You don’t need credentials to show that Rod’s wrong on this – think of a simulation as a type of mathematical proof available for all to use, view, critique:
Green trace is the output of a “pure square law” device B1, properly biased to get a single ended output
(yellow is on top of green at 1KHz in the fft, B1 circuit is inverting in the wave plot, B2 is inside a +1 unity gain feedback loop, input signals adjusted for identical output level)
With no feedback (Green trace B1 circuit) we get a +/- 8 V amplitude fundamental – our amplified signal and a
-20 dB second harmonic => 10% distortion, you can see that there are some numerical errors causing harmonics at ~ -160 dB and below
Yellow is the output from an identical nonlinear square law controlled current source B2 in a unity gain feedback loop with E1 voltage controlled voltage source
The yellow out2 trace in the fft shows a ~50 dB reduction in the 2nd harmonic – this is about the value expected from assuming distortion is reduced by the closed loop gain: the E1 gain of 10x and ~ 20x more from the B2 source when using a unity gain feedback (compare input sine amplitudes in V1 and V5)
You can also see a full harmonic series of all even and odd harmonics where the Green B1 open loop case has essentially none beyond the 2nd - clearly linear feedback around a nonlinear gain stage has "created" new harmonic distortion components
[as usual free LtSpice .asc just rename removing .txt ]
Green trace is the output of a “pure square law” device B1, properly biased to get a single ended output
(yellow is on top of green at 1KHz in the fft, B1 circuit is inverting in the wave plot, B2 is inside a +1 unity gain feedback loop, input signals adjusted for identical output level)
With no feedback (Green trace B1 circuit) we get a +/- 8 V amplitude fundamental – our amplified signal and a
-20 dB second harmonic => 10% distortion, you can see that there are some numerical errors causing harmonics at ~ -160 dB and below
Yellow is the output from an identical nonlinear square law controlled current source B2 in a unity gain feedback loop with E1 voltage controlled voltage source
The yellow out2 trace in the fft shows a ~50 dB reduction in the 2nd harmonic – this is about the value expected from assuming distortion is reduced by the closed loop gain: the E1 gain of 10x and ~ 20x more from the B2 source when using a unity gain feedback (compare input sine amplitudes in V1 and V5)
You can also see a full harmonic series of all even and odd harmonics where the Green B1 open loop case has essentially none beyond the 2nd - clearly linear feedback around a nonlinear gain stage has "created" new harmonic distortion components
[as usual free LtSpice .asc just rename removing .txt ]
It took me the longest time to mentally visualize the mechanism involved. As seems common with difficult concepts they appear obvious in hindsight. The typical topology of a circuit with negative feedback has three ports: the positive input, the negative input and the output. For an opamp it's obvious, for a triode the grid, cathode and plate respectively. For me the ah-ha was the realization that neither input appears undistorted on the output. It was very educational, being from the thermionic camp, to drive the cathode (or feedback return node) of a very linear triode and see the distortion components on the output or plate. There's no magic here, if a signal on the negative input appears distorted on the output terminal, the same must happen to the feedback 'correction' signal as well. It doesn't 'correct' perfectly, the result will be distortion of the fundamental plus distortion of the postive input's distortion components on the output. Higher harmonics. The degree and severity depend on implementation.
Quoted
---It is one of the basics of circuit theory that feedback closed over non-linear circuit creates new harmonic components, that were not present in the signal before applying feedback.---
As no circuit is really linear, the amplifying process, be it with or without a feedback, always has some harmonic components on sinewaves. However, as the pleasure of hearing sinewaves is limited, we usually prefer complex input signals in real life.
As soon as they are two or more sine components in the input signal, intermodulation occurs at the output of any unperfect amplifier. It is directly related to the non-linearity of the amplifer. A non-feedback amplifier whose forward amplification is less linear than of a high feedback amplifier will have much more intermodulation with components not harmonically related to the input.
As far as the input signal is not too complex, the intermodulation seems to add some details, richness and spaciousness to the sound, hence the success of 300B and the like amplifiers.
