Amplification, feedback, and sub-biasing: Applying analog techniques to high-speed, low-power digital CMOS circuits

Analog and digital circuit design architectures are usually regarded as being separate spheres - with very different objectives, and thus consequently with very different methods of approaching the design challenges. Some twenty-five years ago, several remarkable United States patents were issued to the late Albert Vinal; these patents were developed and issued at a time when it had not even yet become clear that CMOS was actually viable as (among other things) a high-speed digital circuit technology (as evidenced by the enormous interest drawn by BiCMOS at the time). What makes these patents remarkable is not only that they reflect a clear understanding that CMOS was indeed on the verge of being a high-speed technology; they also employ a number of analog-like techniques (such as serial-amplification and feedback) as the bases for their architectures. This paper will revisit four of the aforementioned patents, particularly with regard to the CMOS technology of the period. The behavior of these architectures will then be evaluated in the context of the evolution of CMOS technology into the present day - with respect to lower power supply voltages, low-power operation, and circuit operation in the moderate and weak inversion regimes.