Articles that I have read over the years on amp distortion tend to deal with the distortion characteristics in the voltage amplification stage and then just lump on the output stage without any further comments in this respect.
If one considers a class A output (for convenience purposes) what is the distortion added by this stage?
Following on from the recent forum topic of no NFB in an amp because of the possible (probable) effects of speaker "back EMF" to the input stages, would it not be possible to design an amp that only had NFB applied to the voltage gain section and run the output in a non NFB manner. In other words the current output acts as a "buffer" between the input stage and the speakers?
Has this been tried before?
I seem to remember from some time back a power amp that was built from and Op-amp input with complimentray pair output transistors and, from memory, the NFB was only on the op-amp.
I assume one of the problems faced would be the output resistance of the amp?
Once upon a time the amp's "Damping Factor" was all in vogue but this seems to have bitten the dust of late (at least in some of the specs I have seen). I take it that the "damping factor" would also play an important part in reducing the feedback from the speakers to the input stage?
Regards,
If one considers a class A output (for convenience purposes) what is the distortion added by this stage?
Following on from the recent forum topic of no NFB in an amp because of the possible (probable) effects of speaker "back EMF" to the input stages, would it not be possible to design an amp that only had NFB applied to the voltage gain section and run the output in a non NFB manner. In other words the current output acts as a "buffer" between the input stage and the speakers?
Has this been tried before?
I seem to remember from some time back a power amp that was built from and Op-amp input with complimentray pair output transistors and, from memory, the NFB was only on the op-amp.
I assume one of the problems faced would be the output resistance of the amp?
Once upon a time the amp's "Damping Factor" was all in vogue but this seems to have bitten the dust of late (at least in some of the specs I have seen). I take it that the "damping factor" would also play an important part in reducing the feedback from the speakers to the input stage?
Regards,
polsol said:
Hi Polsol
Yes it is possible and it can be easely done.
The feedback is taken from the colector of the VAS transistors and the output stage only do the current gain.
Many comercial amps use that aproach , come to mind Densen in the DM10 and Onkyo in the Integra 9711 series .
They are advertised as Non Negatif Feedback amps , when they must really be called Overall Feedback Free amps.... a la Nelson Pass...
Regards
Tony,
Many issues in your question; it's complex. Pardon me if I cover a little ground before a fuller explanation.
First, the voltage amplifier of any amplifier is operating in common emitter mode. This is the most distortive of all the configurations, known loosely as CE, common collector (emitter follower) and common base.
The distortion introduced is an order of magnitude higher than that of an emitter follower (EF) and compounded by the interelectrode capacitance between base and collector which changes with applied voltage and current. This capacitance, usually called Miller capacitance, is effectively multiplied by the beta of the transistor, and is considerable, leading to HF rolloff and some nasty distortion effects.
In any global feedback loop there is a propagation delay which at some very high frequency turns negative feedback into positive feedback. This abruptly turns the amplifier into an oscillator, so it is immediately clear that something must be done to the voltage amplifier to cut its gain back to below unity so that by this very high frequency the now positive feedback does not bring a destructive paroxysm of oscillation.
Thus, the heart of any amplifier becomes the voltage amplifier, principally because all the compromises lie in this circuit element.
It is certainly possible to drive the output stage in open loop. I've been down this path, and the results are interesting. In brief, taking feedback from the voltage amplifier rather than the output stage makes the amplifier sound very like a tube amplifier, with a lovely, twee sound and not too much bass control. Incidentally, imaging is superior, and you can drive almost any load with ease....... so, you can't have it all, where would you put it?
Obviously, this configuration reduces damping factor, which explains the effect on the bass. But it shrinks the feedback loop, and propagation delay no longer encompasses the output stage, so the pole frequency, the point at which the amplifier would ordinarily go up, is much higher, making stability much easier to ensure. We can use less lag compensation across the VAS to achieve this; this is good for sonics.