This is not a criticism to those who like 300B's. But I can't agree when some of them say that the sound is more real with these amplifiers : it may be nice but it is fallacious. Some 300B's or 845's owners know that, love that and live happy. Some others don't admit it at all and become very angry when you tell them.
---It is one of the basics of circuit theory that feedback closed over non-linear circuit creates new harmonic components, that were not present in the signal before applying feedback.---
As no circuit is really linear, the amplifying process, be it with or without a feedback, always has some harmonic components on sinewaves. However, as the pleasure of hearing sinewaves is limited, we usually prefer complex input signals in real life.
As soon as they are two or more sine components in the input signal, intermodulation occurs at the output of any unperfect amplifier. It is directly related to the non-linearity of the amplifer. A non-feedback amplifier whose forward amplification is less linear than of a high feedback amplifier will have much more intermodulation with components not harmonically related to the input.
As far as the input signal is not too complex, the intermodulation seems to add some details, richness and spaciousness to the sound, hence the success of 300B and the like amplifiers.
This is not a criticism to those who like 300B's. But I can't agree when some of them say that the sound is more real with these amplifiers : it may be nice but it is fallacious. Some 300B's or 845's owners know that, love that and live happy. Some others don't admit it at all and become very angry when you tell them.
Forr,
agreed. I was just holding theoretical basis to explain mechanism of creation of new components by feedback over non-linear stage, regardless their amplitude, which is the case.
In real circuits, as you probably now, I prefer feedback designs and I take care about low distortion. I really do not care if higher harmonics of my design lie below -110dB, and it is achievable.
I have made one exception - the preamp with no global feedback, to verify sound difference from global feedback designs.
agreed. I was just holding theoretical basis to explain mechanism of creation of new components by feedback over non-linear stage, regardless their amplitude, which is the case.
In real circuits, as you probably now, I prefer feedback designs and I take care about low distortion. I really do not care if higher harmonics of my design lie below -110dB, and it is achievable.
I have made one exception - the preamp with no global feedback, to verify sound difference from global feedback designs.
feedback for ouput transistor protection...
Hi,
Looking at the Arcam alpha schematic I have there is some sort of negative feedback output stage protection feedback circuit in adition to the global feedback. It looks like current from the output stage is low pass filtered and fed back into the driver stage.
However, my reading of the schematic is that this output protection is in continual operation so it must be negative feedback..
Is this paper a good read ?
http://www.aes.org/e-lib/browse.cfm?elib=5991
Hi,
Looking at the Arcam alpha schematic I have there is some sort of negative feedback output stage protection feedback circuit in adition to the global feedback. It looks like current from the output stage is low pass filtered and fed back into the driver stage.
However, my reading of the schematic is that this output protection is in continual operation so it must be negative feedback..
Is this paper a good read ?
http://www.aes.org/e-lib/browse.cfm?elib=5991
Hi,
I maybe my simulations are wrong , but i could not see any addictional harmonics above -140db , when the feedback was applied. Maybe erronous is to use so much feedback as op amp supplies.
First picture is without any feedback(only emitter resistor); second is with strong feedback with op amp in circuit.
regards,
Lukas
I maybe my simulations are wrong , but i could not see any addictional harmonics above -140db , when the feedback was applied. Maybe erronous is to use so much feedback as op amp supplies.
First picture is without any feedback(only emitter resistor); second is with strong feedback with op amp in circuit.
regards,
Lukas
Attachments
You see the new harmonics immediately if you simulate with a circuit that creates 2nd harmonic only in open loop.
BUT, having an amp with low enough open loop distortion and high enough feedback these harmonics drop below noise threshold.
Mike
EDIT:
http://www.diyaudio.com/forums/showthread.php?postid=843339#post843339
BUT, having an amp with low enough open loop distortion and high enough feedback these harmonics drop below noise threshold.
Mike
EDIT:
http://www.diyaudio.com/forums/showthread.php?postid=843339#post843339
OK, I think I see where everyone is coming from now. I think I misunderstood the initial paramaters of the discussion. Mea culpa. (Which just proves you shouldn't post when tired. . . )
I also think this proves that you should avoid bad amp design in the first place! If you start with a bad design, I think we all agree feedback won't fix it. Assuming you got something close to right the first time through, I think we can also agree that feedback can substantially reduce the distortion present in an amp.
