In Defense of Uselessness

■ James Rautio

really like doing things that are use- forced by measurement limitations? Per- This article first appeared in IEEE ful. This fellow was suggesting that my haps. Would communication systems Microwave Magazine, vol. 5, no. 1, work was useless. What’s more, it was engineers be able to take advantage of March 2004. It has been slightly a reasonable and fair question to ask! a stop-band rejection of more than 100 revised. Not having considered this question dB? Maybe. Would circuit designers be before, I did not have a good answer. able to design it? The difficulty would Within an hour, however, I had talked be increased many times with insuffi- was feeling pretty goo d about a with an experienced filter designer. He cient analysis dynamic range, or worse paper I had just presented [1] dem- told me that their filters are typically yet, unknown dynamic range. Ionstrating a dynamic range exceed- specified for stop-band rejection down This is an odd defense of useless- ing 120 dB for a method of moments to 80 dB. Looking at it from the analysis ness. I spend my time searching for (MoM) electromagnetic (EM) analysis point of view, we typically trust results justification that my own work is of a fairly complicated circuit. I had actually useful when someone sug- even taken care to point out that tiny ggests it is useless. That is exactly the noise ripples at 120 dB down were rreason I think uselessness should consistent with a numerical noise bbe defended: We don’t always see floor 155 dB down. “Pretty neat!” I uusefulness in advance. Roger Har- thought to myself. rrington related to me (as described Then came The Question. It was eearlier in this publication [2]) how from a good friend, a well-known hhis work on MoM was considered and capable professor. He asked uuseless because a computer could me privately, after the presenta- nnot invert even a 100 x 100 matrix; tion, “Why would anyone ever need tthe magnetic tape would wear out results with such a large dynamic ggoing back and forth. range, you can’t even measure that Three years into my work with far down.” Wow! I had not thought © MASTERSERIES MMoM (19 June 1986, 2:35 p.m., to ask that question. We engineers MMeadowbrook Lodge, Blue Moun- down to about 20 dB above the noise tain Lake, New York), a prominent floor. A noise floor at –100 dB works microwave designer told me that all James Rautio ([email protected]) nicely. Since many EM analyses can- this numerical EM stuff was useless is with the Sonnet Software, Syracuse, New York. not push any where near that far down, academic research with no practical maybe my work might be useful after all! application. I clearly recall the sink-

Digital Object Identifier 10.1109/MMM.2012.2189515 This causes me to wonder: Is the ing feeling I had as I realized he was Date of publication: 7 May 2012 80-dB typical stop band specification right. After all, the best I could do was

148 June 2012 to invert a 100# 100 matrix in about inverted, it was really neat. When I saw When I am talking with research- an hour and then only by using hand- my first analysis showing a dynamic ers or reviewing papers, I try to drive coded assembly language. What kind range pushing all the way down to 180 carefully. I will never condemn a new of practical circuit could I do with dB, it was really neat. We do it because EM analysis because it is slow or inac- only 100 subsections? Nothing! Maybe we love it. We do it because it is really curate. Someday it might grow up to I should just drop all this really neat neat. Usefulness is an afterthought. be fast and accurate. However, if some- but useless EM stuff and get real. The But we are engineers; usefulness one claims an analysis is fast when it is answer is clear today: We can use lower is important. If something is useless, really slow, or accurate but they give no upper decomposition (LUD) to invert that might be OK, as long as there is a quantitative indication of error, I point 100,000 # 100,000 matrices in about two chance, someday, that it might be useful. it out. I always try to give the child a hours, and numerical EM is a required So we do have to keep an eye on useful- chance to grow and to gently but firmly part of modern high-frequency design. ness, and if some course of our endeav- direct attention to areas that need work. Fortunately, I decided to ignore that ors really starts to look like it will never As long as it feels “really neat” and early (and accurate) designation of be useful (e.g., pig perfume [4]), then it is there is at least some remote potential uselessness. important to be realistic, cut our losses, for future usefulness, let’s drive care- In these—and many other—cases, and get on with life. We have all had our fully and refrain from using “useless- uselessness is an accurate designation failures, and, like losing lottery tickets, ness” as a reason to discard someone’s upon the start of a major piece of work. we tend not to publicize them. When we work. But perhaps some might feel this The people telling Roger Harrington strike out, it is important to realize it as advice is itself useless. They might be that his work was useless were abso- quickly as possible, shake it off, and get correct. But I hope they will consider lutely correct. The people saying my on with the game. carefully reading this essay a second work was useless knew exactly what What about these people who take time, because it is they for whom I they were talking about. My work on ideas whose greatness still lies in the write. EM dynamic range and quantitative future and blithely call them use- accuracy (it is incredibly difficult to get less? Are these people useless, igno- References papers published on this topic!) is also rant morons? I don’t think so. In fact, [1] J. C. Rautio, “Testing limits of algorithms as- sociated with high frequency planar electro- correctly described as useless, at least these people are basically just like the magnetic analysis,” in Proc. European Micro- for now, for some people. rest of us. I’ll go even further: These wave Conf. Dig., Munich, Germany, 2003, pp. Why do we creators of useless people are us. Any one of us can and 463–466. research continue to work so hard? have pointed to a future great idea and [2] J. C. Rautio, “Planar electromagnetic analysis,” IEEE Microwave Mag., vol. 4, pp. 35–41, Mar. When Roger Harrington did the calcu- called it useless. It happens. The down- 2003. lations for his classic text on EM theory side is we might kill a future great idea [3] R. F. Harrington, Time-Harmonic Electromagnet- [3] in the mid 1960s using an advanced, with our quick callous of judgment. It’s ic Fields. New York: McGraw-Hill, 1961. state-of-the-art electromechanical calcu- like a car running down a small child; [4] M. Golio, “Continuing adventures of Bo Cam- bert and Leary McFly,” IEEE Microwave Mag., lator, it was really neat. When I got my as the parents cry over the small coffin, vol. 3, no. 2, p. 144, June 2002. first MoM matrix successfully filled and we wonder what might have been.

MTT-S Ombuds Officer (continued from page 138)

services, [email protected]. To add a new service electronically, please visit the IEEE at www.ieee.org and click on renew. To purchase an IEEE product by e-mail: [email protected] or call the numbers listed above. Please feel free to contact me by let- ter, telephone, or e-mail concerning any complaint you may have or any assis- tance you may need in obtaining mem- bership services from the IEEE and MTT-S. My contact information may be found on the first page of this column.

150 June 2012