Mastery of the complex

http://www.memagazine.org/contents/current/features/mastery/mastery.html




MECHANICAL ENGINEERING DESIGN

mastery of the complex

After a half-century of development, CAD continues to extend control over the design of ever-more-challenging systems.


by Jean Thilmany, Associate Editor
Consider the iPhone. Even in the likely event you don't own one, you've heard the buzz: The Apple hybrid combines a mobile phone, a widescreen iPod, and Web-surfing capability all rolled into a handheld device about the size of a playing card. It runs the same operating system as a desktop computer. Twenty years ago, engineers would never have been able to compress the electronics required for so many functions and get it to fit that small silver case.

But this story isn't about the iPhone. That's the last you'll hear of it here. This is the story of the design technology introduced only about 45 years ago that helped create that gadget, and that plays a role in the ongoing miniaturization of everything from the telephone to the computer. Integrated circuits, each no bigger than a square inch, which act as the brains for digital devices like cell phones, laptops, and similar items demonstrate what computer-aided design software has made possible.

The earliest integrated circuits, introduced in the 1960s, contained a few transistors. Today's circuits can contain millions of transistor equivalents-logic gates, flip-flops, multiplexers, and other circuits. Only a few centimeters wide, integrated circuits now contain millions of geometrical features arranged just so and drawn in exacting detail. No piece of paper on Earth is large enough to allow an engineer who works with pencil and paper to represent the complex circuitry that fits so precisely within the tiny space. By using CAD, engineers can stuff a lot more into these small packages.

CAD came along at the same time as computer graphics programs. Both technologies allowed shapes to be depicted on the computer screen that had been dominated until then by blinking letters and numbers. Like its graphics counterpart, CAD advancements have continued apace. Today's systems allow for 3-D views that let engineers slice into their digital designs to look inside.

Although integrated circuits are probably the best example of CAD's little-known capability for heavy lifting, for another example we could turn to the printed circuit boards inside the computers themselves. The boards require an exacting layout effort.

"To figure out the geometry of them manually and to do that engineering drawing on paper and determine the manufacturing instructions necessary to cause them to be etched on copper and making sure everything matches is a virtual impossibility," said Joel Orr, a consultant at Cyon Research, a Bethesda, Md.-based CAD analysis and consulting firm.

Along the way, CAD and the manufacturing software that drives manufacturing tools have married, to a certain degree. So, at least in theory, designers can send even their most complicated CAD drawings to be machined simply by hitting a button. In practice, the engineering-to-manufacturing handoff can still be clunky.

In the early days of CAD, engineers (mainly those who worked for large companies) did their work on 16-bit computers like this Commodore Amiga 500, circa 1987.


While CAD software has changed the ways engineers, architects, and graphic designers work every day, it has also seeped into modern life in other ways. Design software not only replaced the drafting table, it has strongly affected consumer choice perhaps without our even noticing. Look at what CAD evangelist John Baker (Yes, he holds that actual title at UGS, the software company in Plano, Texas.) calls the elegance of the iMac. As ever, industrial designers still model their concepts in traditional media like clay. What's relatively new is the ability to scan these clay pieces and import exact data about their curves and bends back into a CAD system.

Industrial designers can sculpt their product—say, the shell of an iMac—and import its dimensions back into a CAD system, where an engineer plays with it to digitally fit the product's guts—the circuit boards, the hard drive, the fan—into the shell. Then, engineers use analysis software to test the digital design to predict that it holds up. The laptop couldn't have been developed without such now-common analysis programs. Engineers can study how heat will flow through the device in order to best place the fan.

"You couldn't sit and build laptop after laptop to test for that," Baker said. "CAD brings us levels of precision of design and duplication and precision in manufacturing that were undreamed of before."

In an earlier generation, only 15 or so years ago, the analysis step was generally the purview of an analyst with a Ph.D. who spent full days and weeks working on a supercomputer to analyze a part. Today's systems can return the same results in minutes or hours. Easier analysis is one reason today's amusement park rides are so much more elaborate, scary, and higher than those of 50 years ago, to cite one unlikely example.

The process of testing a digital part, fixing its problems, and analyzing again what happens on a CAD model long before an actual prototype is made is called virtual prototyping. Making the physical mockup is saved for much later in the design cycle than previously. And virtual prototyping is what you can blame or exalt for the many different models of cell phones, MP3 players, and flat-screen televisions introduced every year. Digital prototyping has sped consumer design from the equivalent of 10 miles per hour to 60 mph in the past 20 years.

And it's a double-edged sword.

"Do we have more products because we can do more or because people demand more?" Baker said. "Probably a little of both."

The story of computer-assisted design could start with Patrick Hanratty, who is often called the father of CAD and is still active in the industry. But it just as easily could begin with Ivan Sutherland, now a researcher at Sun Microsystems and widely credited with creating the modern graphical user interface. Like so many software developments, it's difficult to pin down an exact moment of inception or to isolate an application's true creator.

