Why do Boyk and Sussman do the analysis at constant input instead of constant output?
They do put the poor tube under a bad constraint- extremely poorly chosen operating conditions. Under better ones, the distortion performance at these input levels is remarkable, though gain is certainly limited compared to undegenerated BJT or FET.
They do put the poor tube under a bad constraint- extremely poorly chosen operating conditions. Under better ones, the distortion performance at these input levels is remarkable, though gain is certainly limited compared to undegenerated BJT or FET.
Nelson Pass said:
Thanks Jan. I was remembering the wrong paper.
I do have an issue with their conclusions about degeneration in Fets, as my
own measurements with both Jfets and Mosfets do not show this behavior
with degeneration. I did not see enough information on the simulation they
did to decide if there is some other factor. I do think it's important to separate
out the effects due to Vgs/Ids versus Vds/Ids.
Nelson,
Your work is quite interesting as it confirms some of Baxandall's stuff. I remember that in another thread Bob Cordell repeated some of Baxandall's tests as well. If I take all this in account, including Boyk & Sussmann, noting that many of the curves are different from each other, about the only conclusion you can draw is that yes, under specific circumstances, depending on the device and circuit and biasing details, it is possible that increasing the nfb factor can lead to increasing (higher order) distortion. In particular, most work has been done on single devices and not on complete amps.
Furthermore, who is to say what is worse: 10% 2nd and 3% 3rd on the one hand, or 1% 2nd, 0.3% 3rd, 0.1 % 4th and 0.03%5th on the other hand?
You can't say that, as a rule, increasing nfb in an amp does increase (audible) distortion of the amp.
Jan Didden
Then I'm really surprised by the level of distortion you get. I would not have thought that a simple follower produces already several percent THD (Class B). I imagine that part of the reason is that you probably used mosfets.
The bipolar output stages in D. Self's book perform much better (again ClassB).
Have fun, Hannes
h_a said:
First off, we understand that Self's idea of Class B is what we think
of a Class AB.
I presume that you are looking at fig 6, since fig 5 has nothing above
1%.
Fig 6 shows a Mosfet output stage, but fear not, I can make a
bipolar stage do that trick also.
Lastly, all this stuff is by way of illustration. You can argue that
a gain stage can be made that has better than 1% harmonic
figures, and I would agree, although there is plenty of stuff out
there that isn't that good.
Hi Nelson,
sorry for the confusion. I just realized I erronously thought the dB-values were already relative to the amplitude of the fundamental. In that case the -40dB peak would already mean 1% distortion.
Thanks a lot for the clarification! I thought you skipped that for sake of brevity in the article. So it's indeed 50% duty cycle.
Have fun, Hannes
sorry for the confusion. I just realized I erronously thought the dB-values were already relative to the amplitude of the fundamental. In that case the -40dB peak would already mean 1% distortion.
Thanks a lot for the clarification! I thought you skipped that for sake of brevity in the article. So it's indeed 50% duty cycle.
Have fun, Hannes
Great article and discussion. A few comments and questions-
On the use of feedback - for a simple single gain stage, does increasing feedback, in general always increase the order of the non linearity function? If so, it would seem like you could mathematically predict distortion order based on amount of feedback, at least for some range of circuits. Yes? No?
There has been work done by Earl Geddes on the perception of distortion that is interesting. His conclusions are based on masking of the harmonics, and indeed also support conclusions that high order harmonics are more audible than low order ones. From his paper -
"The masking effect of the human ear will
tend to make higher order nonlinearities
more audible than lower order ones.
• Nonlinear by-products that increase with
level can be completely masked if the order
of the nonlinearity is low.
• Nonlinearities that occur at low signal levels
will be more audible than those that occur at
higher signal levels."
http://www.gedlee.com/downloads/Distortion_AES_I.pdf
This would support the general feeling by many that low order distortion, even at higher levels, is less offensive than high orders at lower levels. And that distortion at low signal levels (from say crossover effects) is also bad.
Bob
On the use of feedback - for a simple single gain stage, does increasing feedback, in general always increase the order of the non linearity function? If so, it would seem like you could mathematically predict distortion order based on amount of feedback, at least for some range of circuits. Yes? No?
There has been work done by Earl Geddes on the perception of distortion that is interesting. His conclusions are based on masking of the harmonics, and indeed also support conclusions that high order harmonics are more audible than low order ones. From his paper -
"The masking effect of the human ear will
tend to make higher order nonlinearities
more audible than lower order ones.
• Nonlinear by-products that increase with
level can be completely masked if the order
of the nonlinearity is low.
• Nonlinearities that occur at low signal levels
will be more audible than those that occur at
higher signal levels."
http://www.gedlee.com/downloads/Distortion_AES_I.pdf
This would support the general feeling by many that low order distortion, even at higher levels, is less offensive than high orders at lower levels. And that distortion at low signal levels (from say crossover effects) is also bad.
Bob
It's pretty clear that (loop) negative feedback increases the order
(complexity)of distortion, and yes, I think you can in principle
predict the distortion in an amplifier.
I pretty much agree with Geddes, but my major point was that we
can easily create high order components even with simple low
order type circuits by virtue of:
a) cascading circuits
b) negative feedback
c) complex signal
It's difficult to get rid of (c).
