Amplifier Servo Problems

[Duo]

After some mindful observations of servo circuits in audio amplifiers I and a friend have noted some interesting problems.

First: Since a good servo integrates the output of the amplifier to generate an offset signal, class D came to mind. My thought was if you have a signal that has a duty cycle over 50%, the servo will integrate that into an offset voltage which will thus output bad DC levels into your speaker or other load.

I know this doesn't happen that much in music, but I have observed it. Anyway, we put a servo under test to see what would happen. This servo was the exact design taken from national semiconductor's AN1192 application note which includes the use of their servo on the LM3886 in bridge/parallel configuration.

On a square wave input at many different frequencies, we were able to observe the predicted output voltage by varying the duty cycle. The output voltages in the circuit approached over 5V output for some levels.

An other noted problem is low frequency distortion. Since the servo integrates, any low frequency operation will be distorted somewhat. I don't think this enough to seriously affect audio with the design in question, however, who knows... Maybe some serious listening to some test circuits must ensue...

I put this set of observations forth to see if anyone else here has observed these symptoms of servo use; and if so, what is their opinion?

[peranders]

I think you have missed the point a bit. A servo works as any feedbacked system. If the opamp has distortion, the speaker output will also have distortion. The thing is to choose an opamp with low distortion at least up to 1 kHz. If you have a squarewave with anything less than 50% duty cycle you'll have DC in the signal and the servo tries to remove the DC. Nothing odd by that.

1 The servo must be good in the frequency bands where it's active.

2 The servo must of course not be overdriven (= distortion)

[Duo]

I understand what you're saying about feedback circuits and the fact that they are servos by nature.

However, the external integrating servo seems to have some less than desirable characteristics in operation than the normal full range servo application which maintains the gain of the opamp in the first place.

EDIT: I'm not trying to say that they're bad in general; just that there are a couple of things that pique my interest as far as operational charactistics go.

[Cobra2]

Originally posted by peranders

1 The servo must be good in the frequency bands where it's active.

I have seen many servos based on the cheapest/oldest op-amps you can find...
Do you mean there will be a sonic improvement for upgrading the servo - op-amp?

Arne K

[tiltedhalo]

Servo's suck no if and or but's about it. They are designed for many different applications from controlling voice coil travel on subwoofers to distortion cancellation on amps, to DC offsets and many others. Bottom line is they not only correct things you want them to correct, but things you dont want them to correct also.
Case in point LM's servo's for the overture chips yep they correct that offset by taking it off the output, putting it out of phase and applying it back to the front end, and boom badda bing no more offset, also frequency, range, dynamics, sound stage all down the crapper
Because for all the good, alot of the good appears bad to a servo

[peranders]

Originally posted by Cobra2


I have seen many servos based on the cheapest/oldest op-amps you can find...
Do you mean there will be a sonic improvement for upgrading the servo - op-amp?

Arne K

Possibly! If you take two opamps and put them into LTSpice, use the same values as the AN-1192 app note, and observe how much which is fed back from the servo you'll also see how important a good servo is. Bear in mind than the "gain" is max 0.1, meaning 1/10 of the signal goes back into the main amp.

If you check my servos, I have also a filter which removes the rest of the servo signal, higher up in the audio band.

According to me making a good servo is no problem only if you let it work within it's limits.

[peranders]

tiltedhalo, a servo is equally good as the opamp used and how good are opamps these days? Pretty good I'd say. Have you examined a servo theoreticly and practically?

According to me they add very very lttle.

[Eva]

Duo :

Congratulations, you have discovered that output filters are not linear and that common op-amps are not very good at integrating several dozen volts and several hundred kilohertz square waves.

Next time that you try a DC servo on a class-D amplifier, take the feedback *after* the output filter and apply a really low pole to it in order to get further ripple attenuation. It will work fine this way.

[Duo]

Um... You jumped to the conclusion that I put my class D servos before the lowpass filter. no, I don't... I use nested feedback and global feedback, but the nested bit is taken inside the linear stages and not around the switching stage.

[peranders]

I missed also that the whole thing was about a class D amp but still I see no theorectial problem with using a DC servo as long as you'll keep HF noise out of it.