DC servo simulation for low offset

[rtarbell]

I am simulating a basic three stage audio amplifier in the full version of TopSpice, mainly to experiment with using a DC servo in the feedback loop to nullify the output offset.

The DC servo itself is a basic opamp in the integrator configuration (input R at 10k, feedback C of 100n). My current problem is this: When I simulate the circuit to get a DC bias point using an ideal opamp (a VCVS with gain = 999,999), the output offset voltage is in the uV (great!). When I replace the ideal VCVS with a spice model LF411 or LTC1151 opamp, the output bias point rails either high or low. Now I understand that the opamp models have important features and parasitics modeled into them, but I wouldn't *think* that it such parameters would make such a big difference (I was hoping to get an offset of at least a few mV, not -99V!).

==> Any ideas as to what I could be doing wrong or overlooking?

[darkfenriz]

Op-amp input phase reversal at overdriven input maybe...
Antiparallel diodes at input of an op-amp do the trick.

[analog_guy]

Originally posted by rtarbell
I am simulating a basic three stage audio amplifier in the full version of TopSpice, mainly to experiment with using a DC servo in the feedback loop to nullify the output offset.

The DC servo itself is a basic opamp in the integrator configuration (input R at 10k, feedback C of 100n). My current problem is this: When I simulate the circuit to get a DC bias point using an ideal opamp (a VCVS with gain = 999,999), the output offset voltage is in the uV (great!). When I replace the ideal VCVS with a spice model LF411 or LTC1151 opamp, the output bias point rails either high or low. Now I understand that the opamp models have important features and parasitics modeled into them, but I wouldn't *think* that it such parameters would make such a big difference (I was hoping to get an offset of at least a few mV, not -99V!).

==> Any ideas as to what I could be doing wrong or overlooking?

Try grounding the + input of the opamp and placing 0.1 uf between Vout and the - input. Then connect a 1meg resistor between the output of the amplifier and the - input of the opamp. Finally, connect the output of the opamp to a ~100k resistor to the input of your amp. Depending on whether the amp is inverting or not, you will need to select the proper location in the input diff amp. The 100k resistor acts as a voltage divider. Since the amount of Voffset is typically less than 1 volt at the amp's output, it does not take much voltage to correct it. Also observe the voltage at the opamp's output. It should not be near either rail.

The original RC values you specified (10K, 100 nf) yield a time constant of 6.3e-3, which is well within the audio range. You want the time constant of the servo to be well below 1/20 Hz (50 ms) . Otherwise the servo amp will act as a lowpass filter and remove audio content. When running SPICE, remember to simulate for a long enough time to observe the RC time constant of the servo amp. For the values I gave, this is on the order of 1 second.

[rtarbell]

==> When running SPICE, remember to simulate for a long enough time to observe the RC time constant of the servo amp. For the values I gave, this is on the order of 1 second.



Aha! So you are saying that I should run a transient simulation to determine if the dc servo is setting the right bias point, not just an .OP simulation?

[analog_guy]

Originally posted by rtarbell
==> When running SPICE, remember to simulate for a long enough time to observe the RC time constant of the servo amp. For the values I gave, this is on the order of 1 second.



Aha! So you are saying that I should run a transient simulation to determine if the dc servo is setting the right bias point, not just an .OP simulation?

Right. You should see the outut of the amp slowly converge to near zero V DC with a 1st order RC behavior. It is lso easier to simulate the servo functon if there is no signal drivin into the amp.