SI经典之二红宝书—高速数字系统设计en.pdf

发布时间：2022-05-29 发布人：admin 分类：说明书资料大小：6.49M 资料格式：pdf 举报版权申诉

qq_27715745-8632839-4744302543382822580.pdf-第1页.png

第1页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第2页.png

第2页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第3页.png

第3页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第4页.png

第4页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第5页.png

第5页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第6页.png

第6页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第7页.png

第7页 / 共327页

qq_27715745-8632839-4744302543382822580.pdf-第8页.png

第8页 / 共327页

文本预览

High-Speed Digital System Design—A Handbook of Interconnect Theory and Design Practices Stephen H. Hall Garrett W. Hall James A. McCall A Wiley-Interscience Publication JOHN WILEY & SONS, INC. New York • Chichester • Weinheim • Brisbane • Singapore • Toronto Copyright © 2000 by John Wiley & Sons, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4744. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10158-0012, (212) 850-6011, fax (212) 850- 6008, E-Mail: PERMREQ@WILEY.COM. For ordering and customer service, call 1-800-CALL-WILEY. Library of Congress Cataloging-in-Publication Data: Hall, Stephen H. High-speed digital system design: a handbook of interconnect theory and design practices/Stephen H. Hall, Garrett W. Hall, James A. McCall p. cm. ISBN 0-471-36090-2 (cloth) 1. Electronic digital computers—Design and construction.2. Very high speed integrated circuits—Design and construction.3. Microcomputers—Buses.4. Computer interfaces.I. Hall, Garrett W.II. McCall, James A.III. Title. TK7888.3 H315 2000 621.39'8—dc21 00-025717 10 9 8 7 6 5 4 3 2 1

Preface Overview This book covers the practical and theoretical aspects necessary to design modern high- speed digital systems at the platform level. The book walks the reader through every required concept, from basic transmission line theory to digital timing analysis, high-speed measurement techniques, as well as many other topics. In doing so, a unique balance between theory and practical applications is achieved that will allow the reader not only to understand the nature of the problem, but also provide practical guidance to the solution. The level of theoretical understanding is such that the reader will be equipped to see beyond the immediate practical application and solve problems not contained within these pages. Much of the information in this book has not been needed in past digital designs but is absolutely necessary today. Most of the information covered here is not covered in standard college curricula, at least not in its focus on digital design, which is arguably one of the most significant industries in electrical engineering. The focus of this book is on the design of robust high-volume, high-speed digital products such as computer systems, with particular attention paid to computer busses. However, the theory presented is applicable to any high-speed digital system. All of the techniques covered in this book have been applied in industry to actual digital products that have been successfully produced and sold in high volume. Practicing engineers and graduate and undergraduate students who have completed basic electromagnetic or microwave design classes are equipped to fully comprehend the theory presented in this book. At a practical level, however, basic circuit theory is all the background required to apply the formulas in this book. Chapter 1 describes why it is important to comprehend the lessons taught in this book. (Authored by Garrett Hall) Chapter 2 introduces basic transmission line theory and terminology with specific digital focus. This chapter forms the basis of much of the material that follow. (Authored by Stephen Hall) Chapters 3 and 4 introduce crosstalk effects, explain their relevance to digital timings, and explore nonideal transmission line effects. (Authored by Stephen Hall) Chapter 5 explains the impact of chip packages, vias, connectors, and many other aspects that affect the performance of a digital system. (Authored by Stephen Hall) Chapter 6 explains elusive effects such as simultaneous switching noise and nonideal current return path distortions that can devastate a digital design if not properly accounted for. (Authored by Stephen Hall) Chapter 7 discusses different methods that can be used to model the output buffers that are used to drive digital signals onto a bus. (Authored by Garrett Hall) Chapter 8 explains in detail several methods of system level digital timing. It describes the theory behind different timing schemes and relates them to the high-speed digital effects described throughout the book. (Authored by Stephen Hall)

Chapter 9 addresses one of the most far-reaching challenges that is likely to be encountered: handling the very large number of variables affecting a system and reducing them to a manageable methodology. This chapter explains how to make an intractable problem tractable. It introduces a specific design methodology that has been used to produce very high performance digital products. (Authored by Stephen Hall) Chapter 10 covers the subject of radiated emissions, which causes great fear in the hearts of system designers because radiated emission problems usually cannot be addressed until a prototype has been built, at which time changes can be very costly and time-constrained. (Authored by Garrett Hall) Chapter 11 covers the practical aspects of making precision measurements in high-speed digital systems. (Authored by James McCall)

