Yes, that's the idea.
The thing is that the symmetry is not the only factor relevent to distortion cancellation, it is also the shape of the curve. And as has been said the balance of the spectral content is very important to the sound so removing only some distortion won't necessarily be better.
Not all tube amplifiers are richer in even order than odd order, that is usually only correct for single ended triode amplifiers. Push-pull, be it triode or penthode usually have at least as much third order as 2nd order distorsion products.
I dont think it is important to reduce distorsion to very low levels, so even if even order products would cancel in a bridge I dont believe it will improve the sound of a well designed tube amplifier in a significant way.
I think an amplifier in order to sound good must have low distorsion before feedback is applied so the new higher order products that are inevitably produced due to feedback are neglible.
BTW, contrary to many others who play with tube amps I believe that feedback improves sound quality, (if applied correctly).
Regards Hans
Tubevr,
I'm of the opinion that all distortion is undesirable, although given some distortion exists it's spectra will greatly affect the sound.
If anyone has SE valve amps and has tried bridging them I'd love to know how they sounded. I suppose SE FET amps might be good candidates for this too.
I also hold no prejudice against feedback provided it is sympathetic to the characteristics of the system to which it is applied.
You say that feedback inevitably produces higher order distortion products. Why?
I'm of the opinion that all distortion is undesirable, although given some distortion exists it's spectra will greatly affect the sound.
If anyone has SE valve amps and has tried bridging them I'd love to know how they sounded. I suppose SE FET amps might be good candidates for this too.
I also hold no prejudice against feedback provided it is sympathetic to the characteristics of the system to which it is applied.
You say that feedback inevitably produces higher order distortion products. Why?
Have you folks agreed on anything yet? ;-)
I have read all kinds of things on this thead.
This is what I know:
Asymmetric distortion, that is normally spoken of as even order, WILL CANCEL, if you mix it with an inverted signal of the same magnitude. This is normally called push-pull design.
All active components normally create lots of even order distortion, because the voltage gain of all individual active components normally increases with increasing current.
Analog magnetic tape creates [compressive] odd order distortion (S-curve), because away from 0 level, the output loses [gain] in both directions. This cannot be fixed by push-pull operation. However, you can create a different form of 3'rd harmonic [expansive] or inverted S curve from the input signal, and mixing it with the compressive signal distortion, will cancel it out. This is rarely done, because often, the residue created by the process has even more higher order distortion, which is MUCH worse sounding than third harmonic, in almost any amount.
I have read all kinds of things on this thead.
This is what I know:
Asymmetric distortion, that is normally spoken of as even order, WILL CANCEL, if you mix it with an inverted signal of the same magnitude. This is normally called push-pull design.
All active components normally create lots of even order distortion, because the voltage gain of all individual active components normally increases with increasing current.
Analog magnetic tape creates [compressive] odd order distortion (S-curve), because away from 0 level, the output loses [gain] in both directions. This cannot be fixed by push-pull operation. However, you can create a different form of 3'rd harmonic [expansive] or inverted S curve from the input signal, and mixing it with the compressive signal distortion, will cancel it out. This is rarely done, because often, the residue created by the process has even more higher order distortion, which is MUCH worse sounding than third harmonic, in almost any amount.
Yes agreed but if the difference between an amplifier with 0.1% or .0001% is not noticeable when listening I dont see why there should be a need to go lower then 0.1% especially when other problems can pop up due to the methods used to lower the distorsion, (for instance cancellation methods).
Assume an amplifier with feedback have a certain amount of 2nd and 3rd order distortion. If feedback is applied an amount of the distorded signal is brought back to the input in opposite phase and thus will lower gain and 2nd and 3rd order distortion.
But! the 2nd and 3rd order products brought back to the input by the feedback loop will also be distorted by the amplifier and will therefore produce higher order products like 4th, 5th, 6th aso. This effect of feedback is well understood and it can be showned that for small amount of feedback the sonic result is often worse then for the amplifier without feedback, (lower order products are reduced but higher order products are added.
Regards Hans
Ok. Now considering we can only hear up to 20kHz, just how much feedback factor and how quick does it need to be in a typical design to ensure you cannot hear the distortion?
I'm supposing that if the harmonics escalate out of the audio band you won't hear them so they don't matter. With transistors having ft in the 10MHz+ region why do we still hear the feedback added harmonics?
