I do understand what you're saying. But that's just damping. Does the damping effort of the amplifier affect what current flows through the feedback loop? Yes it certainly does, and the relationship is more complex than the usual inverting and non inverting gain equations that we use in design. And the better the damping is, the less these artifacts affect the final performance of the circuit.
And output impedance is dynamic and changes with frequency. It is not really a simple phenomenon to express mathematically.
Yes I do! It's the ONLY way to get rid of them beyond any intrinsic low output impedance (which is another word for damping). And without feedback you can only get rid of it so much - it's FB that lets you decrease it by the loop gain - often a factor 1000.
The amp or the FB don't know where the 'nasties' come from. Whether they come through the air, or as harmonic distortion from the output stage; if it appears at the output, and it is not what's in the input, FB will reduce it. And the only way to do that is by running it as an error signal back to the input stage and through the amp. Aagain. And that is PRECISELY where spectral growth comes from. I never said the spectral growth comes from external things.
I don't know what you mean by 'skewed by FB' but it I agree it is caused by the distortion of each stage. If you wrap FB across a perfectly linear amp it does not produce harmonics you also have no spectral growth. That is why spectral growth is not an issue in amps that are reasonably linear before the application of FB. Spectral growth DOES occur with even the slightest non-linear path of course, but it is then so low level that you don't see (and hear) it. To migitate you 1) design for best open loop linearity and 2) use as much FB as you can get away with, stability-wise.
Jan
Last edited:
Are you referring to the circuit referenced in this thread? There is lots of feedback; it is just not global.
Look at the circuit again. It is just a buffered op amp. Although the circuit is unity gain, the op amp provides plenty of loop gain.
Read my post above. We are in agreement. I apologize, because I misjudged your proficiency level. I do not really know you, and thought your question was from someone with an intermediate understanding of circuits. But in fact it was a good question and you explained it well. Please share more.
I always try to help people get up the next rung of the ladder, but I suspect that you could probably teach me something.
Several circuits referenced in this thread, so not sure. But was of the misconfusion that you were talking about zero feedback around the output. Now that's cleared I understand where you are coming from.
Not that output stages with no feedback around them are bad, I've been nursing one for 17 years now. Just I know it could be a lot better!
Jan would be too modest to mention it, but he's chairman of the AES for northern Europe, publishes one of the top audio technical journals, and has written quite a few papers on the theory and application of feedback. I've been reading his articles since I was a kid.
He said that on purpose - in fact I think he only made this post so he could say that 😉
Jan
'Back emf' is part of what produces the speaker impedance; you can't avoid processing that unless you want pure current drive - which will give a very lumpy frequency response with almost all speakers. RF can be filtered away.Fast Eddie D said:
Simply untrue. As I said, in some cases the main reason for the feedback is to reduce output impedance so the amp can better control the speaker. Putting "period" after an untruth does not make it true.
If the speaker impedance is non-linear then this can contribute to spectral growth, although it is likely that most will come from the amp itself.
Did I say it was simple? Can maths only deal with simple things? Fortunately most audio amps are sufficiently linear that a quasi-linear approximation can work, but if not there is always a Volterra series to fall back on.
nonlinear
Quasi-linear works fine, and although not automatically leading to insights, fine-enough-mesh simulators and good device models work well with nonlinear systems, absent "catastrophes".
Anecdote: when I was in the doghouse having taken the responsibility for a failure mode mostly due to substandard parts, I was testing another consultant's amplifier with a new driver for a featherweight subwoofer. At some level the system erupted in a hideous cacophony with all kinds of nonharmonic energy unrelated in any obvious way to the input signal. Of course the consensus was it had to be electronic in origin.
But it persisted when the little amplifier was replaced by a lab amp with reliable and well-documented performance. The problem was the new driver. The person who designed it had cut some interesting corners and the device had built-in problems setting in at large excursions. The situation had the makings of a good "laboratory" unit in association with a course on nonlinear systems.