Acknowledgments Many people have contributed directly or indirectly to this book. We have been fortunate to keep the company of excellent engineers and fine peers. Among the direct, knowing contributors to this book are: Dr. Maynard Falconer, Intel Corporation Mike Degerstrom, Mayo Foundation, Special Purpose Processor Development Group Dr. Jason Mix, Intel Corporation Dorothy Hall, PHI Incorporated We would also like to recognize the following people for their continuing collaboration over the years, which have undoubtedly affected the outcome of this book. They have our thanks. Howard Heck, Intel Corporation; Oregon Graduate Institute Michael Leddige, Intel Corporation Dr. Tim Schreyer, Intel Corporation Harry Skinner, Intel Corporation Alex Levin, Intel Corporation Rich Melitz, Intel Corporation Wayne Walters, Mayo Foundation, Special Purpose Processor Development Group Pat Zabinski, Mayo Foundation, Special Purpose Processor Development Group Dr. Barry Gilbert, Mayo Foundation, Special Purpose Processor Development Group Dr. Melinda Picket-May, Colorado State University Special thanks are also given to Jodi Hall, Stephen's wife, without whose patience and support this book would not have been possible.

Chapter 1: The Importance of Interconnect Design OVERVIEW The speed of light is just too slow. Commonplace, modern, volume-manufactured digital designs require control of timings down to the picosecond range. The amount of time it takes light from your nose to reach your eye is about 100 picoseconds (in 100 ps, light travels about 1.2 in.). This level of timing must not only be maintained at the silicon level, but also at the physically much larger level of the system board, such as a computer motherboard. These systems operate at high frequencies at which conductors no longer behave as simple wires, but instead exhibit high-frequency effects and behave as transmission lines that are used to transmit or receive electrical signals to or from neighboring components. If these transmission lines are not handled properly, they can unintentionally ruin system timing. Digital design has acquired the complexity of the analog world and more. However, it has not always been this way. Digital technology is a remarkable story of technological evolution. It is a continuing story of paradigm shifts, industrial revolution, and rapid change that is unparalleled. Indeed, it is a common creed in marketing departments of technology companies that "by the time a market survey tells you the public wants something, it is already too late." This rapid progress has created a roadblock to technological progress that this book will help solve. The problem is that modern digital designs require knowledge that has formerly not been needed. Because of this, many currently employed digital system designers do not have the knowledge required for modern high-speed designs. This fact leads to a surprisingly large amount of misinformation to propagate through engineering circles. Often, the concepts of high-speed design are perceived with a sort of mysticism. However, this problem has not come about because the required knowledge is unapproachable. In fact, many of the same concepts have been used for several decades in other disciplines of electrical engineering, such as radio-frequency design and microwave design. The problem is that most references on the necessary subjects are either too abstract to be immediately applicable to the digital designer, or they are too practical in nature to contain enough theory to fully understand the subject. This book will focus directly on the area of digital design and will explain the necessary concepts to understand and solve contemporary and future problems in a manner directly applicable by practicing engineers and/or students. It is worth noting that everything in this book has been applied to a successful modern design.

1.1. THE BASICS As the reader undoubtedly knows, the basic idea in digital design is to communicate information with signals representing 1s or 0s. Typically this involves sending and receiving a series of trapezoidal shaped voltage signals such as shown in Figure 1.1 in which a high voltage is a 1 and a low voltage is a 0. The conductive paths carrying the digital signals are known as interconnects. The interconnect includes the entire electrical pathway from the chip sending a signal to the chip receiving the signal. This includes the chip packages, connectors, sockets, as well as a myriad of additional structures. A group of interconnects is referred to as a bus. The region of voltage where a digital receiver distinguishes between a high and a low voltage is known as the threshold region. Within this region, the receiver will either switch high or switch low. On the silicon, the actual switching voltages vary with temperature, supply voltage, silicon process, and other variables. From the system designers point of view, there are usually high-and low-voltage thresholds, known as Vih and Vil, associated with the receiving silicon, above which and below which a high or low value can be guaranteed to be received under all conditions. Thus the designer must guarantee that the system can, under all conditions, deliver high voltages that do not, even briefly, fall below Vih, and low voltages that remain below Vil, in order to ensure the integrity of the data. Figure 1.1: Digital waveform. In order to maximize the speed of operation of a digital system, the timing uncertainty of a transition through the threshold region must be minimized. This means that the rise or fall time of the digital signal must be as fast as possible. Ideally, an infinitely fast edge rate would be used, although there are many practical problems that prevent this. Realistically, edge rates of a few hundred picoseconds can be encountered. The reader can verify with Fourier analysis that the quicker the edge rate, the higher the frequencies that will be found in the spectrum of the signal. Herein lies a clue to the difficulty. Every conductor has a capacitance, inductance, and frequency-dependent resistance. At a high enough frequency, none of these things is negligible. Thus a wire is no longer a wire but a distributed parasitic element that will have delay and a transient impedance profile that can cause distortions and glitches to manifest themselves on the waveform propagating from the driving chip to the receiving chip. The wire is now an element that is coupled to everything around it, including power and ground structures and other traces. The signal is not contained entirely in the conductor itself but is a combination of all the local electric and magnetic fields around the conductor. The signals on one interconnect will affect and be affected by the signals on another. Furthermore, at high frequencies, complex interactions occur between the different parts of the same interconnect, such as the packages, connectors, vias, and bends. All these high-speed effects tend to produce strange, distorted waveforms that will indeed give the designer a completely different view of high-speed logic signals. The physical and electrical attributes of every structure in the vicinity of the interconnect has a vital role in the simple task of guaranteeing proper signaling transitions through Vih and Vil with the appropriate timings. These things also determine how much energy the system will radiate

