How can I improve noise perforcmance in Gm-C filter?

[hsez]

Hi,
I have designed a 5th order Gm-C lowpass filter with CMOS0.25um Tech.
the cut-off frequency is 2MHz
The OTA structure is folded cascade with negative source degeneration using nmos channel resistor.The gm value is 10uA/V.The simulation result shows that the filter input noise is 4uV/sqrt(Hz)@300KHz and 2uV/sqrt(Hz)@1MHz.It's terrible.
The total filter current consumption is 3.4mA,1dB Comp. is -0.1dBm. These two performance are something good.
How can I improve the noise performance by comsuming more power?
Pls give me some tips, thanks a lot.

 

[rfsystem]

Power efficiency is defined by:

1. How much % of supply is your max signal swing
2. Does the filter topology and coefficents show frequency peaking which limit signal swing.
3. Is the bias current a factor greater than the max signal current or possible lower because of AB-type OTA

All what you can do is changing these three major parameters.

My personal experience is that Gm-C achieve not more than 50% swing related to supply. Folded cascodes increase bias current by 2. Linearity is difficult to control. Continous online tuning is a myth.

Look for active RC. There you can built an AB-type opamp getting typical 10% bias related to maximum peak current.

Gm-C only applies if the parasitic pole of the opamp have a technology related dependence which could not be effective compensated. That happens around 100MHz. Disk drive filters or wired comms uses Gm-C. Below 100MHz GM-C works but get easy outperformed by active RC.