 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
75.168.123.245
In Reply to: RE: The better the design of the amp and its execution posted by 13th Duke of Wymbourne on November 19, 2024 at 10:37:51
are you thinking of advancement as providing more gain-bandwidth allowing more feedback?
Yes, among other things.
Its also possible to apply feedback in such a way that it is not distorted in the process of doing its job. Opamps are a good example of how that's done.
Follow Ups:
Its also possible to apply feedback in such a way that it is not distorted in the process of doing its job. Opamps are a good example of how that's done.
Could you elaborate a bit on this
Could you elaborate a bit on thisOpamps derive feedback from a Voltage divider network which mixes feedback with the incoming signal right at the input of the opamp.
Most audio amplifiers don't do this- instead apply the feedback to the cathode of the input tube, emitter of an input transistor (in the case of more primitive solid state amps) or to the gate or base of a semiconductor that is part of a differential input pair.
In all cases the device to which the feedback is applied is non-linear. So in the process of mixing with the audio signal, the feedback signal is distorted. This causes it to generate higher ordered harmonics and in some cases, IMD.
Resistors of course are a lot more linear.
Norman Crowhurst pointed out this problem 60 some years ago but didn't propose a solution. 15 years later Peter Baxandall also pointed it out but simply proposed 'more feedback' which really didn't work.
In either case if the feedback is wrapped around the amplifier as seen in opamp circuits, it will not be distorted so it can simply do its job.
Edits: 11/20/24 11/20/24 11/20/24
Thanks for the explanation, I can understand the logic behind it.
I would say that almost all solid state amplifiers follow the Lin topology (input differential Gm stage - transimpedance stage - output buffer/follower) which is, basically, the same as used by op amps. So, the feedback signal will connect to the input gate/base of the input diff pair just the same in audio power amps as in op amps.
I think this idea that feedback intermodulates with the signal in the amplifier is specious. As we know, the closed loop performance is given by A/(1+AB) and as A becomes larger and larger A effectively drops out of the equation and performance is defined by the feedback network, usually resistors. A portion of the output is subtracted from the input signal to create the error signal - this is what is amplified by A. The error signal is pre-distorted and amplified so that distortion at the output is minimized. IMO, it is a red herring to think of what is passing through the amplifier - yes, it is not 'pure' but it is exactly what is required to make the output 'pure'. The only thing we care about what goes on inside A is that the small signal gain remains high and there is no saturation in either the time domain or amplitude domain.
I don't know Norman Crowhurst's writings but I do know Baxandall's analysis of applying feedback to a pure second-order device. The way to think about what is happening is that the feedback is straightening the very curved transfer curve. Even if you have a transfer characteristic that looks straight visually it will still have some curvature and wiggles that equate to high order terms in the power series expansion of that characteristic and they create high(er) order distortion. You can play with this in MS Excel by plotting a straight line and doing a power series fit that will have only the linear term. Then add points to make subtle deviations to that line and the power series fit will then calculate the added high order terms. Only a perfectly straight line produces no distortion and there is no perfectly straight transfer characteristic. Just as there is no perfectly square law characteristic to start with.
In Baxandall's think-piece, second-order and total distortion decreases as feedback is increased but all other terms increase until you reach an inflection point (approx 20dB loop gain) and increasing feedback above that point then reduces all distortion components. This is explained in Bruno Putzey's 'F-Word' article. The worst thing would be to apply only 'a little' feedback because of the widely held view that 'a lot' is bad but puts you right at that inflection point where the higher harmonics are their worst! Nelson Pass has written of the Baxandall example that if the purely second harmonic is not objectionable then why apply any feedback if you need to apply a lot to lower the high order harmonics back to where they are with no feedback - though he also writes about using 'a little' feedback in his amplifiers? Maybe hedging his bets.
I'll leave the subjective impacts of distortion spectra aside because it's, er, subjective and I see that amplifiers of all different topologies are highly regarded so, maybe, non of this really matters?
I would say that almost all solid state amplifiers follow the Lin topology (input differential Gm stage - transimpedance stage - output buffer/follower) which is, basically, the same as used by op amps.Yes, many amplifiers are very similar to opamp design- nearly the same. But the way feedback is applied is different. In most power amps the differential input is used for the input signal on one side and the feedback is applied to the other input of that differential amplifier. So the feedback signal is mixed with the incoming signal within the differential stage itself.
In an opamp, the inverting input usually has both the audio and feedback applied to it; the gain of the opamp thus defined by the ratio of the input resistance (in series with the input signal) and the feedback resistor. If a single-ended connection, the non-inverting input is tied to ground. That seemingly small difference is important as the differential input circuit of a power amp isn't linear and the coupling between the devices is usually degenerated slightly to prevent distortion driving the next stage. By contrast, the resistor divider network of an opamp's feedback is considerably more linear!
So yes, I'm sure.
Edits: 11/21/24 11/21/24
IIUC, you are contrasting non-inverting feedback with inverting feedback. You can do either with an op amp or with a Lin-style audio power amp. I get the point you are making that for the inverting feedback case the input differential amplifier appears 'within' the feedback loop but appears 'outside' the feedback loop for the inverting feedback case. I say appears because there may be no significant difference as the feedback will drive the input differential signal to be the same for both cases with the same closed-loop gain.
Diff pairs, whether FET of BJT are the most linear amplifier configurations for small signals. If you apply a lot of feedback the input differential input should be tiny. Degenerating the diff pair is usually done to cope with transient overload of the error signal - the appendix in Putzeys' 'F-Word' article explains this really well.
Putzey's explanation is excellent.
But it also points to one of the reasons using the feedback internally isn't the best move. That's because for the feedback to do its best job, the other transistor of the differential pair has to be coupled as tightly to the transistor receiving the feedback as possible. It isn't due to the degeneration. This ignores of course the simple fact that the two devices aren't linear.
Best, IMO/IME to avoid that problem altogether and simply use both inputs to the differential pair as inputs to the amp- and then mix the feedback externally rather than internally. According to Baxandall this should help things out.
Of course we've not talked about the implications of insufficient Gain Bandwidth Product on distortion but Bruno covered that pretty well.
A test for the curious would be to take an op-amp and configure it in non-inverting mode with gain X and then re-configure to inverting mode, again with gain X, and compare the output spectra.
I still think, to a first order, the differential input voltage will be the same so the distortion will be the same but, to the second order, non-inverting does have a large common-mode on both inputs (i.e. the input signal swing) whereas the inverting effectively has no common-mode swing. So, CMRR could have an effect.
'Putzey's explanation is excellent' - I had another read through and agree very much and that is a good place to end.
Power amplifiers have internal feedback. Opamps do not. That is the point I'm making.
When the feedback is distorted by the feedback node, it can't do its job properly. The feedback isn't distorted nearly so much when mixed externally.
I don't disagree with you as such but I don't see insight in your position.
I hope my reasoning is laid out clearly enough so that any flaws are apparent. I don't have access to SPICE at present but I will simulate these cases hopefully in the not too distant future.
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
