Book review: Variation-Aware Design of Custom Integrated Circuits

12/13/2012 12:15 PM EST

Chapter 3 A pictorial primer on probabilities: Intuition on PDFs and circuits

This chapter talks about probability density functions, Monte Carlo sampling and yield estimation. Thankfully, for me, they chose to use a graphical way to explain most of these concepts rather than equations that make my eyes glaze over. They do eventually resort to equations for some of them towards the end.

Chapter 4 3-sigma verification and design: Rapid design iterations with Monte Carlo accuracy

I will run an excerpt of this chapter over on EDN.

This chapter deals what typical analog, RF and I/O circuits where the target yield falls within three-sigma. It applies the general techniques and mathematical methods introduced in the previous chapters to global and local process variations in a statistical manner. How many Monte Carlo samples are required? This chapter attempts to answer that question. It starts by looking at various design flows and describes how sigma driven corners are selected. Sigma-driven corner extraction is a crucial element of the suggested design flow and attempts to find corners that represent the bounds of performance. Three design examples are provided to demonstrate the techniques. As with other chapters, useful appendices are provided that go into more depth on some of the theoretical aspects.

Chapter 5 High-sigma verification and design: The accuracy of five billion Monte Carlo samples in minutes

Some blocks in a design are replicated many times over, such as memory cells or digital standard cells. Even higher confidence is required for these and this chapter deals with them. It starts by examining a 6 transistor bit cell and the number of samples that would be required to get the necessary level of confidence. Existing high-sigma approaches are examined before outlining the High-Sigma Monte Carlo method and showing the technique applied to five circuits.

Then it looks at applying this analysis to the system level and applying it to an SRAM memory slice

Chapter 6 Variation-aware design: Manual sizing, automated sizing and an integrated approach

The previous chapters looked at a design and provided analysis tools to assess if the design would have suitable yield under a range or operating conditions. But the design itself can be changed to help ensure that it will produce the desired yield, and that is the focus of this chapter. It concentrates on device sizing and looks at manual, automatic and an integrated approach where the user is guided towards making the greatest-benefit design choices first.

Chapter 7 provides some conclusions.

A few nits - but as you will see these are all very minor and do not get in the way of reading this book. I wish books were written in a gender neutral manner. It is not difficult to do and it is about time we removed gender bias where possible. RTL has not stood for Resistor Transistor Logic for a long time and mistakes like this show when non industry knowledgeable people are used as editors. I wish the references were marked as to where they were used.

So there you have it. If you are interested in variation-aware design then this book will probably be a good one to have on your book shelf. This is especially the case if you are working on custom logic or highly replicated cells such as standard cells used in memories or in the digital process.