The Trouble With EM — It Always Gives An Answer!
This is the first in a series of guest columns by Dr. John Dunn of AWR Corporation.
Now I know what you are thinking. How can this be a problem? As an overworked, overloaded, and overcommitted designer, I'm sure you have bigger issues to worry about than a CAD tool that (almost) always gives you an answer. The question is: does it give the right answer?
EM simulation tools follow the standard engineering paradigm: GIGO (Garbage In, Garbage Out). An EM simulator works by meshing up the physical geometry of the problem it has been given. Then, it solves for the S-parameters at the various ports that have been placed in the geometry. Ultimately, there are only two ways the simulator won't give an answer: if the setup of the geometry is so messed up that the simulator has no idea what to do (e.g. there are no ports), or if the problem is so large numerically that it will never solve it in a reasonable amount of time, and the job is eventually aborted. In all other cases, you will get an answer.
I've seen many naïve designers happily take the S-parameters from an EM simulator and rely on them as a definitive result. Nonetheless, I assume most designers will at least question a result of S11 = 1 when they are studying an interconnect line, but even so, the problem is much more subtle. For example, you get: |S11| = - 43.4 dB at the center of your filter's pass band. Do you believe it? Why? Unfortunately, many designers never question the assumptions that went into that result: the mesh size, the boundary conditions, and the ground return assumptions, to name a few of the top culprits.
So what is a designer to do? Be skeptical. Try some simple cases you know are right first. When confronted with a large, complicated problem, the experienced EM tool user will first try simpler problems that give insight into the more important and complex issues. As a starting point, change your mesh settings to see if it makes a difference. If you have to evaluate the layout of a 16-line bus feed, begin with two lines to see what mesh is reasonable to get what you need. Try to get it close to known results by comparing answers from standard models. Simplified problems tend to run quickly, and you can build your intuition. Don't make the mistake of not taking a few hours to gain experience on problems you understand before you tackle that two-day simulation monster.
EM simulators are wonderful tools, but please remember they almost always give an answer. It's your task to ensure the answer is a sensible one. And when in doubt, don't believe it unless you can prove it to your satisfaction!
About The Author
Dr. John Dunn, a recognized expert in EM modeling and simulation for high-frequency and high-speed circuit applications, is a senior applications engineer at AWR and develops and presents AWR training material to customers world-wide. Before joining AWR, he was head of the interconnect modeling group at Tektronix and a professor of electrical engineering at the University of Colorado, Boulder, where he led a research group in EM simulation and modeling. Dr. Dunn received his Ph.D. and M.S. degrees in applied physics from Harvard University and is a senior member of IEEE. You can contact him at jdunn@awrcorp.com.
About AWR
AWR is the innovation leader in high-frequency EDA. Its software solutions quicken the pace at which high-tech products like cell phones and satellite systems are developed. When AWR software is part of the design process, engineers can deliver cutting edge, affordable products faster, more reliably, and at a lower cost. Headquartered in El Segundo, CA, AWR is a privately-held, growing company with thousands of users world-wide. Learn more: www.awrcorp.com