Agreed, but what are the pros and cons of the alternative injection points of the DC servo correction current? At the minimum, it may either be injected into the input (non-inverting) node or the feedback (inverting) node. The DC resistance to ground at these nodes might vary greatly, and therefore the pull-range (referenced to the output) might also vary greatly, depending on the injection point.
It appears (from my preliminary LTSpice sims) that the stability and sonics also vary quite a bit, depending on the injection point.
Hi,
If injecting into the inverting input (or non-inverting input) of the signal circuit, any non-ideal behavior of the servo will be amplified like the signal, in fact, it becomes an (unwanted) part of the signal. The Feedback loop will do nothing to attenuate any (AC) output of the servo.
If we inject the servo output elsewhere in the circuit (within the NFB Loop) the feedback loop (if one is used of course) will attenuate any any non-idea behaviour of the servo.
If we (for example) inject the servo signal into the output of the VAS in a typical Amp and with optimised pull range by selecting a suitable value series resistor for the servo we should have minimised any impact, as there will be no AC gain whatsoever at this node and we get the full effect of the negative feedback to correct any non-idea behaviour of the servo.
Of course, non of this says anything about audibility and so those who wish to delude themselves thusly may still believe that all servo's are inaudible.
Ciao T
If injecting into the inverting input (or non-inverting input) of the signal circuit, any non-ideal behavior of the servo will be amplified like the signal, in fact, it becomes an (unwanted) part of the signal. The Feedback loop will do nothing to attenuate any (AC) output of the servo.
If we inject the servo output elsewhere in the circuit (within the NFB Loop) the feedback loop (if one is used of course) will attenuate any any non-idea behaviour of the servo.
If we (for example) inject the servo signal into the output of the VAS in a typical Amp and with optimised pull range by selecting a suitable value series resistor for the servo we should have minimised any impact, as there will be no AC gain whatsoever at this node and we get the full effect of the negative feedback to correct any non-idea behaviour of the servo.
Of course, non of this says anything about audibility and so those who wish to delude themselves thusly may still believe that all servo's are inaudible.
Ciao T
Sy,
If you inject the Servo output into the feedback node without adding more lowpass filtering (which increases the rolloff order and incidentally can cause enough phaseshift to induce the system oscillate at very low frequencies) the DC and AC gain from the output of the servo to the output of the Amplifier are the same.
If of course you elect a different node inside the feedback loop things may differ).
Finally, considering the effort and cost in implementing a servo, I would normally suggest to add J-Fet input buffers (source followers) to any bipolar signal Op-Amp (if such must absolutely be employed) to eliminate the input bias current as source of offset and to use separate DC and AC feedback path with a modest value (1uF or thereabout) film type dc-blocking/feedback coupling cap.
This entirely omits the need for a servo AND keeps the DC blocking capacitor inside the AC feedback loop, unlike the common approach to a DC blocking capacitor in series with the feedback divider (which usually needs to be a large value electrolytic capacitor), which, frequent and widespread use non-withstanding, strikes me a severely subideal and not as something a careful engineer should like to employ.
Ciao T
If you inject the Servo output into the feedback node without adding more lowpass filtering (which increases the rolloff order and incidentally can cause enough phaseshift to induce the system oscillate at very low frequencies) the DC and AC gain from the output of the servo to the output of the Amplifier are the same.
If of course you elect a different node inside the feedback loop things may differ).
Finally, considering the effort and cost in implementing a servo, I would normally suggest to add J-Fet input buffers (source followers) to any bipolar signal Op-Amp (if such must absolutely be employed) to eliminate the input bias current as source of offset and to use separate DC and AC feedback path with a modest value (1uF or thereabout) film type dc-blocking/feedback coupling cap.
This entirely omits the need for a servo AND keeps the DC blocking capacitor inside the AC feedback loop, unlike the common approach to a DC blocking capacitor in series with the feedback divider (which usually needs to be a large value electrolytic capacitor), which, frequent and widespread use non-withstanding, strikes me a severely subideal and not as something a careful engineer should like to employ.
Ciao T
Hi Sy,
I actually do not do that (for obvious reasons discussed above). I actually always consider Servo's like Isaac Asimov's Salvor Hardin considered violence and only use them as a last resort, preferring instead to deal with the actual underlying problem.
However this very method is still very commonly seen in all sorts of designs.
And judging by comments made above some seem to feel that this method (inject servo output into feedback node) is actually a good idea and defend it by declaring any other method to be "a bunch of bull".
So it seems necessary to remind people of this. Well, I guess I could just recommend to read Bob's book, he has a good section on Servo's from the objectivist viewpoint.
Ciao T
I actually do not do that (for obvious reasons discussed above). I actually always consider Servo's like Isaac Asimov's Salvor Hardin considered violence and only use them as a last resort, preferring instead to deal with the actual underlying problem.
However this very method is still very commonly seen in all sorts of designs.
And judging by comments made above some seem to feel that this method (inject servo output into feedback node) is actually a good idea and defend it by declaring any other method to be "a bunch of bull".
So it seems necessary to remind people of this. Well, I guess I could just recommend to read Bob's book, he has a good section on Servo's from the objectivist viewpoint.
Ciao T
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.