Sorry, you lost me there: what is this theory that only works on a linear amplifier and therefore must be discarded? It can't be feedback theory because that is specifically intended for non-linear amplifiers. And any delay will be small and is the same thing as a phase shift which varies linearly with frequency.
Tubes vs FETs and Bipolar
Found about triode: Distortion without feedback is complex, resembling that of the FET amplifer with feedback, but distortion products are distinctly lower in level that in the no-feedback FET amplifer at the same input level, or
even the no-feedback BJT amp at its much lower level.
I think the reason why some amplifiers sound "harsh" is the generation of higher harmonics (5th, 7th and so on) with a quite high level, compared to second and third harmonics.
Found about triode: Distortion without feedback is complex, resembling that of the FET amplifer with feedback, but distortion products are distinctly lower in level that in the no-feedback FET amplifer at the same input level, or
even the no-feedback BJT amp at its much lower level.
I think the reason why some amplifiers sound "harsh" is the generation of higher harmonics (5th, 7th and so on) with a quite high level, compared to second and third harmonics.
Hi Fireworks !
Last comparative test was between a Technics SU-V8 (very good, no crossover distortion, but lots of feedback) and an Electrovoice EV1177 (germanium outputs, interstage transformers, low feedback). Seems an unfair match, 110x2 against 18x2, 0.00000... THD against 1% THD. Same souce (CD) and same speakers (Wharfedale).
Main difference is in the attacks (mostly with jazz and rock), I can understand words more with low feedback, get more 3Dimensional effect and... it's much more like being there !
Had the same comments by several friends, to whom I made the same test, just to make sure I was not biased ...
Ciao !
Orit
Yes that's a good paper (and why I referenced it just above your post
) There are two things that amplifiers have that many people vehemently deny however:1) Non-linearity caused by non linear devices
2) Delay of a signal from input to output (related to slew rate)
I'm not aware of any others but I suppose there may be, but those two alone will cause trouble that can be measured as THD and IMD, and measured by the ear as flatness etc.
These two items cause feedback to be a difficult subject as 1) and 2) vary so much from amp to amp, therefore one level/situation does not fit all.
In this Wikipedia article:
Tube sound - Wikipedia, the free encyclopedia
it is said that negative feedback causes inter-modulation distortions but the references are unreachable.
Can someone give me some serious references that explain this phenomenon and maybe model it, etc. ?
I am curious if this type of distortion is the reason some audiophiles reject the circuits with negative feedback. Maybe they are onto something ?
Any comments and ideas are appreciated.
Hi fireworks et all,
following the thread,and its twistings,I'm thinking that,what you really wanted to ask is, what kind of amplifying devices are more "design-able",as far as, feedback is concerned,in order to minimize the artefacts.
Correct me if I'm in an error.
Regards
B.L
I don't know.
But first I'd be interested in what their experience would be without the benefit of knowing they were listening to a solid state amp.
se
There is as far as I know no measurable delay in the feedback circuit, but it does take a measurable amount of time for a movement on the amp input to create a movement on the output. This is your delay (t).
This means that trying to correct something already wrong on the output by feeding in a new signal to the input is not going to work perfectly because the correction will be delayed by (t). At different frequencies this will be seen as different phase distortions - worst of course in an amplifier with many gain stages at high frequency.
What you are largely referring to here is the work done by Baxandall that I mentioned at the beginning of this thread. It is true, both in theory and practice, that, starting with no negative feedback at all, adding negative feedback around a device will create some new harmonics if they were not there in the first place because of the re-mixing of the fed back distorted signal with the original signal.
However, you have to put in the numbers and put it in perspective. Moreover, just throwing up your hands and saying it is a nonlinear system to which linear approximations should be thrown out is not helpful. Also bear in mind that SPICE simulation handles nonlinear systems quite accurately, so all of these assertions are readily testable by SPICE simulation.
First, keep in mind that real amplifiers have multiple stages and start out with harmonics of many orders to begin with. This is especially true of distortions contributed by the output stage in the form of crossover distortion.
Secondly, it has been shown that virtually ALL kinds of feedback placed around a nonlinearity count toward getting to the point where the feedback decreases all orders of distortion rather than increasing any of them. Most interestingly, emitter degeneration counts as well. And yes, starting with a stage with only one order of distortion (e.g., second), even adding emitter degeneration initially adds/creates higher orders of distortion.
This second point also reinforces the notion that one should start with a rather linear open-loop amplifier in the first place, since getting that open loop amplifier linear most likely involved the use of local negative feedback (even if only degeneration) that counts toward the amount of total feedback needed to get to the point where further increases in negative feedback always decrease distortion of all orders.
All of these issues and theories apply equally to tubes and transistors.
Cheers,
Bob
What is the value-range of (t), how did you measure it?
What is the effect of a certain (t) on the overall distortion (e.g. THD), how big in numbers is the change in THD, or whatever measure you think is appropriate to represent the effect.