By the early 1960s, many researchers knew that they could program computers to display shapes rather than just letters and numbers. Still, they couldn't conceive how such a display might look. Then, in 1963 while still a Ph.D. student at the Massachusetts Institute of Technology, Sutherland created his Sketchpad system, the first program to use a graphical user interface to interact with users. The program comprised an x-y plotter display and the recently invented light pen, a computer-input device shaped like a wand.

"In the past, we have been writing letters to, rather than conferring with, our computers," Sutherland wrote in his thesis. "For many types of communication, such as describing the shape of a mechanical part or the connections of an electrical circuit, typed statements can prove cumbersome. The Sketchpad system, by eliminating typed statements in favor of line drawings, opens up a new area of man-machine communication."

He essentially came up with the idea of drawing on the screen.

Sutherland's system displayed vector graphics rather than the raster graphics we're used to today. Sketchpad users controlled the cathode ray tube's electron beam via light pen to draw vectors on screen, creating shapes line by line. It was like operating a ray gun. You'd turn it on, draw a line, turn it off, move to the next point, and turn it on again in a process not entirely unlike the workings of an Etch-a-Sketch (which is, in fact, a simplified vector plotter).

The phosphate that displayed the drawings tended to fade so users had to continually refresh the display. With very complicated displays they'd have to refresh often, said Bernhard Bettig, a mechanical engineering professor at the West Virginia University Institute of Technology, who teaches a course on CAD history.

"And the system blinked a lot and when you got really complicated, you got a lot of blinking," Bettig said. "And there was no shaded view. But you still had lines on the screen, so it was a big deal."

In comparison, today's graphics are raster images made up of a rectangular grid of densely packed pixels, each individually colored to create the overall effect. Blown up, such images lose their definition. You can see they're composed of a quiltlike pattern of individual dots, or pixels.

In 1961, as Sutherland experimented with his Sketchpad, Hanratty was at work at General Motors' research laboratory in Warren, Mich. He'd already written software that some in the industry regard as the first programming language to automate machine tools. At GM, Hanratty developed programming for the numerically controlled machine tools portion of the company's design-augmented-by-computers project, an early attempt at CAD.

Howard Crabb also worked on the design-augmented-by-computers project. In 1964, he wrote a program for GM that, much like Sutherland's system, relied on a cathode ray tube, a light pen, and an alphanumeric keyboard. At the time, there were two such machines in the world, one at an IBM research facility and the other, created by Crabb and fellow engineers at the GM Technical Center, he said in an interview shortly before his death three years ago. Engineers used the light pen and created drawings from geometrical entities.

These first CAD programs used simple algorithms to display patterns of lines in two dimensions.

After the drawing was complete, the engineer printed it, after a fashion. That early CAD system was linked to a drawing machine, which replicated the design via a diamond stylus that scraped a Mylar coating off the surface of paper.

At the manufacturing plant, the drawing was inserted into a shadow-box-style machine that cast the pattern on a wall. Shop floor employees would literally hold their molds against the projection to ensure that they fell within proper tolerances.

Players like Ford (where Crabb later worked), Lockheed Corp., Renault, and McDonnell Douglas also developed their own proprietary CAD systems in the 1960s. Large companies in the aircraft, automotive, and electronic fields carried out these early CAD experiments mainly because their products' heavy engineering requirements were made easier with CAD. Large companies could more readily cover the expense of the mainframe computers that ran the programs, which were intended to automate the repetitive drafting projects engineers needed to perform daily. If computers had a hand in smaller projects, like bolt design, engineers could spend more time on larger issues, like the fuselage. Those first CAD applications slashed design time significantly at GM and Ford. The manufacturers' internal technology development groups mainly worked on these first-generation systems.

The first CAD systems were mainly developed in-house by the aircraft, automotive, and electronics companies. They could afford to house the mainframes that ran the programs.


After the design-augmented-by-computers project took off at GM, a rendering that took a talented, experienced draftsman 50 hours to draw could take an engineer working with the cathode ray tube and the light pen 12 minutes to create, Crabb said. Immediately, the difference between manual and machine drawing was made clear.

By the middle 1960s, only one company had tried to make a commercial CAD system. Control Data Corp.'s Digigraphics division offered a system, but the cost, including computer hardware, was too high.

Many in the industry point to Hanratty's 1971 founding of Manufacturing and Consulting Services Inc. as the turning point in CAD history. The company sold one of the first successful CAD products, Hanratty's Automated Drafting and Machining software, or ADAM, which ran on any 16-bit computer. The company supplied code to the likes of McDonnell Douglas, Computervision, Gerber Scientific, and Control Data to power their own products, which eventually mutated into name-brand CAD software available today. Many industry analysts estimate that 70 percent of all CAD systems can trace their roots to Hanratty's code.

