Preface
The story of Digital Circuit Boards at Mach 1 GHz starts with my friend Daniel Beeker. Dan is a senior field applications engineer for Freescale Semiconductor. He was instrumental in getting me interested in circuit board design problems. He was the one that spurred me into finishing the 5th edition to my book “Grounding and Shielding,” which was published by John Wiley in 2007. I have rewritten this book five times since 1967 and when this fifth writing was finished, I really thought I was through writing books. Obviously, I was mistaken.
Dan sees the problems encountered by his customers. He recommended to his management that the users must be provided some help in the form of seminars. They agreed, and as a result Dan took on a new set of responsibilities. He was tasked to find speakers and arrange for seminars for Freescale customers. To locate speakers, Dan turned to his own personal library. The first book he took from the shelf was a copy of my “Grounding and Shielding.” He then looked into my web site and found my address. The result was that I was invited to participate in the Freescale Forum in Orlando and later to give a seminar to his customers in the Detroit area.
The seminar I gave was based on my book and was well received. For those familiar with my books, I use very simple physics to explain how interference is generated, how it enters circuits, and how the circuits can be protected. The principles are the same whether the problem is analog, digital, or rf. I had little trouble bringing transmission line theory into my discussions. I found I had to catch up on the language of circuit boards and how they are built. I had to find out what a BGA was, what prepreg meant, and what is an interposer board. I had to learn the difference between a blind and a buried via.
Fortunately, Dan followed through with additional seminars where the speakers understood the details of circuit board design and could relate more closely with the details of components and materials. I admire Dan for recognizing that the fundamentals must come first. Even though I knew little about the details of board design, I could show the users how and why layout geometry was critical if they were going to build successful boards.
Dan wanted me to get closer to the circuit board problems, so he arranged for me to attend the PCB conference in Santa Clara, given by the UP Media Group. At the conference I sat in on courses given by experts in the field. I learned a lot about how circuit boards were designed and built. The talks introduced me to the designers' problems. Many of the speakers used a combination of circuit theory and lore to explain circuit board behavior. This is just the problem I had been dealing with throughout my career. I was in a different field with a new language. I had a lot to learn. I wanted to understand the digital layout problem based on physics, not on lore.
I found that digital engineers were working in areas of nanosecond delays and picosecond rise times. This was an area where I understood the physics but not the details of board construction. I recognized that real time delays were involved, and this was not covered by circuit theory. The speakers related their real world experiences and how they resolved many difficult problems. I remember speakers saying that energy could be drawn from the ground/power plane faster than from a capacitor. This got the wheels turning. How fast is fast and how much energy is there? How fast are capacitors?
I knew the basic physics so the challenge was to learn the language and make sense out of all the material that was being presented. When I got home from the show one of the first things I examined was how energy is moved from a ground/power plane. I assumed a coaxial connection to the conducting planes. If a step load was placed on this connection, the wave that propagated outward moved in a circular pattern. The characteristic impedance of the wave depended on the radial distance from the point of coaxial connection. I recognized that there were continuous reflections as the wave propagated outward and that the energy returned to the source increased with time. Much to my surprise, I found that the power curve depended on conductor spacing and not on the dielectric constant. With a greater dielectric constant, the energy was pulled from a smaller area. With a higher dielectric constant and with multiple demands for energy there would be less cross talk.
This exercise helped me understand the problem of connecting to the ground/power plane. I recognized that vias were the accepted method of making connections between layers. Typically, a via geometry has an aspect ratio of unity, where the characteristic impedance is about 50 ohm. When I assumed that a short lead length of 50 ohm was used to connect the ground/power plane to a load, I found out a very important fact. A short section of coaxial line placed between the load and the ground/power plane increased the rise time significantly. What was happening was that a large number of wave reflections were required to move energy across this short connection.
This one fact really caught my attention. It had a far reaching impact on my thinking. How many places were there, where short sections of transmission lines were used to connect a load to a source of energy?