If it where so that feedback could be applied unlimited this would never be a problem but many amplifiers has limited open loop bandwidth therefore limited gain and limited feedback is therefore available at higher frequencies shown as distortion increasing, (at higher frequencies).
In such an amplifier this and similar effects are shown by the IM spectrum for higher frequencies which contains many higher order products.
If it is heareable or not I think depends on who is listening, I have heard effects in this kind of amplifiers that I believe is related to "feedback induced distortion" but I should admit that even if these where so called high end amplifiers they where quite old, (from the eighties).
Instead of relying on feedback as the universal cure, (as it is used many times) it make more sense of trying to reduce distortion in the first place, (before feedback is applied) and also to keep open loop bandwidth as high as possible.
Regards Hans
Well I know a couple of essential reasons for employing feedback.
1) Transistors and valves are current output devices. Without some form of voltage feedback you can't control the voltage across a non-resistive load.
2) Transistors and valves produce odd and even distortion products, so bridging or push-pulling won't cancel the distortion properly.
Tubetvr menitoned that distortion doesn't really matter if it is <0.1% or so because you can't hear it.
So isn't the key here to work out how much feedback is required, for a given level of open loop distortion, up to 20kHz such that the open-loop products are reduced to inaudible levels and the feedback-induced side-products are also low enough below 20kHz (regarless of what they are above 20kHz)?
Work this out, apply that level of global feedback and muck about at high f to make it stable and you're done.
Seems a fundamental question to me.
1) Transistors and valves are current output devices. Without some form of voltage feedback you can't control the voltage across a non-resistive load.
2) Transistors and valves produce odd and even distortion products, so bridging or push-pulling won't cancel the distortion properly.
Tubetvr menitoned that distortion doesn't really matter if it is <0.1% or so because you can't hear it.
So isn't the key here to work out how much feedback is required, for a given level of open loop distortion, up to 20kHz such that the open-loop products are reduced to inaudible levels and the feedback-induced side-products are also low enough below 20kHz (regarless of what they are above 20kHz)?
Work this out, apply that level of global feedback and muck about at high f to make it stable and you're done.
Seems a fundamental question to me.
I like to throw in again my distortion killer circuit as a functional diagram 
Right side mirrors difference between left side and input.
Output stages are classA.
In the sim world K2 up to K9 are improved from -120 dB to - 155 dB.
With a more complicated circuit K2 to -180 dB and K3 ... K9 to -160dB.
Do I have to fear higher order artefacts & bad sound in the real world ?
At the moment I'm too lazy to build that
Right side mirrors difference between left side and input.
Output stages are classA.
In the sim world K2 up to K9 are improved from -120 dB to - 155 dB.
With a more complicated circuit K2 to -180 dB and K3 ... K9 to -160dB.
Do I have to fear higher order artefacts & bad sound in the real world ?
At the moment I'm too lazy to build that
Attachments
Bernhard said:
unless you prove otherwise, the default answer is always "yes".
BTW, are you sure your idea works? I am not sure if those little "blocks" mean real amps or you mean literally comparators (or opamps without local feedbacks). Anyway, assuming perfect amps, the output on the left channel should be precisely +V1. the output on the right channel should be precisely +V1 , or a function of V1 if you assume less-than-perfect opamps.
then exactly what do you expect to happen on the 8ohm load?
It is the very simplified version of schematic in post #19.
Left side: V1(sin t) + distortion
Right side: Left side - V1(sin t) = same distortion like left side.
+in - -in = out
On the right side = 0 = virtual ground (+ same distortion like left side).
I assume that the distortion of the right side is very low and neglible, because the voltage swing (<1mV for 8V across load) is very small.
In any case it is like that in the sim.
Each side alone is -120dB.
Right side carrying distortion of left side is -46dB.
Across load is -155dB.
Left side: V1(sin t) + distortion
Right side: Left side - V1(sin t) = same distortion like left side.
+in - -in = out
On the right side = 0 = virtual ground (+ same distortion like left side).
I assume that the distortion of the right side is very low and neglible, because the voltage swing (<1mV for 8V across load) is very small.
In any case it is like that in the sim.
Each side alone is -120dB.
Right side carrying distortion of left side is -46dB.
Across load is -155dB.
I'm sorry, I might be missing something, but I don't see any amazing distortion reduction in the circuit in #37. It looks like a bridged output to me. We have been making amps like this for 1/3 century. They always have low, but real distortion. Also, you have a short on the input. This can be fixed by assigning the opposite input polarities on one side.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.