Hi All,

I am trying to make a very low cutoff frequency filter (~ 1 Hz), and the choice is between using a gmC variant based on an active stage biased with very low current, and a MOS transistor biased for very large resistance with a capacitor. The whole idea of course is to make it with a small cap (~100 pF).

The gmC approach seems to work OK but introduces some offset, and I was thinking of using a MOS as a switch to get offset-free behavior.

However, the MOS obviously acts as a half-wave rectifier with this large cap when there are large out-of-band signals. This worsens the out-of-band suppression limiting it to about -40 dB which would be enough still, but this also introduces a large offset which is the consequence of the rectification process. It is the latter which is the problem.

Could someone suggest what one may do to minimize this large rectification-induced-offset? Or is this a fundamental limitation that I cannot get around?

Otherwise, I could make the MOS-C filter with far lower area than the gmC filter. For better linearity, I am controlling the gate drive to track the input signal.

Thanks,

Vivek

Hi Vivek,

MOS-C I think is the best approach for very low cut-off frequency filters. In order to minimize the distortion the MOS might introduce you could put several MOS devices in series to achieve a very large RON. In this way you reduce the effective swing accross each of them and therefore reducing the overall distortion.
There are also other techniques which control the gate voltage to cancel out the introduced distortion. There is an old paper describing these techniques:

"Continuous-Time MOSFET-C Filters in VLSI" Y. TSIVIDIS, IEEE JOURNAL OF SOLID-STATECIRCUITS, FEB 1986.

Regards
Tosei