Variation-Aware Design of Custom Integrated Circuits: Practical Field Guide for Engineers | Circuit Design, Semiconductor Manufacturing & Custom IC Development

This is a truly excellent book for circuit designers to learn how they can check the robustness of their designs against process variations and predict the resultant manufacturing yield. The book teaches how to achieve this objective with a little, well-targeted and affordable extra simulation effort. What I liked most in this book is that it focuses on techniques that have shown to work well in practice instead of iterating on techniques that are outdated and simply do not work for modern sub-micrometer technologies. Surprisingly such techniques are still quite popular among researchers and "unsuspected" practitioners and this book brings down some myths. Of course all important techniques are reviewed, but those that have pitfalls and limitations are covered in a concise manner and the emphasis is given only on the most recent and powerful ones. It is important also to stress that the most recent and powerful techniques have never been documented before in a book. This book serves as a single entry point to the field and is concise and easy to navigate. The authors are true experts with a long hands-on experience and themselves have done some of the most remarkable contributions. If you don't want to be baffled by the many techniques out there, get this book and you will get to learn quickly techniques that are both handy and are shown to work very well in practice.The introduction lays out the basic concepts and gives a comprehensible and to-the-point overview. The second chapter will teach you how to quickly identify the worst-case process, variation, and temperature (PVT) corners of a design instead of blindly simulating irrespectively all PVT corners (which is very time-consuming, if possible at all). The third chapter refreshes the memory of the reader on some basic knowledge in statistics for better following the remaining chapters. The fourth chapter will teach you how to verify designs that have a target yield between 95-99.86% (e.g. analog, RF, I/O). In this chapter, the authors first describe a variety of techniques proposed to date to arrive to a novel "sigma-driven corner" technique with very appealing characteristics. The chapter ends with an overview on sampling techniques aiming at reducing further the simulation time by replacing the standard pseudo-random sampling. The more standard Latin hypercube sampling and quasi-Monte Carlo techniques, as well as an efficient novel "optimal spread sampling" technique are discussed. The fifth chapter deals with the more difficult problem of verifying designs that have a target yield higher than 99.86% or, equivalently, designs that should fail not more than a few times in millions or billions of fabricated instances (e.g. bitcells in memories and designs employed in safety-critical automotive or medical applications that have "zero defective parts per million" requirements). Again, the chapter gives an overview of well-known techniques summarizing their limitations and goes on to explain in detail an efficient novel "high-speed Monte Carlo" technique. The final chapter six completes the puzzle by describing methodologies on how to explore the design space to make the design robust.