(complexity)of distortion, and yes, I think you can in principle
predict the distortion in an amplifier.
I pretty much agree with Geddes, but my major point was that we
can easily create high order components even with simple low
order type circuits by virtue of:
a) cascading circuits
b) negative feedback
c) complex signal
It's difficult to get rid of (c).
A policeman came across a drunk in the middle of the night crawling
on all fours beneath a street light. "I lost my car keys", explains the
drunk. "Did you lose them here?", asks the cop. "No", says the drunk,
I lost them down the block, but the light's a lot better here."
So it is with using single tone distortion analysis. It's easy to do, and
it's possible to get astonishingly good performance on this test with
an amplifier, but it is a simplified representation of the problem.
on all fours beneath a street light. "I lost my car keys", explains the
drunk. "Did you lose them here?", asks the cop. "No", says the drunk,
I lost them down the block, but the light's a lot better here."
So it is with using single tone distortion analysis. It's easy to do, and
it's possible to get astonishingly good performance on this test with
an amplifier, but it is a simplified representation of the problem.
Golden design?
So reading a lot of all, building the Hiraga (yes, not Nelsons baby...) I would think, that the ZenV9 must be the best amp in the world...
Single stage, low feedback, only second order harmonics...
If I would build a bridged version, leave out the coupling capacitor: This would be the "final" amp to praise Nelson??? (And cancel 2. harmonics...)
Thinking a step further: Use Nelsons V9 design, replace the Jfet with a triode, replace the Fets with pentodes and run it at 5kV...
Nelson like electrostatic drive? Maybe a cascoded Toshiba jfet / IRF610 input to get enough gain...
Would this spoil the purity of the ZenV9 design????
I'm thinking of such a project... Anybody thought of this (you, Nelson??? ;-))
Dreaming of it....
Dirk
So reading a lot of all, building the Hiraga (yes, not Nelsons baby...) I would think, that the ZenV9 must be the best amp in the world...
Single stage, low feedback, only second order harmonics...
If I would build a bridged version, leave out the coupling capacitor: This would be the "final" amp to praise Nelson??? (And cancel 2. harmonics...)
Thinking a step further: Use Nelsons V9 design, replace the Jfet with a triode, replace the Fets with pentodes and run it at 5kV...
Nelson like electrostatic drive? Maybe a cascoded Toshiba jfet / IRF610 input to get enough gain...
Would this spoil the purity of the ZenV9 design????
I'm thinking of such a project... Anybody thought of this (you, Nelson??? ;-))
Dreaming of it....
Dirk
Professor smith said:
Hello Professor,
If I may be a bit cynical: When you consider that people enjoy tube amps, solid state, hybrid, low fb, high fb, error correction, current dumping, class A, AB, B, D, H, G etc, each selecting their preference over the others, and that for at least the past 40 years, what would *you* think?
Jan Didden
Nelson Pass said:
I pretty much agree with Geddes, but my major point was that we
can easily create high order components even with simple low
order type circuits by virtue of:
a) cascading circuits
b) negative feedback
c) complex signal
It's difficult to get rid of (c).
I'm thinking about (c), as this is independent of the circuit design, unlike a) and b). I guess maybe the point is the more complex the signal, the more likely you'll hit the stops (limits of fairly linear mode of operation) randomly with 6 sigma peaks. And those peaks will show the worst case nonlinearities of the circuit. This is the old peak to rms ratio expressed another way, which is NOT nice and tame like one or two pure tones is.
Yes, well the real fact is that we don't know all that much about
hearing. Steven Dear (the bat guy) has been pointing me in a lot
of interesting directions, and it continues to be be very fascinating,
but mostly the conclusion I have to draw is that we don't really
even know what we don't know.
The ear is not really a microphone, and the neural networks which
process the feed from the ear do not really resemble any analog or
digital system we are familiar with.
We vaguely understand that high order distortions are more
audible, but we are operating in the realm of "ironic" engineering,
to steal a phrase from John Horgan, in which we are speculating
without either much in the way of facts or even decent theories
which suggest good experiments. Sort of like literary criticism.
Ironic engineering come in the flavors "naive" and "sophisticated",
the former having strong beliefs in the "truth" of things like
harmonic distortion analysis, and the latter more sceptical, maybe
having listened to audio with single ended tube amps and such.
So does the drunk really appreciate the limitations of his search
method? Maybe he does, but likes working where the light is.
hearing. Steven Dear (the bat guy) has been pointing me in a lot
of interesting directions, and it continues to be be very fascinating,
but mostly the conclusion I have to draw is that we don't really
even know what we don't know.
The ear is not really a microphone, and the neural networks which
process the feed from the ear do not really resemble any analog or
digital system we are familiar with.
We vaguely understand that high order distortions are more
audible, but we are operating in the realm of "ironic" engineering,
to steal a phrase from John Horgan, in which we are speculating
without either much in the way of facts or even decent theories
which suggest good experiments. Sort of like literary criticism.
Ironic engineering come in the flavors "naive" and "sophisticated",
the former having strong beliefs in the "truth" of things like
harmonic distortion analysis, and the latter more sceptical, maybe
having listened to audio with single ended tube amps and such.
So does the drunk really appreciate the limitations of his search
method? Maybe he does, but likes working where the light is.
