From Oxymoron to Mainstream: The Evolution and Future of RF CMOS

Well into the latter half of the 1990s, few experts believed that CMOS would ever become a credible RF technology. A scant decade later, CMOS dominates in many RF applications below 10GHz. Recent developments show credible performance at millimeter-wave frequencies, and continued scaling promises near-THz operation within a decade. This paper tracks key milestones in the evolution of RF CMOS to identify broad themes that permit plausible speculation about possible futures.

I. Introduction

Today, CMOS is so prevalent as an RF technology that it is hard for newcomers in particular to imagine a time when “RF CMOS” was regarded as oxymoronic (if not simply moronic). Aside from its notoriously poor transconductance per unit current, CMOS was thought to suffer so much from poor passive components and high noise that it would be forever an uncompetitive medium for realizing RF circuits. Yet, persistent work has overcome these deficiencies to such an extent that CMOS has actually become dominant in a great many RF applications. It is remarkable that the transition from “absurd” to “obvious” occurred in the space of a decade. It is perhaps even more remarkable that CMOS RF ICs for use at 60GHz are nearly commercial realities, and that possibilities of near-THz CMOS are now contemplated with seriousness. Tracing the history of this dramatic transition is interesting for more than simply nostalgic reasons. Identification of important enabling developments teaches valuable lessons that aid in predicting possible futures for RF CMOS specifically, and for wireless integrated circuit technology in general.