It is adviseable to run Class A in any output stage which is running open loop. This is because the crossover disjunction, the so-called 'dead zone', is no longer important because neither of the two output devices is ever switched off. And the high quiescent current of Class A will ensure lower large signal distortion arises in the Emitter Follower(s), and thus helps to improve damping factor. Large signal distortion is caused by higher current flow through collector/emitter as the signal swings higher (and lower); to generate more current flow, the base must be driven harder, and the result is that the step voltage between input and output, the base/emitter voltage, increases. This has the effect of compressing the waveform, adding harmonic overtones, generally H2 to H6. Ordinarily, this distortion mechanism is well nulled in a global feedback loop, but if the output stage is open loop, that is, outside the feedback loop, it can be minimized by biasing both ouput devices at high current as in Class A.
Cheers,
Hugh
Many issues in your question; it's complex. Pardon me if I cover a little ground before a fuller explanation.
First, the voltage amplifier of any amplifier is operating in common emitter mode. This is the most distortive of all the configurations, known loosely as CE, common collector (emitter follower) and common base.
The distortion introduced is an order of magnitude higher than that of an emitter follower (EF) and compounded by the interelectrode capacitance between base and collector which changes with applied voltage and current. This capacitance, usually called Miller capacitance, is effectively multiplied by the beta of the transistor, and is considerable, leading to HF rolloff and some nasty distortion effects.
In any global feedback loop there is a propagation delay which at some very high frequency turns negative feedback into positive feedback. This abruptly turns the amplifier into an oscillator, so it is immediately clear that something must be done to the voltage amplifier to cut its gain back to below unity so that by this very high frequency the now positive feedback does not bring a destructive paroxysm of oscillation.
Thus, the heart of any amplifier becomes the voltage amplifier, principally because all the compromises lie in this circuit element.
It is certainly possible to drive the output stage in open loop. I've been down this path, and the results are interesting. In brief, taking feedback from the voltage amplifier rather than the output stage makes the amplifier sound very like a tube amplifier, with a lovely, twee sound and not too much bass control. Incidentally, imaging is superior, and you can drive almost any load with ease....... so, you can't have it all, where would you put it?
Obviously, this configuration reduces damping factor, which explains the effect on the bass. But it shrinks the feedback loop, and propagation delay no longer encompasses the output stage, so the pole frequency, the point at which the amplifier would ordinarily go up, is much higher, making stability much easier to ensure. We can use less lag compensation across the VAS to achieve this; this is good for sonics.
It is adviseable to run Class A in any output stage which is running open loop. This is because the crossover disjunction, the so-called 'dead zone', is no longer important because neither of the two output devices is ever switched off. And the high quiescent current of Class A will ensure lower large signal distortion arises in the Emitter Follower(s), and thus helps to improve damping factor. Large signal distortion is caused by higher current flow through collector/emitter as the signal swings higher (and lower); to generate more current flow, the base must be driven harder, and the result is that the step voltage between input and output, the base/emitter voltage, increases. This has the effect of compressing the waveform, adding harmonic overtones, generally H2 to H6. Ordinarily, this distortion mechanism is well nulled in a global feedback loop, but if the output stage is open loop, that is, outside the feedback loop, it can be minimized by biasing both ouput devices at high current as in Class A.
Cheers,
Hugh
Dean,
Just a quick reply to say thanks for a very comprehensive reply - much appreciated - which I will study in detail tomorrow (I'm just about to head home from work - now 4pm in Kuwait).
I mentioned Class A as I couldn't see a way around trying to get around the cross-over distortion in class AB.
Cheers
Just a quick reply to say thanks for a very comprehensive reply - much appreciated - which I will study in detail tomorrow (I'm just about to head home from work - now 4pm in Kuwait).
I mentioned Class A as I couldn't see a way around trying to get around the cross-over distortion in class AB.
