Why not Pavel Dudek's feedback network.

[HAYK]

Pavel Dudek (RIP) has a circuit using a different kind of decoupling for DC unity gain using low non polarized capacitor. Why no one uses it?

 

[JonSnell Electronic]

It is more useful, to build an active filter rather than use a passive one like this.

 

[MarcelvdG]

It could be a lack of awareness or because of the larger sensitivity to input bias currents.

That is, suppose 1 uA of DC current is drawn out of the right side of each circuit. You then need to apply 500 mV at the input of the upper circuit and only 22 mV at the input of the lower circuit to get the output voltages back to 0 V. Hence, you have a larger transresistance from input bias currents to output offset voltages when you apply these circuits in the feedback network of a negative-feedback amplifier. Whether that is a problem or not depends on the input bias currents and output offset requirements.

 

[HAYK]

I don't understand what is your argument. I made piecwize model with 1M ohm and 1ua bias current. OK, I have 1v on each input but the offset on the output is 7uv. Of course the bias difference will provoke much higher output offset.

 

[MarcelvdG]

I don't understand what is your argument. I made piecwize model with 1M ohm and 1ua bias current. OK, I have 1v on each input but the offset on the output is 7uv. Of course the bias difference will provoke much higher output offset. View attachment 1184549

I was thinking of a circuit with DC coupled input and the circuit of post #1 in the feedback. You can indeed get some compensation by AC coupling the input, provided the bias currents match. That 7 uV is of course unrealistic, bias currents won't match within 7 ppm.

 

[EdGr]

This is a good question.

The bias offset voltages can be made the same by setting R3=22K and decreasing R2 to 500 ohms. C1 can still be much smaller (22uF) than C2.

I suspect the downside is that C1 is at signal level whereas C2 has one side grounded. A large capacitor is probably best connected to ground so that the can acts as a shield.
Ed

 

[BSST]

James Bongiorno used that bias network in the Sumo Nine. He used LF353 BiFet opamps, so bias current was low. Using OP's designators in post 1, he used R3=10M, C1=0.1uF

Service manual is attached, schematic at the end.