Hi all,
I'm working on a low power downconv(0-IF) Gilbert Mixer at 400 Mhz. Tail current source in the bottom, RF diff pair, 4 switching transistors and resistive load. There is a shunt cap at the load to 1st order filter out the higher freq comp. The vdd is 1V so I do not have much headroom and max power consumption is half mW. My question is about switching pair. I used minimum L (180nm) and biased them at 600 mV. LO is a sine and has +-400 mV swing and AC coupled to switches. The problem is voltage drop across switches. In order to reduce the ON resistance and voltage drop across them I used very big W which is 400 um to reduce V drop and increase 1/f NF performance. But still more than 600 mV drop across it. Which causes only 67mV over RF transistors and 115 mV across tail current source.
The questions are
1) What is the maximum limit for W of these switches?
2) How can I reduce voltage drop of SWs further?
3) Do I need to bias switches at around 600mV so when L0 swings (from .6-.4=.2 to .6+.4=1) they will become ON and OFF? (if I increase bias V, the voltage drop reduces (which is what I want!!) but in this case transistors may not became OFF which may cause operation problem.  right???)
-3b) (extension of 3rd question) if SWs not become OFF completely (chopping vs. multiplying) what kind of consequences will happen?
4) Linearity of mixer is directly related to RF pair. If I increase L of them (let's say 500nm-600nm) is this a good practice in terms of linearity?

There are many things embeded in this questions for me. Any expert replies will be very enlightening.
Thanks so much again.