What are "different phase distortions"?
Of course its not perfect - it might never be - but how is a certain high feedback corrected amp - with lower THD values (%) than a certain no feedback amp - worse than said no feedback amp.
What do I have to measure to see this, if THD won't do the trick?
Hi,
an amplifier does not oscillate if the phase margin (into a resistive load) is significantly greater than 1degree.
It is also stable if the phase margin still remains +ve when a reactive load is connected.
But look at the response to a filtered squarewave that has frequencies limited to within the passband of the amp. Low phase margin results in severe overshoot and ringing.
Medium phase margin results in severe overshoot and highly damped ringing.
It's not until the phase margin approaches 80degrees that the overshoot is well controlled. Is this the same as saying that the highest frequencies in the output are >100degrees out of phase with the same frequencies at the input?
Any less a phase margin and the amp is adding something to the filtered test signal.
Could the difference in phase margin into reactive and resistive loadings have an influence on what the filtered test signal sounds like? I think the literature on this supports the audibility of gross differences in phase margin.
I think this "Phase Margin" thing goes some way towards explaining what Kra was asking
an amplifier does not oscillate if the phase margin (into a resistive load) is significantly greater than 1degree.
It is also stable if the phase margin still remains +ve when a reactive load is connected.
But look at the response to a filtered squarewave that has frequencies limited to within the passband of the amp. Low phase margin results in severe overshoot and ringing.
Medium phase margin results in severe overshoot and highly damped ringing.
It's not until the phase margin approaches 80degrees that the overshoot is well controlled. Is this the same as saying that the highest frequencies in the output are >100degrees out of phase with the same frequencies at the input?
Any less a phase margin and the amp is adding something to the filtered test signal.
Could the difference in phase margin into reactive and resistive loadings have an influence on what the filtered test signal sounds like? I think the literature on this supports the audibility of gross differences in phase margin.
I think this "Phase Margin" thing goes some way towards explaining what Kra was asking
Last edited:
Please can we stop all this loose talk about time delays affecting feedback? Consider a typical SS amp, with a Cdom around the VAS. Lets assume that the open-loop bandwidth is 100Hz. This gives a time delay of about 1.6ms at low frequencies. Sounds alarming, doesn't it? But at low frequencies this is just a fraction of the period so does no harm. What about high frequencies? Well, here the Cdom just gives a 90 degree phase shift and very little delay, it is just an integrator. The feedback can cope with a 90 deg phase shift, so no problem.
The real issue with this architecture is not time delay, but the integrator. This integrator in the forward path means that the residual distortion gets differentiated, which may make it more unpleasant to our ears as it reduces the high-order components less than it reduces low-order components. If you want to solve a problem, first identify the right problem! If the forward path is reasonably linear then the feedback has less to do so all the problems become smaller.
The real issue with this architecture is not time delay, but the integrator. This integrator in the forward path means that the residual distortion gets differentiated, which may make it more unpleasant to our ears as it reduces the high-order components less than it reduces low-order components. If you want to solve a problem, first identify the right problem! If the forward path is reasonably linear then the feedback has less to do so all the problems become smaller.
t is a variable and will vary hugely depending upon the amplifier.
THD is an irrelevant and outdated measure that is crazy. THD effectively weights the 2nd harmonic with the same importance as the 7th harmonic, whereas the ears rate the 2nd as pleasant and the 7th as awful. As ultimately we use our ears we must ignore THD, as it's useless.
If I played you an amp with 5% 2nd harmonic compared to an amp with 2.5% made up of 5th, 7th and 9th you'd think the one with twice the THD as far far better
TIM is a much better measure of what is going on in an amp, but not infallible.
I think we also need a weighted AHD or Audible Harmonic Distortion measure, that tell you how much distortion you will actually hear.
No
...because the time delay is the causal factor of the phase distortion. You tell me that phase distortion increases with frequency, I say to you that it is a fixed time delay. I prefer to use a single number rather than a formula - it's just preference and I prefer to think of the causal factors - not the overlaid theory.
Please can we stop all this loose talk about zero delay amplifiers?
Do you know amps that display such awfull level of THD ?...
In most amps , high range harmonics are lower than
low order ones...
That said, we want neither of these THD , be it even order...
No
...because the time delay is the causal factor of the phase distortion. You tell me that phase distortion increases with frequency, I say to you that it is a fixed time delay. I prefer to use a single number rather than a formula - it's just preference and I prefer to think of the causal factors - not the overlaid theory.
Please can we stop all this loose talk about zero delay amplifiers?
Yes it is a delay that leads to phase distortion, by whatever name you call it.
But nobody ever showed that even large phase distortion can or is audible; the tests I know are done with carefully controlled single-tone tests and even then it takes lots of phaseshift to notice it.
I agree with you on the THD measurements; there are many proposals to apply an order weighting factor to traditional THD measurements which somehow haven't been adopted. Probably because a single THD number is so conveniently measured (and understood).
jd