Abstract
Receiver design noise considerations for FET front ends for fiber-optics amplifiers as presented by Smith and Personick are revisited. The MOSFET noise model used is simple yet more accurate than that used previously. The device equivalent circuit for noise is derived from first principles. We are thus able to optimize the amplifier sensitivity with respect to hot-carrier channel thermal noise in terms of the FET drain-to-source voltage and the effective channel length. The importance of drain-source overlap capacitance in determining amplifier sensitivity, which has hitherto not been formulated, is also quantified. It is thus concluded that in spite of hot-carrier noise effects, fine-line NMOS amplifiers designed for gigabit-rate applications will continue to see sensitivity improvement for effective channel lengths down into the quarter-micrometer range. Investigation in the subquarter-micrometer range is in progress.