I am designing a sum-product circuit (based on gilbert-cell). I have 2 input current vectors X and Y. Z is the output current vector. All the MOS transistors are operated in weak inversion (subthreshold). The paper which I am following says that monte-carlo simulations should be run to fix the width of the transistors. Can anyone explain me how to go about this problem?
I don't know which paper you are reffering.
But as per my knowlege the monte carlo analysis is performed to find the yeild analysis, which gives the yeild for the of design for process variation.
Yes, what you said is correct. But I have to tradeoff between area vs variations in o/p wrt transistor mismatch. As the standard deviation is inversely proportional to square root of (WL).