Gabe's EDA Update
In June 2012 Gabe Moretti will celebrate 44 years in EDA. Gabe has contributed to the industry first as a developer, then as a senior manager and now as an editor and industry observer. He is a Senior member of the IEEE and the recipient of the IEEE RonWaxman Meritorious Award. Gabe has worked … More »
A Difficult Book About A Very Difficult Thing
November 3rd, 2012 by Gabe Moretti
There is a new book out. It is not cheap at $119 a copy but I think the price is justified by the depth of its contents and the criticality of the subject. The book by Trent McConaghy, Kristopher Breen, Jeffrey Dyck and Amit Gupta has the imposing title of: “Variation-Aware Design of Custom Integrated Circuits: A Hands-on Field Guide” published by Springer.
My first reaction to its contents was to think that things cannot continue the way they are going. Complexity is killing productivity and financial returns. Of course one should not be surprised. Asking unnatural things from light is both difficult and expensive.
Reading the book and getting meaningful information from it requires some understanding of statistical analysis, so do not be scared off by the equations. After all our profession requires precision in communications, and there is nothing more precise than a mathematical expression. The goal is to teach engineers the technology of PVT analysis. Process variation (P), power supply voltage (V), and temperature (T) are fundamental components in determining whether at the end you have soup or hogwash. And if the result is the latter it will be a very expensive one.
The book is not a reference publication, unless you are already very familiar with the techniques covered in the first three chapters. If you are, then chapters 4 and 5 will give you theory and techniques for 3-Sigma and High-Sigma verification and design, and chapter 6 will guide you through building better SPICE models. Otherwise read and digest the first three chapters. If by then you have not had the urge to change profession, go on and find out how to design circuitry that will yield more good die than before.
Just reading the foreword by Jim Hogan and the Table of Contents reminded me of the reason it is taking so long to get to commercial use of 20 nm process. it is clear that we need to change how we do things at the front end, since the problem generated in place and route are too expensive to solve. That will require discipline on the part of both architects and designers, as well as the willingness to accept less than optimal use of real estate.
Be a Real Estate Developer
Living in Florida, the land of mega developments, and talking about real estate gave me an idea. Let’s look at the die surface exactly the way local government looks at real estate: through zoning. Some areas are good for commercial development, some for parks and recreation, and some for residential buildings. And of course you can use even finer distinctions: condominiums and apartments, versus single family homes, for example. or heavy industry versus retail stores. You get the point. In this way we can have design rules that are specialized for the type of circuitry you want to create while at the same time you can only place that type of circuitry in zones of the die reserved for it. We know that blocks have different physical, and of course, logical characteristics. Let’s not handle them all in the same manner, as if they were just a collection of transistors and wires.
Now that you have found a place to build your development, you need to figure out where to put the streets (the data busses), how to distribute utilities (power grid), which type of housing goes where (condo building versus individual homes, versus attached villas), a central recreation center, and so on. Each of these must conform to very strict codes, both at the federal and at the local level. These are your design rules.
The book teaches how to verify that the design rules have been met. But it does not tell you how to plan your die, which blocks should go in which location. The assumption is that if you know how to verify you will know how to design. This is not always true. Knowing how difficult the verification task may push designers to be more cautious in spite of marketing wanting them to be as daring as possible.