OK Scott, thanks. I will have to try the library, because although I have a lot of telecomms books, this exact subject isn't covered. If I want to pursue this otherwise I will have to spend a few days practising algebra, I know there are a few wrinkles I've forgotten. I was kind of hoping there would be someone who is everyday fluent in this stuff, although obviously it's a lot to ask for somebody to concentrate long enough to mark my homework. I'm pretty sure the inverse transform contains an infinite series, being IMD, perhaps I should start by trying to write out the answer and then work backwards.
I beg to differ. The feedback is the quality control mechanism, and if it doesn't work better than other parts of the circuit then the end result is compromised - it will effectively inject very nasty, "non-linear" distortion into the mix - the very reason that those very high feedback units, some decades ago, probably earned such a reputation for unpleasant sound ...
In simulations of amplifier behaviour, when the unit is being pushed to the edge of its "correct" functioning it's easy to see how the feedback actually makes things worse, by trying to force the topology to work "better" than it's capable of doing, in an attempt to correct the output.
a light summary paper https://web.archive.org/web/2004072.../akustikk/meetings/DAFx99/schattschneider.pdf
Distortion Analysis of Analog Integrated Circuits - Springer I don't own it but have looked at a library copy
Distortion Analysis of Analog Integrated Circuits - Springer I don't own it but have looked at a library copy
I'm afraid it's true, Scott ... it only takes a bit closer scrutiny of what simulations of circuits in real world environments show, to see where the weaknesses are. Sometimes engineers only consider what occurs within a fixed, stable outer setting, and miss the bigger picture - a classic example is that they scrutinise the non-linearity of the circuit OR the supply rejection capability, and ignore non-linearity of the circuit AND supply rejection behaviour.
Last edited:
This is just what you get when you use feedback to polish a turd. In order for feedback to work properly, there needs to be excess loop gain and excess bandwidth available, so the circuit can quickly react to its own errors.
Put another way, any well designed piece of audio equipment should not be "pushed to the edge of its correct functioning" in the process of reproducing music. Those horrible old solid-state power amps were not well designed by today's standards. Calling them "high feedback" does a disservice to feedback, they probably had less effective feedback than a good modern design under the conditions where it is needed.
Executive summary: If you try to use feedback to polish a turd, don't blame feedback when it goes wrong.
Actually, their problem was insufficient feedback, as you implied. However, it is difficult for people brimming with typing energy and self-confidence, but with close to zero engineering knowledge and no experience designing, measuring, or building to grasp this simple fact. So the urban legends persist.
Exactly. The feedback mechanisms, and everything that interacts with them, have to be the best performing parts of the circuit - slightly 'incorrect' feedback most likely will be worse than none at all, especially in the treble. Everything in circuitry conspires against the treble frequencies being handled as competently as elsewhere in the audible spectrum, and this is so obviously the problem still when listening to most audio systems.
Again I think it is unfair to blame feedback for this. The fundamental problem in amplifier design is not that the feedback is somehow inaccurate, but that the output stage is both nonlinear and relatively slow. The slow speed of the output stage means that you can't apply as much feedback as you would like at treble frequencies, or it becomes unstable.
Since the era of the nasty "high feedback" amps, output stages have got at least 10x faster thanks to MOSFETs and ring emitter BJTs. A modern design can safely run even more feedback at 20kHz.
I've also noticed that the treble is the first area to show problems. For me it is cymbals and hi-hats that start to sound unnatural. The last time I noticed this, the problem turned out to be old and tired fabric dome tweeters.
Oh and merry Christmas
Since the era of the nasty "high feedback" amps, output stages have got at least 10x faster thanks to MOSFETs and ring emitter BJTs. A modern design can safely run even more feedback at 20kHz.
I've also noticed that the treble is the first area to show problems. For me it is cymbals and hi-hats that start to sound unnatural. The last time I noticed this, the problem turned out to be old and tired fabric dome tweeters.
Oh and merry Christmas
Last edited:
Not "blaming" feedback here - I said "Everything in circuitry", not "Everything in FB circuitry" - but rather the overall design which is too simplistic in the thinking. An interesting exercise is simulating an ordinary type circuit, which performs quite well with ideal voltage rails - but then replace them with real world supplies having to deal with the load requirements: the distortion figures collapse by literally orders of magnitude, because the feedback has to deal both with the modulating supplies, and the non-linearities.
Speakers have generally behaved well - the worst that has happened for me, using very ordinary drivers, is that the voice coils have apparently overheated, and I get some rubbing noises.
Best wishes for the season, to all ...
Speakers have generally behaved well - the worst that has happened for me, using very ordinary drivers, is that the voice coils have apparently overheated, and I get some rubbing noises.
Best wishes for the season, to all ...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.