The first area I considered was the capacitor. The available books and all the speakers discussed the natural frequencies of capacitors. The limitations in performance were assigned to the series inductance. This was a good explanation for axial lead capacitors but what about surface mounted components? Because capacitors of any geometry have a natural frequency when tested with sine waves, the assumption that resonance is related to a simple series inductance is made. In my view, I was dealing with step functions and reflections, and these ideas of circuit theory did not exactly fit. I decided then that I needed to look at digital processes in a consistent manner. I could not mix step function discussions with sine wave terminology. I had to clean up my understanding and explain things in an unambiguous manner. Resonant frequency concepts using sine waves was not compatible with real time reflections on a transmission line.
Calling a capacitor a short transmission line was the first step. Then I realized that the symbol was misleading as it implied a midpoint connection. Then some obvious issues came to mind. How can you take energy out of a capacitor and put it back in at the same time? What kind of construction allows wave energy to enter between the conductors? If you want energy with a short time constant, how do you construct a low impedance connection to a capacitor? In working on these problems, a book was beginning to form.
I have been doing engineering and consulting while writing books for some time. I have come to the conclusion that we often use electrical symbols in a careless manner. We need the symbols, but they represent complex conductor geometries that are intended to store or dissipate field energy. This is the correct view for all electrical activity whether we describe circuit behavior in terms of step functions or sine waves. Unfortunately, the field view is cumbersome, difficult mathematically, and impractical most of the time. Circuit theory is always correct, but it requires that the simplifications that are made are applicable. For example, simple circuit theory does not allow for time delays and that eliminates transmission lines. It implies that the interconnecting leads do not affect circuit performance, as we assume zero resistance and inductance for conductors regardless of length. We can always include parasitic elements in any analysis, but when we need them they are still just approximations. A resistor at 1 MHz can be represented as having a single parasitic shunt capacitance. At 100 MHz, the lead inductance starts to play a role and a distributed parasitic capacitance is needed. At 1 GHz, the resistor becomes a part of a transmission line path and the return path geometry is critical. At 10 GHz, there is really no such thing as a resistor even if we add many correcting terms. It is a lossy conductor geometry that modifies an electromagnetic field. The problem we face is how do we represent this device? The symbols we draw are very misleading. So far, there are no suggestions on what else we might do.
In my “Grounding and Shielding book,” I recognized the need to use field theory to explain electrical activity. I kept it very nonmathematical. The ideas are important and not the exact numbers. This book is no different. The basic physics is again the starting point. In this book, the emphasis is on events that occur in picoseconds where the signals are in the volt range. In “Grounding and Shielding,” the problems were usually related to microvolt signals, where the frequencies of interest were under 1 MHz. Move the signal level and signal bandwidth up to six orders of magnitude to volts and gigahertz, and the same physics solves a different set of problems.
I treat the basics in Chapter 1. It is important for the reader to appreciate that all the answers to their problems rest somewhere in this Chapter 1. If the reader needs a more in-depth review, I suggest reading the first chapters of “Grounding and Shielding” book. Of course, there is always that textbook from school that rests on the nearby bookshelf.
This book is intended to help the reader understand the problems of laying out digital circuit boards for fast logic. It is not intended as a treatise on how circuit boards are made although this knowledge can often be very helpful. Understanding the manufacturing process helps in understanding practical design. An engineer needs to know the range of trace widths that can be accommodated or how thin a dielectric can be used or what materials will withstand soldering. He needs to understand the cost of different laminates and why they are needed. All of this comes from the experience gained in doing designs and working with a manufacturer. On top of all this knowledge, the engineer needs to know the basics of signal transmission so that the logic will function.
Learning is an ongoing process. Board manufacturers will continue to improve their art. The only constant thing will be physics. It was my intent in writing this book to stick with the basics and use the present art as an example when it seems relevant. It is interesting to note that the materials we use today are the same ones that were used 20 years ago. Change is usually a refinement in processing raw materials caused by the continuous demand to improve performance and reduce cost. I hope this book will make a few design tasks a bit easier.
Ralph Morrison
Pacifica, CA
April 2012