into space, which will lead to determining whether the system complies with governmental emission requirements. We will see in later chapters how to account for all these things. When a conductor must be considered as a distributed series of inductors and capacitors, it is known as a transmission line. In general, this must be done when the physical size of the circuit under consideration approaches the wavelength of the highest frequency of interest in the signal. In the digital realm, since edge rate pretty much determines the maximum frequency content, one can compare rise and fall times to the size of the circuit instead, as shown in Figure 1.2. On a typical circuit board, a signal travels about half the speed of light (exact formulas will be in later chapters). Thus a 500 ps edge rate occupies about 3 in. in length on a circuit trace. Generally, any circuit length at least 1/10th of the edge rate must be considered as a transmission line. Figure 1.2: Rise time and circuit length. One of the most difficult aspects of high-speed design is the fact that there are a large number codependent variables that affect the outcome of a digital design. Some of the variables are controllable and some force the designer to live with the random variation. One of the difficulties in high-speed design is how to handle the many variables, whether they are controllable or uncontrollable. Often simplifications can be made by neglecting or assuming values for variables, but this can lead to unknown failures down the road that will be impossible to "root cause" after the fact. As timing becomes more constrained, the simplifications of the past are rapidly dwindling in utility to the modern designer. This book will also show how to incorporate a large number of variables that would otherwise make the problem intractable. Without a methodology for handling the large amount of variables, a design ultimately resorts to guesswork no matter how much the designer physically understands the system. The final step of handling all the variables is often the most difficult part and the one most readily ignored by a designer. A designer crippled by an inability to handle large amounts of variables will ultimately resort to proving a few "point solutions" instead and hope that they plausibly represent all known conditions. While sometimes such methods are unavoidable, this can be a dangerous guessing game. Of course, a certain amount of guesswork is always present in a design, but the goal of the system designer should be to minimize uncertainty.

1.2. THE PAST AND THE FUTURE Gordon Moore, co-founder of Intel Corporation, predicted that the performance of computers will double every 18 months. History confirmed this insightful prediction. Remarkably, computer performance has doubled approximately every 1.5 years, along with substantial decreases in their price. One measure of relative processor performance is internal clock rates. Figure 1.3 shows several processors through history and their associated internal clock rates. By the time this is in print, even the fastest processors on this chart will likely be considered unimpressive. The point is that computer speeds are increasing exponentially. As core frequency increases, faster data rates will be demanded from the buses that feed information to the processor, as shown in Figure 1.4, leading to an interconnect timing budget that is decreasing exponentially. Decreased timing budgets mean that it is evermore important to properly account for any phenomenon that may increase the timing uncertainty of the digital waveform as it arrives at the receiver. This is the root cause of two inescapable obstacles that will continue to make digital system design difficult. The first obstacle is simply that the sheer amount of variables that must be accounted for in a digital design is increasing. As frequencies increase, new effects, which may have been negligible at slower speeds, start to become significant. Generally speaking, the complexity of a design increases exponentially with increasing variable count. The second obstacle is that the new effects, which could be ignored in designs of the past, must be modeled to a very high precision. Often these new models are required to be three-dimensional in nature, or require specialized analog techniques that fall outside the realms of the digital designer's discipline. The obstacles are perhaps more profound on the subsystems surrounding the processor since they evolve at a much slower rate, but still must support the increasing demands of the processor. Figure 1.3: Moore's law in action.

分享到：

赞收藏

资料库

SI经典之二红宝书—高速数字系统设计en.pdf

相关推荐

开发技术

热门标签

最新资料