We're using the above circuit to drive a sinusoid with constant current into a custom sensor which is best modeled as a series capacitance & resistance (~100ohm, 1uF). The op-amp is single supply, and biased to 2.5V. A 1kHz signal with fixed 500mV DC offset is driven into the circuit. We measure the voltage over the sensor in order to determine changes in its resistance. An additional sense resistor (precision 100ohm) is included in the feedback loop to better estimate the current through the sensor (voltage measurement circuitry omitted). A 5Mohm resistor is placed in parallel with the sensor to provide a DC path.
The problem we have is that a DC potential is developed across the sensor each time the circuit is powered up to take a measurement. This happens because the sensor sources the charge required to bias up the DC blocking capacitor. This potential eventually decays through the DC resistor in parallel with the sensor, but not quickly enough to avoid long term damage to the sensor (which is very sensitive to DC voltages).
We do not wish to reduce the size of the DC resistor because it will affect the accuracy of the measurements. At the moment we implement a clumsy solution whereby we temporarily short out the sensor with a digital multiplexer whilst the op-amp is powered on, and the circuit stabilizes to 2.5V. Does there exist a neater solution to quickly bias up the circuit to 2.5V whilst not developing any DC voltage across the sensor?