Hi,

IMO the analysis looks very wishy washy. Its a little surprising because i've read some of Ali Niknejad's previous papers and they were very good. Perhaps i just need to spend more time reading it...

Correct me if i'm wrong, but it looks like equation (2) doesn't agree with equation (1). i.e. if you take the noise of the transconductor, 4kTγg_do, you should be able to just multiply it by the conversion gain to get something like,

16kTγg_d0/π²*|Z_f(fout)|²Δf

Not so sure if equation (3) is correct either. It seems to suggest that the noise of the switches is zero is Z_gm' is infinte. However, if Z_gm' is infinite, then you can treat each half of the double-balanced mixer as something like a differential pair with a transimpedance amplifier at the output. No reason why it should have zero noise contribution...

Anyway just a last comment, it seems like the switches would have to be pretty enormous in order for this to be a current-mode mixer. In which case, the maximum operating freqeuncy would be limited by the transistor average g_d0/C_GS while operating in the triode region with VDS near zero. This is apparently around 2.5 GHz.

I'll look over it again and see if I have to take back any of my comments  :P

cheers,
Aaron

EDIT:
after looking more carefully, I guess (1) and (2) do agree due to the noise folding. I guess the assumption is, however, that the transconductor high freqeuncy noise is not filtered...