Cheers
polsol said:
In order to use a class AB emitter follower current amplifier outside of the NFB loop, as just an emitter follower, you have to bias the two transistors separate of the rest of the circuit that is enclosed in the NFB loop. Using current sources that are equal so that no major DC offset is present, and a bias servo for thermal compensation establishing an overall temperature coefficient as close to 1 as possible to eliminate crossover. If you make one of the current sources slightly adjustable, DC offset can be adjusted and output EF stage can be capacitor coupled to input drivers. Also output impeadence is more or less fixed and lower impeadences may cause problems.
There is one major drawback to not including the output stage in the NFB loop. If you desire to have say 6A peak output, then the transistor must be fairly linear up to 6A or symetrical harmonic distortion will occur. However, if included in the NFB loop, and the amplifier circuit before the emitter folower output stage has been over designed for peak voltage & current, and can operate at high frequencies (much higher than audio), then harmonic distortion will be generated equal and opposite in phase to the distortion created by the saturation of the device, and will mathmatically cancel out leaving a much lower distorted sine wave. This will compensate somewhat for certain nonlinearities inherent to all transistors. Simply stated, transistors are not linear devices, just have a region of operation that is very close to linear. This is where a hybrid SS feedback driver stage combined with a power tube output stage not included in the NFB loop would be more prudent in achieving linearity if this is the route of choice. Tubes are very linear by design. More expensive and usually requires a lot of iron but would be worth the $$$.
cunningham said:This is where a hybrid SS feedback driver stage combined with a power tube output stage not included in the NFB loop would be more prudent in achieving linearity if this is the route of choice. Tubes are very linear by design. More expensive and usually requires a lot of iron but would be worth the $$$.
Hi Cunningham
But a tube output stage without overall feedback will have a much higher output impedance (lower damping factor) then a triple follower...even if we use power triodes...
There seems to be a correlation between output stiffness (low output impedance) and variously described unpleasant sonic characteristics.
Solid state amplifier designers take great pains to keep output impedance low exactly in order to mitigate against feedback from the speaker reactance back to the earlier stages (input or VAS).
Is it possible that today's high efficiency tweeters would prefer to be driven with a higher impedance? I mean, if people are willing to turn amplifiers into room heaters to avoid output distortion, why not put the heat to good use by increasing the output impedance of the amplifier, attempting to approximate the speaker's conjugate impedance?
Any thoughts?
Solid state amplifier designers take great pains to keep output impedance low exactly in order to mitigate against feedback from the speaker reactance back to the earlier stages (input or VAS).
Is it possible that today's high efficiency tweeters would prefer to be driven with a higher impedance? I mean, if people are willing to turn amplifiers into room heaters to avoid output distortion, why not put the heat to good use by increasing the output impedance of the amplifier, attempting to approximate the speaker's conjugate impedance?
Any thoughts?
One of N.P.'s "hits", the famous Stasis amp used this kind of feedback:
http://www.diyaudio.com/forums/attachment.php?s=&postid=109546
And I think I have seen it on a schematic of a GAS or ML as well.
regards
Charles
http://www.diyaudio.com/forums/attachment.php?s=&postid=109546
And I think I have seen it on a schematic of a GAS or ML as well.
regards
Charles
polsol said:
Hi Polsol
Is very easy to use this approach in any type of amp (or schematic) that use a unity gain output stage (99% of them).
Take out the feedback resistor (from output to input)... and connect in the some point at the input, 2 resistors with the double of the value of the original...the other side of this 2 resistors must be connected each at each side of the bias network (VBE) or to each base of the driver transistors.
I agree 100 % with coments about the sound of that approach Hugh made in post # 3.
Cheers
local NFB at output stage
Does anyone have any experience with an output stage that incorporates some gain and local negative feedback?
In other words, take the global feedback from the voltage amplifier as proposed elsewhere. Then, add local feedback with gan in the feedback path (possibly using a good op amp) to build a good unity gain buffer output stage.
