70mV signal and 20V common mode voltage demands the use of a very very precise differential amplifier, not showing any temp and long term drift of common mode rejection. With 1% toleranced voltage dividers about 40dB common mode rejection is the best you can achieve, resulting in about 200mV common mode noise at the input of ADC!

Also, if the transducer isn't symmetrically working, the source impedances of signal and reference voltage of transducer will differ. Then, the differential amplifier cannot work symmetrically and the comon mode rejection will heavily suffer. In such designs the common mode rejection isn't better than 20...30dB usually.

Filtering will only help, if the signal bandwidth is way below mains frequency (50/60Hz).

So, I wouldn't recommend the differential amplifier solution.

Another option is to use a IL300, to provide a "true" galvanic isolation of analog signals. Or you could build a simple delta sigma modulator and use a fast opto-coupler (6N137, PC410) to transmit the digital pulse train. The easiest way to achieve isolation, though, would be using a transducer which already provides a digital output.

Kai Klaas