When Excel isn't enough

Engineering users of math software are a diverse lot. The publishers of some of the more powerful math packages believe that most engineers who need to perform calculations and create graphical displays based on experimental data are more than happy to write programs in specialized languages to perform these tasks. But even these companies (several of them, anyway) acknowledge that they can't overlook many technical users' affinity for the calculation-and-graphing-software market's 800-lb gorilla: Microsoft's Excel spreadsheet package. Excel is ubiquitous; you'll find it on nearly every PC user's desktop because it is part of the Microsoft Office suites that many PC manufacturers preload into every PC they sell. And Excel is inviting. Though perhaps not the most capable spreadsheet program ever published, it has a wide array of capabilities and is relatively easy to learn and use. Unlike the other Office applications (most notably, the bloated, needlessly complex Outlook), Excel actually appears to have been designed by people who use programs of its genre. (Note to Microsoft: You might try making your other applications behave as if knowledgeable users had been involved in their development; if Excel is any example, doing so could cut down on grumbling by users who feel that they have no practical alternative to your software, no matter how horrendous it is. On the other hand, when you own 95% of the market...)

Microsoft recognizes that a general-audience package can't profitably include capabilities that only small user communities need, so the company encourages an Excel-based cottage industry. The spreadsheet package includes an add-in facility, which enables users to easily install programs that enhance its basic capabilities. For example, users of data-acquisition hardware can choose among approximately half a dozen low-cost dedicated add-ins that allow acquiring data directly into Excel worksheets and producing plots and displays that are often more appropriate for presenting experimental data than are Excel's native graphics. Vendors of such packages include Dataq, IOtech. Keithley, Measurement Computing, and National Instruments.

Although their installation routines may place DLLs (dynamic-link libraries) and OCXs (Object Linking and Embedding Custom Controls) on users' hard drives, dedicated, low-cost add-ins, such as the previously mentioned data-acquisition packages, generally don't require installing other software. Conversely, add-ins that turn Excel into a front-end for high-powered math packages usually require the target package's presence on the user's PC.

If you don't normally use a high-performance math package and you need one only to add a few advanced functions to Excel (for example, to perform matrix operations that are more complex than Excel's array functions), the addition may appear to be a case of the tail wagging the dog. After all, most high-powered math packages cost significantly more than a separately purchased copy of Excel. Moreover, because Excel probably came preinstalled on your PC, you might easily have gotten the erroneous impression that the spreadsheet package cost you or your employer nothing.

Nevertheless, math-package publishers have good reasons for designing their Excel add-ins as links from Excel to their packages and not as dedicated programs. Creating a link to a package is easier and less costly than creating a less capable dedicated add-in. As a result, add-ins that link to established packages often cost less and do more than dedicated add-ins. Of course, a user who purchases an expensive high-powered package and uses it only to enhance Excel never realizes these cost benefits. Moreover, the high-performance package almost surely requires PC resources greater than those needed to run the dedicated add-in. However, in this era of 80-Gbyte hard drives that sell for $129, few architects of high-powered math software worry much about resource constraints.

Perhaps more important, publishers of the high-performance software also hope that, once you gain some experience with their packages, you will delve further into them and start using them extensively in their stand-alone form. Users often become unpaid evangelists for these packages. However, if you've been using Excel intensively and you become particularly fond of a new math package, you should think about whether you had been using Excel inappropriately. Publishers of high-powered packages stress that a major hazard of Excel is users' tendency to apply it where a different program could more quickly and easily provide the needed answers.

Excel has bred a generation of spreadsheet jockeys—largely in finance and nontechnical functions—not in engineering. Most engineers regard any software package as merely a tool for doing a job; if a familiar tool appears adequate, engineers rarely see much reason to find a replacement. But the math-software publishers maintain that hard-pressed engineers who have become familiar with Excel often use it in place of other, more appropriate programs. These publishers say that fear of having to spend significant time and effort learning a new program often motivates these engineers to use Excel and to fail to recognize the symptoms of using the wrong tool. The most pronounced symptom is spending too much time on the job—not on the technical aspects, but on finding creative ways to make an inappropriate tool perform functions that its creators clearly never intended. In such cases, a new tool could increase productivity so much that the time savings would dwarf the time spent learning its use.

The beauty of using an Excel add-in as your introduction to a more powerful program is that you can follow a familiar path as you learn to use the new tool. This tactic often represents the best of both worlds; it can take some of the pain out of climbing the learning curve, and it does minimal (and only temporary) damage to your productivity as you learn.

Anyone who wanted to make a career of evaluating math software might just be able to turn the activity into an entire life's work. A directory at http://directory.google.com/Top/Science/Math/Software/ lists more than 300 suppliers and packages, but even that voluminous listing probably isn't comprehensive. If the evaluations took an average of two weeks per package, the job could easily stretch out over more than a dozen years. And by then, it would be well past the time to start all over again.