I'm probably overlooking something silly here, but if anyone knows why this isn't done I'd love to know...
Does anyone have any experience with an output stage that incorporates some gain and local negative feedback?
In other words, take the global feedback from the voltage amplifier as proposed elsewhere. Then, add local feedback with gan in the feedback path (possibly using a good op amp) to build a good unity gain buffer output stage.
I'm probably overlooking something silly here, but if anyone knows why this isn't done I'd love to know...
Is the premise of this thread that including the output devices in the NFB loop will lead to worse sound?
Why should this be...and why should it be that NFB will make one part of the circuit better but not the whole circuit? These are the important questions.
IMO the largest and most audible distortion occurs in the output devices; the rest of the circuit should be easy to make extremely linear as it involves small currents, class A bias and high performance transistors. Leaving the output devices outside the loop is sort of avoiding the issue but is a very practical way to make something sound quite good, or at least to avoid some tricky issues that can make something sound quite bad. I have made many circuits like this myself and they sounded good and are well-behaved.
Why should this be...and why should it be that NFB will make one part of the circuit better but not the whole circuit? These are the important questions.
IMO the largest and most audible distortion occurs in the output devices; the rest of the circuit should be easy to make extremely linear as it involves small currents, class A bias and high performance transistors. Leaving the output devices outside the loop is sort of avoiding the issue but is a very practical way to make something sound quite good, or at least to avoid some tricky issues that can make something sound quite bad. I have made many circuits like this myself and they sounded good and are well-behaved.
On the face of it this makes no sense to me at all. Consider the transistor's point of view - it knows not what confirguration it is in, it amplifies current in the same way regardless. The beta is not affected by the configuration. The miller effect exists in both CE and CC modes and generates the same base current for the same dVcb/dt. The output transistors in CC mode experience large variance in Ic and Vcb - thus variance in beta and Ccb - normally much moreso than the "VAS" transistor. Can you explain in more detail what you mean?
This statement as such is true. But you have to keep in mind that we don't feed our VAS by a current-source usually but something between current and-voltage source. If we now take the nonlinear interrelation of base -current vs -voltage into account we end up with quite some nonlinearity. For this consideration I deliberately left away the nonlinear behaviour of current gain vs collector current.
For the emitter- follower we have this nonlinear behaviour as well but the RELATIVE error on the output voltage is much smaller due to the much smaller voltage-gain in use (i.e. the ratios between the input voltage, output voltage and the absolute magnitude of the error are much more favourable).
Regards
Charles
P.S.: I am currently thinking about formulating the above in a more understandable way.
Bam,
I'm not sure how to respond to your points in your post; I was not aware I said that the sound with output devices outside the NFB loop was worse; I merely said it was different. In fact there are swings and roundabouts........ it just ain't that simple, and categorical statements are not my bag anyway.
You believe that the output devices determine the sonic flavour of an amplifier; I believe it is the voltage amplifier. That's fine; there is enough room under the sun for both beliefs!!
I will admit that these beliefs are the result of detective work on amp design, and there is no single answer here. In truth, both are important, and you'll soon notice if you change the output devices.
Miller capacitance is less important in an emitter follower than a common emitter configuration. This is because of the influence of beta, which in the latter, and via the depletion layer on the reverse biased collector/base junction, causes HF rolloff far sooner than in an emitter follower, which is nimble by comparison. As it happens, we want the VAS to be the slowest part of the circuit anyway, since its OLG must be lower than unity by the pole frequency. No such constraint need bother the output stage.
Charles, if you can explain it better I would be grateful.......
Cheers,
Hugh
I'm not sure how to respond to your points in your post; I was not aware I said that the sound with output devices outside the NFB loop was worse; I merely said it was different. In fact there are swings and roundabouts........ it just ain't that simple, and categorical statements are not my bag anyway.
