DC servo with RC filter question

Member
Joined 2006
Paid Member
I am now designing a simple CFA mosfet amplifier that needs a DC servo.

The servo action is injected in the NFB node.

With the servo output unfiltered I am noticing a spike near 3.5Hz in the OLG plot. This spike is accompanied by a phase inversion.

Filtering the output of the servo with an RC (R19 20k and C6 47u) I can greatly reduce the spike and reduce it's frequency (now small peak @ 1.4hz) and phase is not severely inverted.

Increasing the value of this capacitor gives even better results (As an example see the plot with C6 1000uF).

My question is:
Does the RC filter in the output of the servo reduce it's speed ?
How high can I go with the filtering capacitor ?
Is there a drawback in using such servo ?

1st image - Amplifier circuit
2nd image - Servo circuit
3rd image - OLG plot without filtered servo
4th image - OLG plot with 47u filter
5th image - OLG plot with 1000u filter
 

Attachments

  • CFA amplifier.JPG
    CFA amplifier.JPG
    135.9 KB · Views: 254
  • CFA amplifier SERVO.JPG
    CFA amplifier SERVO.JPG
    51.8 KB · Views: 243
  • OLG without servo filter.JPG
    OLG without servo filter.JPG
    216 KB · Views: 219
  • OLG with servo filter.JPG
    OLG with servo filter.JPG
    223.4 KB · Views: 194
  • OLG with servo 1000uF filter.JPG
    OLG with servo 1000uF filter.JPG
    223 KB · Views: 217
If I connect C6 to GND instead of to GNDI the spike is much much worse.

I am using R18 10ohm as input gnd lift - a very effective trick to avoid hum in my builds.

But my most important doubt is... how large can I go with the servo C6 cap ?
 
You will not get hum in the simulation.
You have to short between GNDI and GND with wire in the simulation.

PS: even with lifted ground, the DC impedance between GNDI and GND is Zero in the real world. You break ground loop, even it is not a loop any more, they are still connected with each other, imagine it.
 
Last edited:
From my experience there are nothing but benefits in using this arrangement. Larger caps will go lower but once you get a couple of decades below 20 Hz there's not much point pushing your luck. I've had best results with FET op-amps (the ubiquitous TL071) as other types sometimes fail to lock the loop. The LF411 would thus appear to be a good choice.
 
I suggest you simulate the loop gain of the DC servo portion. More than one low pass poles in the system, you have to make sure you are not making a “phase shift oscillator”. Check the phase margin and gain margin of the DC servo.