# Common mode RC filter of High-BW differential line

I'm using a fully differential analogue chain, using 20Hz HP, 3MHz LP and x2 Gain. My signal of interest is fully differential with a 200KHz fast modulation and a 100Hz sawtooth ramp. The high frequency chirp but may change within the range of 50KHz to 1MHz. I've therefore ensured a flat pass band (0dB) from 100Hz to 1MHz.

The differential structure will naturally attenuate CM signals, however the industrial environment the system will be working in will contain a number of high amplitude common mode signals. I've therefore decided to use common mode filtering capacitors ($C_{cm}$) on the input differential line, right next to the input termination resistor.

What I need to do, is provide further CM only filtering of 50Hz, 100Hz 1KHz etc but not interfere with my differential signal path.

Unfortunately, adding suitable common mode capacitors $$F_{cm} = \frac{1}{2\pi R_1C_{cm}}$$ acts to change my 3MHz low pass differential response. If the differential filter is $$F_{diff} = \frac{1}{2\pi(R_1+R_1)(C_{diff}+C_{cm}/2)}$$ I cannot get common mode filtering as far below the differential as I'd like, as $C_{cm}$ acts to swamp $C_{diff}$ in the $F_{diff}$ equation.

How should I go about common mode rejection of lets say 50Hz mains hum, while keeping my differential mode 100Hz to 1MHz pass band?

• I'm confused, aren't you going to take the difference (V(+) - V(-)) somewhere in software? That should get rid of the common mode. Or are you expecting lots of CM signal and are worried about dynamic range issues? – George Herold Jan 7 '15 at 17:12
• The ADC is differential so yes CM issues should be removed, however the system is going to operate in a high voltage, high current industrial environment, so yes to the dynamic range issues. – user2286899 Jan 8 '15 at 12:06
• OK you've got a very wide bandwidth, so all I can think of are notch filters at 50,100, 150 Hz? That or better shielding. I can't think of a way to filter common mode, and leave the differential, without taking the difference of the signal. – George Herold Jan 8 '15 at 14:31
• Mind that the capacitors have bad accuracy, meaning you screw up the CMRR by the two caps to GND. Only use one between the line to ensure symmetry – Gregory Kornblum May 23 '17 at 11:46