How about it? Can't we all just get along? 😀
I also think this proves that you should avoid bad amp design in the first place! If you start with a bad design, I think we all agree feedback won't fix it. Assuming you got something close to right the first time through, I think we can also agree that feedback can substantially reduce the distortion present in an amp.
How about it? Can't we all just get along? 😀
Hi Dfdye
---you should avoid bad amp design in the first place! If you start with a bad design, I think we all agree feedback won't fix it. Assuming you got something close to right the first time through, I think we can also agree that feedback can substantially reduce the distortion present in an amp.---
Yes, we can agree.
But is there something like an agreement about what is a bad amp... or a good amp design ?
---you should avoid bad amp design in the first place! If you start with a bad design, I think we all agree feedback won't fix it. Assuming you got something close to right the first time through, I think we can also agree that feedback can substantially reduce the distortion present in an amp.---
Yes, we can agree.
But is there something like an agreement about what is a bad amp... or a good amp design ?
As infinite regression rears its ugly head, I think I will run back to my lab and work on MY amp some more. . . 😀forr said:
Seriously, I think a few basic goals can be agreed upon, sure. I'll go first: high order harmonic distortion should be avoided. As to the circuit topology as to how this is accomplished, I am not so sure we can get a similar agreement!
Your turn 😉
You look with google for "John Curl Interview". There is a part of this interview that said to advoid higher order harmonics, the cct has to produce no harmonic at all at possible.
Like MikeB's said, the cct only producing 2nd harmonic can produce 4th etc if the feedback loop is closed. So to advoid higher, the cct should be producing no harmonic as possible. Maybe your sim is doing this 😀
Hi, Forr, that's a good one 😀 It is not always that the "cleanest" is the best sounding. I think it's because that our ear itself is not the most linear device. Not so distorted, but also not so clean may sound the best. To me it is not the ammount, but the spread of the harmonics, the weight between harmonics that matters.
Late night reading yielded this little comment from Self's "Audio Power Amplifier Design Handbook" that seemed on topic:
As such, though the sims are indeed instructive in an academic sense, once real devices are used, the generation of higher order harmonics only becomes an issue if feedback is used improperly. I think this is where there may be a difference in opinion/perception, but though I am again posting when tired, I have a sneaking suspicion this observation won't end the debate. 😀
David
As such, though the sims are indeed instructive in an academic sense, once real devices are used, the generation of higher order harmonics only becomes an issue if feedback is used improperly. I think this is where there may be a difference in opinion/perception, but though I am again posting when tired, I have a sneaking suspicion this observation won't end the debate. 😀
David
Hi PMA,
---In real circuits, as you probably now, I prefer feedback designs and I take care about low distortion.---
I know your preference because, for some reasons, I carefully followed the Blowtorch thread and found your comments among the most interesting. Here, I repeat some of them :
---I often listen to classical music and I do not like it through tube amps. I can even tell (recognize) that recording was done through tube preamp. Me personally I am bored by euphonic sameness of the tube sound.---
---We sometimes feel that the DIY community is pushed to some sort of "subjective truth", that we do not find universal, so we dare to oppose. I do not think this would be a problem and a lot of e-mails I receive from people with different tastes than non-feedback e.g. convince me that I am probably not wrong.---
---I have also tried a lot of comparisons between global NFB a non-global NFB designs. Sometimes I tend to prefer non-global NFB line stages, especially for jazz and mainstream music. For complex classical music, like great symphony orchestra, the properly designed NFB line stage is much much better, IMHO, regarding accuracy and resolution, especially regarding sound of bow instruments and brass.---
This last comment made me think that, on simple signals, those having not so many harmonics, are nicely rendered when euphonic distorsion of non-nfb circuits is added. For complex and rich signals, the intermodulation products might dominate the harmonic content, and linear (= hi-nfb) amplifiers are wellcome.
A great part of the debate between subjectivists and objectivists and between no or low feedback supporters and hi linearity advocates may rely here
I am not aware of an amplifier changing of features, from euphony to linearity, according to the complexity of the sound. Maybe such a strange amp exists, but according to the level, having a (nice) bad linearity at low levels and perfect linearity at high levels.
