TABLE OF CONTENTS From the MIT Radiation Laboratory days of the 1940s until the successful development of high-frequency gallium arsenide MESFETs in the 1980s, low-noise receiver front ends used mixers. Considerable research on low-noise mixers was carried out in that period, culminating in the classic 1976 paper by Held and Kerr [9.1] describing the first practical noise model of a millimeter-wave mixer. Today, mixer noise modeling, although generalized, is performed in much the same manner as described by Held and Kerr.
Although the theoretical aspects of mixer noise modeling are now well understood, the subject is still complicated by a number of practical dilemmas. Chief among these is the handling of diode or active-device terminations at mixing frequencies beyond those of the RF input and IF output. The problem is further complicated by a number of popular publications that are poorly thought out or, occasionally, erroneous. To minimize the visibility of such publications, we will not reference them here; instead we present the correct theory, and hope that the incorrect simply disappears.
A mixer is fundamentally a multiplier, which multiplies the signal frequency, usually called the RF, by a local oscillator waveform, the LO, ideally sinusoidal, to produce an output at the intermediate frequency, or IF. [1] The signal at the IF is a replica of the RF, shifted to a lower or higher frequency, with (again, ideally) no loss of information.
The mixing process requires a time-varying parameter, which is usually a conductance...
