I wrote another article:
http://www.6moons.com/industryfeatures/distortion/distortion.html
Later today it will also be up on www.passlabs.com
http://www.6moons.com/industryfeatures/distortion/distortion.html
Later today it will also be up on www.passlabs.com
Nelson Pass' distortion tutorial
mod note: 2 threads merged into 1
Nelson Pass has a quite interesting tutorial about amplifier distortion in 6moons here:
http://www.6moons.com/industryfeatures/distortion/distortion_2.html
Recommended reading.
Jan Didden
mod note: 2 threads merged into 1
Nelson Pass has a quite interesting tutorial about amplifier distortion in 6moons here:
http://www.6moons.com/industryfeatures/distortion/distortion_2.html
Recommended reading.
Jan Didden
Very interesting piece.
Nfb is a lot like salt. A little can enhance, a lot can kill.
I also believe that a low piece count is the way to go. Now I guess fewer stages also makes sense.
I am of the opinion that for the most part, errors add together they rarely cancel each other out. Every piece in the chain has an error factor associated with it.
Thanks for the thought provoking article.
I sense there is more to come
Nfb is a lot like salt. A little can enhance, a lot can kill.
I also believe that a low piece count is the way to go. Now I guess fewer stages also makes sense.
I am of the opinion that for the most part, errors add together they rarely cancel each other out. Every piece in the chain has an error factor associated with it.
Thanks for the thought provoking article.
I sense there is more to come
First worthwhile thing I've read in a month of Sundays. Thanks.
One thing that I've always bounced back and forth in my mind is the idea that both the mathematics and real life measurements indicate different curves for the different gain devices. Trivial observation, right? Maybe, maybe not. Suppose you've got an all-bipolar circuit. It might then be argued that the harmonic content of the bipolars might have a--for lack of a better word--optimum feedback ratio. But that ratio might be different for other gain devices. Okay, still no revelation...after all tube gear (as tube lovers like myself have pointed out for years) gets by with far lower feedback and sounds as good or better than equivalent solid state pieces.
But...and I may not have set my thesis up properly; you'll have to look at this in just the right way to see what I'm trying to say...
What if you've got a hybrid piece, meaning not just tube/solid state, but FET/bipolar or bipolar/MOSFET or the like, and you find that things just won't gel? What if the bipolars need X amount of feedback, but the MOSFETs would really do a better job with less (or more)? The 'conflict' (again, I'm struggling to pick the right word here) between the competing requirements of the two or more kinds of gain devices might produce a piece that's neither fish nor fowl, as the expression goes.
Yes, it's possible to combine gain devices (e.g. Lovoltech power JFET cascoded with another device so as to flatten the curve) to produce something greater than the sum of its parts, but unless done intentionally at each and every stage of the circuit, you're in danger of producing an anti-synergy.
One possible solution is no feedback at all. This may or may not fly, depending on any number of other factors. Another would be local feedback loops. For example, if you're building a tube/solid state amplifier, run a feedback loop around the front end (presumed to be the tubed part) only, then another around the back of the amp (presumed solid state) so as to allow each class of device its optimum feedback. Yes, there are high-feedback tube pieces out there, but the predominant strategy is low feedback for tubes. The converse is true for solid state.
This line of thought echoes a point I was trying to make in a series of posts wherein Srajan Ebaen and I had differing views on hybrid circuits. He had just reviewed (with high praise) a tube front/MOSFET back end amp (something by Thorens?). While acknowledging that I hadn't heard the amp, I noted that there has never been a hybrid that was an unqualified, runaway success in the market. Does that, ipso facto[/] mean the circuit sucks? No, of course not. And maybe this one amp that he was so impressed with really is the Holy Grail. And maybe not. I still haven't heard it and can only say that I have no opinion.
One thing I said at the time was that I run a tube preamp with (mostly) solid state amps. It sounds good. But one of the more subtle points is that each chassis, by definition, contains its own feedback loop, which we will assume for the sake of argument is optimized for its sort of gain device.
Grey
P.S.: And, yes, those who are familiar with tubes might easily begin thinking that triodes and pentodes need to be treated differently in terms of feedback.
One thing that I've always bounced back and forth in my mind is the idea that both the mathematics and real life measurements indicate different curves for the different gain devices. Trivial observation, right? Maybe, maybe not. Suppose you've got an all-bipolar circuit. It might then be argued that the harmonic content of the bipolars might have a--for lack of a better word--optimum feedback ratio. But that ratio might be different for other gain devices. Okay, still no revelation...after all tube gear (as tube lovers like myself have pointed out for years) gets by with far lower feedback and sounds as good or better than equivalent solid state pieces.
But...and I may not have set my thesis up properly; you'll have to look at this in just the right way to see what I'm trying to say...
What if you've got a hybrid piece, meaning not just tube/solid state, but FET/bipolar or bipolar/MOSFET or the like, and you find that things just won't gel? What if the bipolars need X amount of feedback, but the MOSFETs would really do a better job with less (or more)? The 'conflict' (again, I'm struggling to pick the right word here) between the competing requirements of the two or more kinds of gain devices might produce a piece that's neither fish nor fowl, as the expression goes.
