I would again like to talk about servos in general.
Some hi end manufacturers still avoid them. Many mid fi manufacturers don't use them, because coupling caps are cheaper to implement.
Still others have tried them, and found them audible in their circuits.
What to do?
Well, first, cap coupling is easiest. Then, slight DC offsets don't matter much, and even with differences in temperature will not make much difference. The caps can also protect adjacent stages in case one active device fails. The drawback is that the caps often have non-linear distortion at low frequencies, and they virtually all have a significant amount of dielectric absorption. This is a 'linear' distortion that doesn't show up as non-linearity, but it can change the signal path to be greater than 1% deviation from ideal, in each capacitor. Is this bad? Well, it's not hi fi.
Servos can be problematic if they behave too much like a capacitor in rolling off the bass. Sometimes, we want a low frequency cutoff in order to remove excess garbage below 20 Hz or so. I have found that using a servo to create this rolloff is not a good idea. It is actually better to use a series cap and a load resistor. With a good cap, especially polypropylene or better, this beats a servo.
However, IF a broadband response is desired, then a servo can be implemented that will only really do much with any offset below 1 Hz. This is important, so that the servo does not actively change the audio signal, especially asymmetrical components. A good, at least linear integration cap should be used, and the output of the servo amp attenuated at least 20 times, (100 is better) before it is put back into the circuit to correct the offset. This means that a 2V offset from the integrator IC will only give 0.1V of correction, for example. This keeps the servo noise and distortion out.
Using these suggestions, servos have a much better chance of being OK in a high fidelity design.
Some hi end manufacturers still avoid them. Many mid fi manufacturers don't use them, because coupling caps are cheaper to implement.
Still others have tried them, and found them audible in their circuits.
What to do?
Well, first, cap coupling is easiest. Then, slight DC offsets don't matter much, and even with differences in temperature will not make much difference. The caps can also protect adjacent stages in case one active device fails. The drawback is that the caps often have non-linear distortion at low frequencies, and they virtually all have a significant amount of dielectric absorption. This is a 'linear' distortion that doesn't show up as non-linearity, but it can change the signal path to be greater than 1% deviation from ideal, in each capacitor. Is this bad? Well, it's not hi fi.
Servos can be problematic if they behave too much like a capacitor in rolling off the bass. Sometimes, we want a low frequency cutoff in order to remove excess garbage below 20 Hz or so. I have found that using a servo to create this rolloff is not a good idea. It is actually better to use a series cap and a load resistor. With a good cap, especially polypropylene or better, this beats a servo.
However, IF a broadband response is desired, then a servo can be implemented that will only really do much with any offset below 1 Hz. This is important, so that the servo does not actively change the audio signal, especially asymmetrical components. A good, at least linear integration cap should be used, and the output of the servo amp attenuated at least 20 times, (100 is better) before it is put back into the circuit to correct the offset. This means that a 2V offset from the integrator IC will only give 0.1V of correction, for example. This keeps the servo noise and distortion out.
Using these suggestions, servos have a much better chance of being OK in a high fidelity design.
Well, with each filter and integrator in series you increase lf phase shift. At some point you need to be sure it remains stable; a servo with lots of filter sections can easily lead to lf instability and oscillations.
jd
It's a regular nfb-loop, so all the nfb-stability rules apply
Gain margin, phase margin, etc.
jd
I promissed some information about a new measurement at Test Factory in Stuttgart.
They found that transistor amplifiers without NFB in the output stage and tube amplifiers can have an advantage when driving real loudspeakers. I found a paper ( in German, sorry) that discribes this measurement. You can download the paper under :
http://www.burosch.de/images/Display-workshop.pdf
Unfortunately i was not able to talk to Mr. Schüller, that came up with that measurement but i will try again.
They found that transistor amplifiers without NFB in the output stage and tube amplifiers can have an advantage when driving real loudspeakers. I found a paper ( in German, sorry) that discribes this measurement. You can download the paper under :
http://www.burosch.de/images/Display-workshop.pdf
Unfortunately i was not able to talk to Mr. Schüller, that came up with that measurement but i will try again.
As far as i understand the test they connect a power amp to a real loudspeaker. They put in an MLS signal and measure the impulse response of the poweramp - loudspeaker combination. Then they convert the impulse respose into the frequency domain. Power amps with no NFB around the output stage and tubeamps usually do better on this test. Power amps with high NFB usually have the impedance curve of the loudspeaker imprinted. They came up with this test because many low feedback amps and tube amps did very well in subjective listening tests although they had more harmonic and intermodulation distortion and a lower damping factor.
I was not able to talk to Mr.Schüller so far but i hope he can give some more elightenment. I visited Test Factory some time ago when they made such a measurement. Unfortunately i can not remember the exact test conditions but i will try to get more information. Amplifiers that did well where Aire, Pass, Brinkmann etc. all with no or little NFB around the output stage.
I was not able to talk to Mr.Schüller so far but i hope he can give some more elightenment. I visited Test Factory some time ago when they made such a measurement. Unfortunately i can not remember the exact test conditions but i will try to get more information. Amplifiers that did well where Aire, Pass, Brinkmann etc. all with no or little NFB around the output stage.
I've been trying to find out exactly what their results are. Is it that when Zout is larger, it is also flatter but provides less damping? Didn't we know that already?
Or is there an effect I'm missing?
Inquiring mind wants to know!
jd
......so Mr.Pass was right all along with STASIS.
Jam
I don't think STASIS has something to do with the fact of an output stage running outside the feedback loop. It's written somwhere in a white paper or so, the X'ed Pass poweramps (not XA...) have the feedback taken before the output stage. No STASIS there.
Just tons of bias current
Tino
- Status
- Not open for further replies.