---In real circuits, as you probably now, I prefer feedback designs and I take care about low distortion.---
I know your preference because, for some reasons, I carefully followed the Blowtorch thread and found your comments among the most interesting. Here, I repeat some of them :
---I often listen to classical music and I do not like it through tube amps. I can even tell (recognize) that recording was done through tube preamp. Me personally I am bored by euphonic sameness of the tube sound.---
---We sometimes feel that the DIY community is pushed to some sort of "subjective truth", that we do not find universal, so we dare to oppose. I do not think this would be a problem and a lot of e-mails I receive from people with different tastes than non-feedback e.g. convince me that I am probably not wrong.---
---I have also tried a lot of comparisons between global NFB a non-global NFB designs. Sometimes I tend to prefer non-global NFB line stages, especially for jazz and mainstream music. For complex classical music, like great symphony orchestra, the properly designed NFB line stage is much much better, IMHO, regarding accuracy and resolution, especially regarding sound of bow instruments and brass.---
This last comment made me think that, on simple signals, those having not so many harmonics, are nicely rendered when euphonic distorsion of non-nfb circuits is added. For complex and rich signals, the intermodulation products might dominate the harmonic content, and linear (= hi-nfb) amplifiers are wellcome.
A great part of the debate between subjectivists and objectivists and between no or low feedback supporters and hi linearity advocates may rely here
I am not aware of an amplifier changing of features, from euphony to linearity, according to the complexity of the sound. Maybe such a strange amp exists, but according to the level, having a (nice) bad linearity at low levels and perfect linearity at high levels.
Just one more comment. IMO linearity is also very important for quite small signals, though not very often measured. I suspect most of class AB designs with high NFB factor to fail in subjective assesment for the reason of rise of high order harmonics at small signals. NFB does not correct properly sharp cross-over region. I have always obtained much better results with Class A or mosfet-error correction AB output stage compared to standard class B/AB output stage. IMO this is not less important than input stage design.
Yes, the harmonics created by crossover distortion are way too high frequency to be really compensated/reduced by global feedback. (especially for trebles)
Mike
Mike
Feedback has it's limitation, especially when headed with problems in output stage. Crossover distortion is really heavy duty for feedback to fix. Maybe it's the main problem 😀
Figure 1 is a simplified power amp topology. The signal tapped from leg A of differential pair, and so on, feedback comes to inverting input.
To see that feedback has it's limitation, try to hear what is the sound likes from the signal of leg B of differential pair. So there is 2 amps left and right, but the feedback node only comes from the right amp. The left amp is open loop, so we must put cap at this output before the speakers.
By figure2 experiment, you will see that the system has a limited capability for fixing things. There's not unlimited energy to fix everything.
The worse the output stages are, the worse you will hear the sound from the left side of the amp (out B). If you biased both output sides very heavily classA, you will be able to hear good music from all sides of output.
If both side of output is low biased classAB, especially mosfet, you will notice that the differential will take side very much for fixing the out A and sacrificing (because there's nothing much power left for fixing after fixing out A) out B. Because feedback is taking side for out A, you can hear undistorted music here, but outB is very-very distorted.
Attachments
maybe it will be not fresh idea, but I also observed that introducing the feedback lowers the 2nd harmonic but cause rise of high order harmonics which make the sound harsh/unpleasant.
Only deep feedback reduces all of the distortions but their character and amount became equal, for example the 2nd harmonic is almost the same like the 9th harmonic.
I also observed that high amount of the 2nd and the 3rd (and even the 4rth) harmonics makes the sound very nice but only in case when higher order are almost absent.
Only deep feedback reduces all of the distortions but their character and amount became equal, for example the 2nd harmonic is almost the same like the 9th harmonic.
I also observed that high amount of the 2nd and the 3rd (and even the 4rth) harmonics makes the sound very nice but only in case when higher order are almost absent.
padamiecki said:
This happens only if open loop distortion is too high (~10%), if ol-dist is low enough (~1%) you get a nice decaying spectrum of harmonics where the high order ones are already below noise level.
Mike
- Status
- Not open for further replies.