Synchronization and Arbitration in Digital Systems

Today’s networks of processors on and off chip, operating with independent clocks, need effective synchronization of the data passing between them for reliability. When two or more processors request access to a common resource, such as a memory, an arbiter has to decide which request to deal with first. Current developments in integrated circuit processing are leading to an increase in the numbers of independent digital processing elements in a single system. With this comes faster communications, more networks on chip, and the demand for more reliable, more complex, and higher performance synchronizers and arbiters. Written by one of the foremost researchers in this area of digital design, this authoritative text provides in-depth theory and practical design solutions for the reliable working of synchronization and arbitration hardware in digital systems. The book provides methods for making real reliability measurements both on and off chip, evaluating some of the common difficulties and detailing circuit solutions at both circuit and system levels. Synchronization and Arbitration in Digital Systems also presents: mathematical models used to estimate mean time between failures in digital systems; a summary of serial and parallel communication techniques for on-chip data transmission; explanations on how to design a wrapper for a locally synchronous cell, highlighting the issues associated with stoppable clocks; an examination of various types of priority arbiters, using signal transition graphs to show the specification of different designs (from the simplest to more complex multi-way arbiters) including ways of solving problems encountered in a wide range of applications; essential information on systems composed of independently timed regions, including a discussion on the problem of choice and the factors affecting the time taken to make choices in electronics. With its logical approach to design methodology, this will prove an invaluable guide for electronic and computer engineers and researchers working on the design of digital electronic hardware. Postgraduates and senior undergraduate students studying digital systems design as part of their electronic engineering course will struggle to find a resource that better details the information given inside this book

Table of Contents:
Preface. List of Contributors. Acknowledgements. 1. Synchronization, Arbitration and Choice. 1.1 Introduction. 1.2 The Problem of Choice. 1.3 Choice in Electronics. 1.4 Arbitration. 1.5 Continuous and Discrete Quantities. 1.6 Timing. 1.7 Book Structure. PART I. 2. Modelling Metastability. 2.1 The Synchronizer. 2.2 Latch Model. 2.3 Failure Rates. 2.3.1 Event Histograms and MTBF. 2.4 Latches and Flip-flops. 2.5 Clock Back Edge. 3. Circuits. 3.1 Latches and Metastability Filters. 3.2 Effects of Filtering. 3.3 The Jamb Latch. 3.3.1 Jamb Latch Flip-flop. 3.4 Low Coupling Latch. 3.5 The Q-flop. 3.6 The MUTEX. 3.7 Robust Synchronizer. 3.8 The Tri-flop. 4. Noise and its Effects. 4.1 Noise. 4.2 Effect of Noise on a Synchronizer. 4.3 Malicious Inputs. 4.3.1 Synchronous Systems. 4.3.2 Asynchronous Systems. 5. Metastability Measurements. 5.1 Circuit Simulation. 5.1.1 Time Step Control. 5.1.2 Long-term τ. 5.1.3 Using Bisection. 5.2 Synchronizer Flip-flop Testing. 5.3 Rising and Falling Edges. 5.4 Delay-based Measurement. 5.5 Deep Metastability. 5.6 Back Edge Measurement. 5.7 Measure and Select. 5.7.1 Failure Measurement. 5.7.2 Synchronizer Selection. 6. Conclusions Part I. PART II. 7. Synchronizers in Systems. 7.1 Latency and Throughput. 7.2 FIFO Synchronizer. 7.3 Avoiding Synchronization. 7.4 Predictive Synchronizers. 7.5 Other Low-latency Synchronizers. 7.5.1 Locally Delayed Latching (LDL). 7.5.2 Speculative Synchronization. 7.6 Asynchronous Communication Mechanisms (ACM). 7.6.1 Slot Mechanisms. 7.6.2 Three-slot Mechanism. 7.6.3 Four-slot Mechanism. 7.6.4 Hardware Design and Metastability. 7.7 Some Common Synchronizer Design Issues. 7.7.1 Unsynchronized Paths. 7.7.2 Moving Metastability Out of Sight. 7.7.3 Multiple Synchronizer Flops. 8. Networks and Interconnects. 8.1 Communication on Chip. 8.1.1 Comparison of Network Architectures. 8.2 Interconnect Links. 8.3 Serial Links. 8.3.1 Using One Line. 8.3.2 Using Two Lines. 8.4 Differential Signalling. 8.5 Parallel Links. 8.5.1 One Hot Codes. 8.5.2 Transition Signaling. 8.5.3 n of m Codes. 8.5.4 Phase Encoding. 8.5.5 Time Encoding. 8.6 Parallel Serial Links. 9. Pausible and Stoppable Clocks in GALS. 9.1 GALS Clock Generators. 9.2 Clock Tree Delays. 9.3 A GALS Wrapper. 10. Conclusions Part II. PART III. 11. Arbitration. 11.1 Introduction. 11.2 Arbiter Definition. 11.3 Arbiter Applications, Resource Allocation Policies and Common Architectures. 11.4 Signal Transition Graphs, Our Main Modelling Language. 12. Simple Two-way Arbiters. 12.1 Basic Concepts and Conventions. 12.1.1 Two-phase or Non-return-to-zero (NRZ) Protocols. 12.1.2 Four-phase or Return-to-zero (RTZ) Protocols. 12.2 Simple Arbitration Between Two Asynchronous Requests. 12.3 Sampling the Logic Level of an Asynchronous Request. 12.4 Summary of Two-way Arbiters. 13. Multi-way Arbiters. 13.1 Multi-way MUTEX Using a Mesh. 13.2 Cascaded Tree Arbiters. 13.3 Ring-based Arbiters. 14. Priority Arbiters. 14.1 Introduction. 14.2 Priority Discipline. 14.3 Daisy-chain Arbiter. 14.4 Ordered Arbiter. 14.5 Canonical Structure of Priority Arbiters. 14.6 Static Priority Arbiter. 14.7 Dynamic Priority Arbiter. 15. Conclusions Part III. References. Index.

About the Author :
David Kinniment is Professor Emeritus at the University of Newcastle, having been Professor of Electronics in the Department of Electrical, Electronic and Computer Engineering from 1979 until 1998. He currently works part time, mainly on asynchronous design and design methodology, in the Computer Science Department. He is very experienced in the measurement and characterization of synchronization and arbitration in digital systems, having been one of the leading researchers in the world in this area of digital design since the 1970s. He wrote the first paper to describe the measurement of mean time between failures (in other words, realiability) for a metastable device in 1972 and this year (2006) he won the best paper award at the ASYNC 2006 conference. Along with this, he has written a number of published journal articles on the topic, and has also collaborated with industry (most recently INTEL) to develop efficient digital hardware design methods, and ways of ensuring reliability in digital processors.