TC-9 Computational EMC Modern Computer Capabilities Enhance Today's Electromagnetic Simulations
	By Colin E. Brench

One aspect of computational modeling that is changing very rapidly is the capability of today's computers to run large, complex problems in very reasonable periods of time. Electromagnetic simulations have always been considered to require significant computer power and indeed there are many such problems that tax or even exceed the capabilities of today's largest computer systems. On the other side of the coin, there are problems that can now be solved quickly and efficiently on a typical notebook computer. These types of problems are probably the ones of most interest to the EMC engineer.

In 1988, this author began using the Method of Moments code, MiniNEC, among other tools, to solve EMI problems. In order to handle the "large" problems of the day, it was necessary to modify the code to run on a Digital VAX computer so that advantage could be made of the 64 M of RAM available. This amount of RAM was infinitely more than what was available on a PC of the day and many times what was actually needed to solve the problem. A few years later an FDTD tool became available and more computer resources were needed. A few more fast processors, each with a few hundred megabytes of RAM available, seemed like heaven; many problems could be set up and would run over night.

In today's environment, processor speeds have greatly increased and available RAM runs to many gigabytes. Models using two gigabytes are not uncommon and two or three times this amount is by no means impossible. These large problems still run over night.

One basic truth to life is that demand will fill up all available resources! If 50 terabytes of RAM became available tomorrow, within a very short time problem complexity would increase to fill it. It is good to sit back and consider what is really needed from computational EMI modeling and what we can do comfortably on today's desktop or even notebook personal computers. This article is being typed upon a personal notebook computer which has a 450 MHz processor and 192M of RAM. This is pretty reasonable by today's standards and, apart from perhaps the available RAM, is by no means unusual. The RAM was increased to permit the execution of FDTD codes but, as the notebook is a working tool, the RAM is very cost effective. On a daily basis I can carry around enough compute power to solve serious electromagnetic problems.

How should this influence the way we approach EMI modeling? Engineers of all disciplines have perhaps three levels of detail to their design work. Starting off on the back of an envelope, they progress to further detailed calculations, and then where necessary to computer simulations. With the resources available today, EMI modeling can move more easily to the earlier stages of this process. In place of the pencil scribbling on a notepad over a cup of coffee, a quick model can be thrown together and run while a fresh pot is brewed.

So what types of EMI problems can be done quickly and easily on the present day PC? For radiation problems, the numerical electromagnetics code (NEC) is a well-proven tool and can provide a great deal of information in seconds for each frequency of interest. NEC can also be used to determine coupling between various arrays of conductors, this is perhaps of greater interest to the EMC engineer. The sizes of the matrices that can be solved by NEC are easily handled in the RAM of a typical PC. By way of comparison to NEC, even the resource hungry, Finite Difference Time Domain (FDTD) can run on a machine with easily obtained amounts of RAM. 64 M RAM will permit many problems to be tackled; add another 128 M and most EMC problems can be managed. The FDTD solution time is measured in minutes, perhaps an hour or so, rather than in seconds as for NEC. Despite the longer run times, FDTD is still a very effective tool for quick models.

Within the computational resource limits of NEC and FDTD, there are other tools that can be used for EMC purposes. The limitation to the uses of any tool is the imagination of those using it. Given the computer resources that we have now, it is time to get creative and find ways to most efficiently tackle our EMI design tasks.

We are still awaiting the personal computer that will permit us to solve the field distributions on very large, complex structures at high microwave frequencies. However, we have more than enough practical sized problems that will keep us busy while we are waiting.

In the meantime, if you have topics concerning computational electromagnetics as applied to EMC that you would like to see addressed in this column, at the EMC Symposium, or through some other medium, please contact the author at colin.brench@compaq.com   EMC

Return to Table of Contents