Yes. The sad thing is most of the hand-waving and hypothesizing is done tacitly assuming quasi-linear. If we start to really dig in to nonlinear systems things get hairy in a hurry. But doing so, I can see where we'll just unleash a herd of terms that will be taken up by people and cited to justify mostly-pseudoscience.
Quasi-linear works fine, and although not automatically leading to insights, fine-enough-mesh simulators and good device models work well with nonlinear systems, absent "catastrophes".
Anecdote: when I was in the doghouse having taken the responsibility for a failure mode mostly due to substandard parts, I was testing another consultant's amplifier with a new driver for a featherweight subwoofer. At some level the system erupted in a hideous cacophony with all kinds of nonharmonic energy unrelated in any obvious way to the input signal. Of course the consensus was it had to be electronic in origin.
But it persisted when the little amplifier was replaced by a lab amp with reliable and well-documented performance. The problem was the new driver. The person who designed it had cut some interesting corners and the device had built-in problems setting in at large excursions. The situation had the makings of a good "laboratory" unit in association with a course on nonlinear systems.
RF anecdote
Some friends had stopped listening to records (pre-CD era here) because a guy down the street using an illegal "linear" for his citizen's band activities would come through loud and clear over their hifi system, the latter with some mid-fi receiver. Complaints lodged with him and others did not help. The friends found the vapidity of his carrying-on to be the worst part. If today the one of them were to eavesdrop on the typical cell conversation I can only imagine her horror.
I presumed, foolishly, that the detector of the RF was in the phono input, and borrowed the receiver to fit some filtering in at those inputs.
It helped...not at all. The signal was common-mode RF efficiently picked up by the speaker cables and detected somewhere in the power amp section, probably at the base-emitter junctions of the input stage. The remedy would have been good common-mode chokes in the speaker lines, perhaps along with some judicious bypassing of the junctions, but I think we all just gave up at that point.
Some friends had stopped listening to records (pre-CD era here) because a guy down the street using an illegal "linear" for his citizen's band activities would come through loud and clear over their hifi system, the latter with some mid-fi receiver. Complaints lodged with him and others did not help. The friends found the vapidity of his carrying-on to be the worst part. If today the one of them were to eavesdrop on the typical cell conversation I can only imagine her horror.
I presumed, foolishly, that the detector of the RF was in the phono input, and borrowed the receiver to fit some filtering in at those inputs.
It helped...not at all. The signal was common-mode RF efficiently picked up by the speaker cables and detected somewhere in the power amp section, probably at the base-emitter junctions of the input stage. The remedy would have been good common-mode chokes in the speaker lines, perhaps along with some judicious bypassing of the junctions, but I think we all just gave up at that point.
So signals entering the amplifier from behind can be amplified and played through the speakers even if no such signal was present at the input. How does that correlate with feedback theory..?
If you read my post above, then you will realize that we agree.
If you read my post above, then you will realize that we agree.
Also realize that I was trying to explain something to someone that I thought had a misconception, and I was mistaken. There is a hierarchy of understanding this stuff, and I try to gear my explanations to what I perceive is a person's proficiency level. And I was mistaken about their proficiency level.
It most certainly can, because non-linear speaker impedance can cause transistors to "move around" on their curve, which will certainly introduce nonlinearity. This non-linearity will be dealt with (theoretically) by the feedback network, because its source is inside of the feedback loop.
And that's what I meant by "complex relationship." It can be explained and hopefully better than I explained it.
And that's a point I was trying to make. By following some reasonable guidelines we can sometimes design a circuit with simplified algebra. In fact, I always shoot for that.
The feedback loop goes right to the non inverting input, or to the emitter of the input transistor. It transmits the signal from the output to the - input.
You'd be surprised how many amplifiers have gain at RF frequency. Most of mine would if I didn't snub the input and output, which alleviates this situation.