raja.cedt wrote on Feb 29th, 2012, 4:22am:no..i guess what he mean is Auto zeroing, or ultimately you are using 1-Z^-1.
You can do it continuous time -- we used to call them servo loops -- basically a low pass filter in the feedback in a loop. But then I believe you still have the problem of generating a low freq pole. For 0.05 Hz, using a 10 pF cap you need to synthesize a 318 G-ohm resistor. A simple gm/c filter would use about 1 pA of current so that won't work! I'm not sure if a simple switched cap would work - I thought you needed an opamp to make those work very well.
It seems to me that the area of an opamp is going to be small compared to the capacitances that you need.
Like Raj, I was also thinking some double sampling method that would give you a (1-z^-1) transfer would be the easiest (i.e. sample twice, subtract the result). In practice this could be done by sampling the input and then subtracting the sampled value from the continuous time input before the amplification (which is essentially what Jerry is suggesting but without an explicit feedback circuit).
I've never done this, but one simple thing that comes to mind is to AC couple the signal into the IA. Periodically "zero" the input of the IA by shorting the IA inputs to a reference. This will sample the input voltage on the AC coupling caps. Then, when you open the shorting switches, the IA will be amplifying the difference, giving you the (1-z^-1) characteristic. The roll-off will be determined by the sampling rate, not the cap. However, the caps will need to be big so that leakage doesn't drain them since your sample time will be longer than 20 seconds.
Seems tough. If common mode leakage currents don't kill you, mismatches in leakage currents probably will. It is better to do it in the digital domain if you can (and also remove the low freq effects of the IA)...
Let us know what works for you.
rg