Clearly, some of the packages in the Web-based directory are of only marginal interest to electronic engineers. Nevertheless, finding the best package for an application is a job of such magnitude that few people would find it worthwhile. It seems more appropriate to adopt a more pragmatic approach: Find a tool that works, and look further only when you start to wonder whether another tool might overcome annoyances and problems you've encountered.

Instead of trying to present any sort of comprehensive listing, the paragraphs that follow discuss a few products that attracted my attention during research on this topic.

A low-cost, easy-to-use package that does symbolic math is Derive, currently in Version 5. Derive has been around for many years, but Texas Instruments acquired it several years ago as an adjunct to its scientific-calculator business. Symbolic math is the ability to manipulate formulas that contain multiple unknowns represented by variable names or symbols rather than by numeric values. Spreadsheet programs such as Excel as well as most math packages, do not perform symbolic operations. Derive also creates attractive color surface plots, which Excel cannot create. Derive costs $199 for a single-user US commercial license. (All software prices in this article are for licenses of this type.) Of course, you can't expect a $199 package to do everything that a truly high-performance $2000 package can do, but within its limitations, Derive is a useful adjunct to a spreadsheet package.

UTS (Universal Technical Systems) offers several interesting tools, the best known of which is TK Solver (see sidebar "Online product selection: Say goodbye to cut and try"). TK Solver, currently at Version 4.0 ($399) incorporates the unusual ability to back-solve. That is, if you have found one acceptable solution to a problem in which there are several unknowns, you can impose constraints on variables—including ones for which you might already have solved—and find new and possibly more desirable input values than those you started with. The example that the sidebar describes uses TK's back-solving capabilities.

A TK feature of special interest is MathLook, which displays and lets you manipulate equations in notation that resembles what you would produce by hand, rather than as confusing strings of nested parentheses and superscripts preceded by carets, which you must use with all spreadsheets and many math programs. UTS also offers a package called MathLook for Excel, an Excel add-in that lets worksheets display formulas in this more convenient manner (Figure 1) and simplifies referring to variables by name instead of by references to spreadsheet cells. UTS hopes that, by the time you read this, it will have a new version of MathLook for Excel that you can install without installing TK Solver and the Excel Toolkit that comes bundled with it. Naturally, though, the company hopes that you'll become intrigued enough by MathLook for Excel to want TK itself. Pricing for the new version of MathLook for Excel was not available at press time, but the earlier version cost $99.

Although Data Translation's Measure Foundry isn't an Excel plug-in, and few people are likely to think of it as a math package (it's a data-acquisition package), the newly announced Version 3.0 ($249) includes "the formula evaluator," a feature that combines data acquisition with math. A perfect application is linearization of inherently nonlinear transducer outputs. Suppose that you use one of the company's analog-input boards or data-acquisition peripherals to monitor the output of a group of strain gauges during a vibration test. As long as you can construct a formula to translate the acquired signal from millivolts to microstrain, you can enter the formula and produce a real-time display of microstrain versus time. With a wideband stimulus, such as a shake table applying dynamic forces, such as white noise, impulses, or step functions to the structure under test, Measure Foundry could do FFT calculations and display microstrain versus frequency. You enter formulas by dragging tools from a graphical palette to a work space. Jobs that involve wideband signals should be right up Measure Foundry's alley because of its patent-applied-for architecture, which avoids unpredictable delays by bypassing Windows' message queue.

A recent major announcement in the math-software market was the arrival of Waterloo Maple Inc's Maple 8 ($1695, Picture). Maple is one of the best known and most comprehensive of the high-performance math packages. Some of Maple's competitors have even struck deals with Waterloo Maple to use parts of Maple's symbolic-math capability. Among the new capabilities in Maple 8 are what the publisher calls a reinvented user interface, an interactive plot builder, and facilities for guiding novice users and students through calculus computations. Additional new capabilities include a database of more than 13,000 scientific constants; a numeric solver for partial differential equations; a code generator for the Java language; the ability to call external Java static methods from Maple applications; and a variety of calculus tools, including vector- and variational-calculus functions.

Regardless of the type of software that handles your math needs, one fact should stand out: This market is about as far as markets get from a one-size-fits-all tool. Excel seems to solve more computation and graphing problems for more people than any other software package, but it doesn’t come close to solving all such problems for everybody. Even with the dozens of available add-ins, it’s questionable whether Excel serves the needs of 80% of math-software users. The enormous number of other packages demonstrates that it doesn’t. In Excel, Microsoft has restrained its usual propensity for adding a plethora of arcane and needless features. Excel owes much of its success to that admirable restraint.

Author Information

You can reach Senior Technical Editor Dan Strassberg at 1-617-558-4205, fax 1-617-558-4470, e-mail ednstrassberg@reedbusiness.com.