What I was talking about at the beginning of the thread was this:
Here is a plot of three signals,
dark blue = input
magenta = feedback (not inverted for the sake of clarity) (has approx 37 degrees phase lag)
yellow = summing junction -> amplifier actual input.
Notice the yellow is =still=a=sinewave= despite being the result of a subtraction of *non-corresponding* parts of the two waveforms, perhaps giving a clue as to why an amplifier might deliver a great sinewave but sound lousy on music.
Here is a plot of three signals,
dark blue = input
magenta = feedback (not inverted for the sake of clarity) (has approx 37 degrees phase lag)
yellow = summing junction -> amplifier actual input.
Notice the yellow is =still=a=sinewave= despite being the result of a subtraction of *non-corresponding* parts of the two waveforms, perhaps giving a clue as to why an amplifier might deliver a great sinewave but sound lousy on music.
Now the situation where we have an input signal that has 15% 3rd harmonic mixed with in it. Same phase lag between input and output.
Notice that the yellow signal that is fed to the amplifier as before, looks nothing like either the input or the output. What will this signal look like when *it* exits the amplifier???
Notice that the yellow signal that is fed to the amplifier as before, looks nothing like either the input or the output. What will this signal look like when *it* exits the amplifier???
lumanauw:
If you want 'fast' feedback then you need a fast amplifier. In a standard three stage design you can accomplish this by using very high speed transistors in the differential and lots of degeneration, and faster than normal output devices. This shifts their poles up, away from the one contributed by the (compensated) Vas. This allows you to reduce the size of the Miller capacitor (or even do away with it altogether in extreme cases), increasing open-loop bandwidth and thus reducing the delay. Make the OL bandwidth wide enough and you can have virtually no phase shift over the audio band and beyond.
This gives you an amp with very high slew rate, but the distortion will be worse. That may be what you're looking for, but generally it's not the best way of going about an audio amp.
Alternatively you can use an inherently fast, low gain topology, such as the folded cascode. This will yield similar results, although it's a more sensible way.
If you want 'fast' feedback then you need a fast amplifier. In a standard three stage design you can accomplish this by using very high speed transistors in the differential and lots of degeneration, and faster than normal output devices. This shifts their poles up, away from the one contributed by the (compensated) Vas. This allows you to reduce the size of the Miller capacitor (or even do away with it altogether in extreme cases), increasing open-loop bandwidth and thus reducing the delay. Make the OL bandwidth wide enough and you can have virtually no phase shift over the audio band and beyond.
This gives you an amp with very high slew rate, but the distortion will be worse. That may be what you're looking for, but generally it's not the best way of going about an audio amp.
Alternatively you can use an inherently fast, low gain topology, such as the folded cascode. This will yield similar results, although it's a more sensible way.
I see what you mean now. You're right, the output might look very different from the input, but only because some frequencies are phase shifted. Mathemematically the output will still be very close to the input. You can analyse it with the component sine waves and get identical results. A more complex input signal in no way alters the way feedback works compared to a simple signal. Luckily our ears are not sensitive to that sort of phase shift. If they were then we would not be able to recognise the sounds coming out of our speakers.Circlotron said:
Intrigued by the idea of extending Miller compensation to encompass the output stage, I did some simulations. It requires rather fast output devices (MOSFETs in the simulated circuit).
It appears to have a huge effect on THD contributed by the output stage, decreasing all harmonics by 20-50dB (yep, that's a lot) in the particular circuit I simulated. I think I'm going to have to implement this in the next amp I build to see if reality agrees.
It appears to have a huge effect on THD contributed by the output stage, decreasing all harmonics by 20-50dB (yep, that's a lot) in the particular circuit I simulated. I think I'm going to have to implement this in the next amp I build to see if reality agrees.
Circlotron said:
I'm with you, Circlotron. My experience is that feedback most harms subtle, delicate low-level signals. But, hey, nice sine wave.
phase_accurate said:
Perhaps that's true, but I'll take a phase distorting speaker reproducing a non-fed-back output any day over a phase distorting speaker reproducing a fed-back output.
1.) I don't think that the low-order distortion from a speaker is less audible than high order distortion from an amp because the first one is coming from a SPEAKER and the latter one from an aAMP. The main reason for the effect is that one is LOW order HD and the other one HIGH order. There are also amps out there that produce mainly low order distortion BTW.
2.) Since the temporal distortion of an amp is by some orders of magnitude less severe (and also much less complex) than the temporal distortion generated by a multiway speaker, one can assume that it is also less audible.
I don't think that NFB per se is bad. It depends on HOW it is done (as always !).
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.