You believe that the output devices determine the sonic flavour of an amplifier; I believe it is the voltage amplifier. That's fine; there is enough room under the sun for both beliefs!!
I will admit that these beliefs are the result of detective work on amp design, and there is no single answer here. In truth, both are important, and you'll soon notice if you change the output devices.
Miller capacitance is less important in an emitter follower than a common emitter configuration. This is because of the influence of beta, which in the latter, and via the depletion layer on the reverse biased collector/base junction, causes HF rolloff far sooner than in an emitter follower, which is nimble by comparison. As it happens, we want the VAS to be the slowest part of the circuit anyway, since its OLG must be lower than unity by the pole frequency. No such constraint need bother the output stage.
Charles, if you can explain it better I would be grateful.......
Cheers,
Hugh
Just to clarify why I initially asked the question regarding the distortion of an output stage:
Rightly or wrongly I feel that when an amp is connected to a speaker system the reactances of the later can have a substantial effect on the instantaneous linearity of the amp if any "back EMF" were to find it's way into the NFB loop - even though the amp may ahve a very low output resistance some current must be fed back to the voltage gain stage thus affecting the output signal. If the output stage could be isolated (what i would refer to as a "crowbar") from the input then the amp should be more "faithfull" in its function. However, I need to understand the affects that this would have on the overall distortion and hence the affect on reproduction (or simply stated "sound"). My idea would be to build an amp that caters for both approaches - "local" (voltage stage) and overall NFB. Fitted with a switch (or jumper) it would be possible to compare the two possibilities and I could judge for myself.
On paper most amps are very good in terms of their performance but amps do sound different and I (along with thousands of otehrs) would like to try and find out why.
Ultimately sound is subjective (and also will differ from individual to individual based on the facial features - e.g. shape of ears) and one mans mead may be anothers poison but at least I can then choose for myself which topology I prefer.
Rightly or wrongly I feel that when an amp is connected to a speaker system the reactances of the later can have a substantial effect on the instantaneous linearity of the amp if any "back EMF" were to find it's way into the NFB loop - even though the amp may ahve a very low output resistance some current must be fed back to the voltage gain stage thus affecting the output signal. If the output stage could be isolated (what i would refer to as a "crowbar") from the input then the amp should be more "faithfull" in its function. However, I need to understand the affects that this would have on the overall distortion and hence the affect on reproduction (or simply stated "sound"). My idea would be to build an amp that caters for both approaches - "local" (voltage stage) and overall NFB. Fitted with a switch (or jumper) it would be possible to compare the two possibilities and I could judge for myself.
On paper most amps are very good in terms of their performance but amps do sound different and I (along with thousands of otehrs) would like to try and find out why.
Ultimately sound is subjective (and also will differ from individual to individual based on the facial features - e.g. shape of ears) and one mans mead may be anothers poison but at least I can then choose for myself which topology I prefer.
That would be quite helpful. I'm confused by your reference to relative voltage error. Eg: if the voltage change is small then isn't the relative error larger?
What I was trying to say is that the distortion contribution from the output devices is much larger than the driving circuitry (if designed well). The circuit is a system - each "stage" interacts with the next. This will lead to changes in the VAS causing significant audible affects. However, this observation does not imply the VAS generates the primary distortion.
I disagree. The only reason the VAS rolls off is because it is fed with a relatively high Z source. Do the same for the EFs and you'll get the same result. And isn't the VAS quite a high Z source?
Are you not confusing back EMF with impedance? I mean, they are the same thing as seen by the amp. So are you not saying that a variable Z load will induce non-linearities in an amp? I don't think there is an unusual effect here and I don't think it has anything particular to do with NFB.
What do you think about the ALTMANN "SPLIF" Amplifier Topology.
http://www.altmann.haan.de/splif_page/
I tested it on a Rotel 971 with good results.
http://www.altmann.haan.de/splif_page/
I tested it on a Rotel 971 with good results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.