Below is a PSN message from Chris Chapman on adjusting the components for a sensor with a different resonant frequency.
If f is the resonant frequency of the sensor, you make 1 / (2 Pi f C R1) = 1.09 according to Roberts. I am not sure exactly why he chose this. It could explain the small hump in the combined characteristic.
The Max gain is R2 / R1
The lower turnover frequency is 1 / (2 Pi C1 R2) This has been set to ~1/2 the lowest response frequency in the Roberts circuit.
You need two units in series.
Roberts reported significant levels of VLF noise, so it might be a good idea to follow up the compensation circuits with a two / four pole high pass Butterworth filter set to about the lower desired frequency.
If you want to use the circuit at long periods, you may need to use some CAZ opamps MAX430, LTC1150 etc. They reject 1/f noise. Otherwise LF411, OP07, LF412 etc are OK
The optimum place to insert this circuit is between the first opamp and the low pass filters.
How about making an Experimenter Board plug in version first?