Yes, it's possible to combine gain devices (e.g. Lovoltech power JFET cascoded with another device so as to flatten the curve) to produce something greater than the sum of its parts, but unless done intentionally at each and every stage of the circuit, you're in danger of producing an anti-synergy.
One possible solution is no feedback at all. This may or may not fly, depending on any number of other factors. Another would be local feedback loops. For example, if you're building a tube/solid state amplifier, run a feedback loop around the front end (presumed to be the tubed part) only, then another around the back of the amp (presumed solid state) so as to allow each class of device its optimum feedback. Yes, there are high-feedback tube pieces out there, but the predominant strategy is low feedback for tubes. The converse is true for solid state.
This line of thought echoes a point I was trying to make in a series of posts wherein Srajan Ebaen and I had differing views on hybrid circuits. He had just reviewed (with high praise) a tube front/MOSFET back end amp (something by Thorens?). While acknowledging that I hadn't heard the amp, I noted that there has never been a hybrid that was an unqualified, runaway success in the market. Does that, ipso facto[/] mean the circuit sucks? No, of course not. And maybe this one amp that he was so impressed with really is the Holy Grail. And maybe not. I still haven't heard it and can only say that I have no opinion.
One thing I said at the time was that I run a tube preamp with (mostly) solid state amps. It sounds good. But one of the more subtle points is that each chassis, by definition, contains its own feedback loop, which we will assume for the sake of argument is optimized for its sort of gain device.
Grey
P.S.: And, yes, those who are familiar with tubes might easily begin thinking that triodes and pentodes need to be treated differently in terms of feedback.
leapcat said:Question
Would a fast amp lower the Im distortion of a complex
signal?
No. The stuff I was talking about is somewhat independent of the
bandwidth of the amp. Of course bandwidth limitations would limit
the ability to pour on feedback, so of course it becomes relevant in
that context.
GRollins said:
Absolutely, and of course that means different techniques to maximize
gain and minimize distortion. Alas, some of those ideas remain on the
table.
I thought about it, but then it occurred to me - "Screw it. I'll use these
to make a million dollars!"
Muuuuuhahahahahahahahaaaa........
Hi Nelson,
Thanks for publishing that article.
One quibble, if I may. You state, "If you see a curve with distortion levels climbing as the output goes down, it implies crossover distortion caused by the gap between the two push-pull gain elements. This implies high-order harmonics."
For the often seen THD+N measurement, and for a very low-distortion amplifier, this kind of measurement can be due to decreasing S/N as the signal level decreases. A guy named Samuel Groner did a nice paper with lots of op-amp distortion measurements. That paper is here (warning: big download). In figure 2.4 (page 17 on the right) he shows the residual distortion of the AP System 1. You can see the increasing residual distortion as the signal level decreases. That's why I hate THD+N. Some people will point to such measurements and say they're due to crossover distortion of the unit under test. Of course, amps with high, but monotonic distortion, will look fine in this test because the distortion is always well above the noise. Amps with very low and monotonic distortion may falsely show indications of non-monotonic distortion in many cases.
Thanks for publishing that article.
One quibble, if I may. You state, "If you see a curve with distortion levels climbing as the output goes down, it implies crossover distortion caused by the gap between the two push-pull gain elements. This implies high-order harmonics."
For the often seen THD+N measurement, and for a very low-distortion amplifier, this kind of measurement can be due to decreasing S/N as the signal level decreases. A guy named Samuel Groner did a nice paper with lots of op-amp distortion measurements. That paper is here (warning: big download). In figure 2.4 (page 17 on the right) he shows the residual distortion of the AP System 1. You can see the increasing residual distortion as the signal level decreases. That's why I hate THD+N. Some people will point to such measurements and say they're due to crossover distortion of the unit under test. Of course, amps with high, but monotonic distortion, will look fine in this test because the distortion is always well above the noise. Amps with very low and monotonic distortion may falsely show indications of non-monotonic distortion in many cases.
lumanauw said:
No special method - IM is just the same old nonlinearity looked at with
multiple tones.
And yes, I looked at Fig 13 for quite a while before deciding it was correct.
Keep in mind that some of this is amplitude changes to the original frequencies
as well as the addition of new tones.
andy_c said:
Absolutely correct. At one point I added a sentence mentioning this as a
possible source of confusion, but edited it for brevity, the feeble excuse
being that I said distortion and not THD+N.
Fortunately (or perhaps unfortunately, the egregious examples of Class B
are pretty obvious in power amplifiers, where the distortion gets nasty well
above the noise level.
IM distortion
Well now I am really thinking. Is there a way to invert the IM distortion and use it as feed back.
If I understand the concepts behind John Broskie's Aikido topology correctly he developed a way to take the power supply noise, invert it and inject it at two points so it would cancel.
This concept sounds desirable. Can the concpet be applied to IM distortion?
Well now I am really thinking. Is there a way to invert the IM distortion and use it as feed back.
If I understand the concepts behind John Broskie's Aikido topology correctly he developed a way to take the power supply noise, invert it and inject it at two points so it would cancel.
This concept sounds desirable. Can the concpet be applied to IM distortion?