Any idea about integrated Image rejection mixer?

Now I design an image rejection mixer, conventional Gilbert cell mixer +poly phase filter the simulation results : conversion gain 6.4 dB, ip3 8.49dBm, ports to ports isolation is bigger than 40dB and the image rejection ratio is bigger than 50dB@all corner the power consumption 1.4mA*3.3V all satisfying except NF, the Noise Figure simulated more than 20dB including the ppf if I improve this parameter the current would increase, How to decrease the NF @ minimum power ,and I use the poly phase filter for image rejection ,because the results would vary with the RC mismatches,and the area really acount, any idea about image rejection? could someone please shed a sight and light me up?

Microwaver,

We need more information in order to be able to help you. Here's what ideally I'd like to see in order to make any judgement:
-Circuit Schematic with Voltages/Currents annotated
-Block-by block breakdown:
1) What is the NF, Gain of a single gilbert cell?
2) What is the NF, Power output of the LO amplifier?
3) What is the insertion loss of the PPF?
4) What is the power level of the input signal at which you are measuring NF?
-Are you doing SSB noise sims, small-signal or large signal? IF large-signal, what is your 1/f noise look like in the system?

In short, there are a lot of factors that may be killing your noise: everything from the bias networks to large signal noise mixing of 1/f. We really need to know more in order to help you.

Greg