion artefacts wich are entirely unrelated to the input.
Still, whether pleasant sounding or not, at the end of the day distortion is, well, distortion.
Cheers,
Nice, Frank. I include the cascode in the list to check will somebody read it. We can treat cascode as a common grid triode stage with nonlinear cathode resistor (bottom triode). You are perfectly right that cascode has higher amount of unpleasant high order disto. This is true for bjt and fet cascodes.
Pavel, if it will be that simple, we will have only products with such specs on the market
We can treat cascode as a common grid triode stage with nonlinear cathode resistor (bottom triode). You are perfectly right that cascode has higher amount of unpleasant high order disto. This is true for bjt and fet cascodes.Pavel, if it will be that simple, we will have only products with such specs on the market
dimitri said:
Dimitri,
I fully agree with the above, it seems from the research you quoted that the best sonic results are obtained either with no overall feedback or with a lot of overall feedback, but not with low feedback.
But, being familiar with the Baxandall study, I remember that it was based largely on research on simple/single stages. What I don't know is whether this can be extrapolated to a multistage and more complex stages. What's your idea about that?
Jan Didden
dimitri said:
Pavel, if it will be that simple, we will have only products with such specs on the market
It is not that simple to reach it, that's why we do not have these products but massive marketing ideology. And it is only necessary, not sufficient condition (I mean very low thd and only very low order harmonics).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.