need help on wideband if filter design

[sharpsheep]

I'm now designing on if filter of fc=50M BW=20M. Gm-C architecture is used in my design. We all know the center frequency is depend on the value of gm/c. To save current of the filter, gm should not be chose too large. So the value of C is chose to less than 1pF. The question is a variation of 1%(about 10fF) of C will bring 0.5M variation of center frequency while the gm is constant. If we consider the variation of gm and parasitic capacitance , the fc will be affected seriouly.

Is there any good method to solve this big problem? Pls.

 

[kspalla]

please share the complete circuit and let us re-evaluate the critical part of it.
May be we can go for an alternative circuit.