Hanratty still acts as MCS president, writes code, and develops software for the company's products.

By the late 1960s and early 1970s, engineers and designers who used CAD did their work on individual terminals tied to their companies' mainframe computers. Although they now drafted on screen, they still did so by connecting two-dimensional vector graphics in a way that greatly resembled the manner in which engineers had always drafted.

The systems allowed engineers to work out potential manufacturing errors on screen, to more readily update their designs, and to render designs faster than they could by hand. But the wireframe drawings could be unclear to those who programmed the manufacturing tools that would turn the designs into parts. Think again of the Etch-a-Sketch figures. What looks to one observer like a cube within a cube might actually be the overhead view of a cube perched atop a cube. Or not.

These early CAD designs couldn't depict volume. And therein lay the ambiguity. As Bettig at West Virginia University explained it, "Did it open from the top, from the left, the right? It was ambiguous in terms of where the surfaces were. You didn't know how to look at it."

Also, engineers couldn't readily draw curves.

As the 1970s dawned, it was time for the third dimension to make the CAD scene. Today's 3-D systems that are based on solid modeling stem from the work of two men on two continents who worked on separate approaches at about the same time. In 1976, mechanical engineering professor Herbert Voelcker's group at the University of Rochester in New York used a process that came to be called constructive solid geometry, essentially a molding and joining of shapes.

Also in the middle 1970s, Ian Braid at Cambridge University in England released his solid modeler, Build, which delineated the boundary between solid and nonsolid to create models. As their methods varied, so did the eventual CAD systems based on their work, though their underlying principle of solid modeling is the same. But both are based on the bedrock of everyday geometry.

It would take two decades after the inception of 3-D modeling before personal computers would be strong enough and cheap enough for companies to reasonably purchase one for every employee. Even today, many smaller shops work in 2-D, albeit with more advanced software than those first 2-D oscilloscopes.

The 2-D holdouts have no real reason to shift to the third dimension. For their basic design, 2-D does the job just fine, and their customers don't complain, Bettig said. Autodesk Inc. was started in 1982 by 13 men who set up shop in Marin County, Calif., and the company continues today as the largest supplier of 2-D CAD software.


What Does an Engineer Do?


Before the software's widespread adoption, engineering students learned in school the formal language of drafting: how to specify design dimensions, how to draw line thicknesses to denote specific design aspects, and when to include notes that would give special instructions to manufacturers or engineers down the line, Baker said.

That language disappeared with the advent of design software. So did the emphasis on certain manual skills that, years ago, separated the best engineers from the rest. When engineers' main medium was pencil and paper, they needed to possess rigorous hand-eye coordination, much as a visual artist does, Orr said. If you couldn't make a drawing that the next guy could read, you were done because the art and science of engineering depended on those paper drawings as the sole means of communication.

Now that an engineer works mainly with software, drafting prowess is barely considered necessary to the job.

"That sounds like a small thing, but it's a big thing because it expanded the population of who could be a designer," Orr said.

In fact, the job of draftsman is obsolete. Just as there is now no such person as a typist, so too does a draftsman walk with dinosaurs. With any extinction, some things are invariably lost. And missed.

"We've lost a little of the art of the drawing, which is a shame to some extent," Baker said. "We've lost a little bit of the apprenticeship aspect, where you work with a master. But when the typewriter replaced the written word, the advantage was being able to read typed writing rather than cursive writing—the unambiguity of that. You lose some skills, but the value of the end product has improved."

So has CAD affected the way engineers think? Probably not. Design is still design. After all, engineers still need to account for the laws of physics, which don't change no matter how far and how fast software advances. A shaft that broke 100 years ago will break today if designed and built in the same way. But 100 years ago, the engineer would have had to wait for the shaft to break to discover a design flaw. Today's engineers can run a series of virtual prototypes to see that they've picked the wrong-size shaft for their design.

And in a strange way—through advertising and slogans— CAD has moved the engineer a little bit more to the forefront of American's consumer consciousness, Baker said.

"Before, the guy who drove the car never saw an engineering design for it. But now those rendered images are part of the advertising—with guys in white coats pointing to them," Baker said. "I'm not saying engineers are any more appreciated today, but people have a better understanding of what's happening behind the curtain. They're not waiting for a curtain to part and a car to drive out as though it's been hammered together by elves."

When General Motors says on national television that engineering is job one, viewers seeing the advertising don't question the assertion.

With that publicity has come a change in job description that today's engineers need to be ready for. When Baker attended Michigan Tech in the 1960s, he said, his professors were just beginning to see that an engineer's role wasn't going to be limited to drafting designs. They'd have a hand in selling their ideas to higher-ups. Today's engineers need to make their CAD images quite compelling to get buy-in from the people with the money and the approval power.

CAD helps because the engineer no longer shows up at a meeting and unrolls a bunch of blueprints. Today's engineers can call upon a CAD tie-in called photorealistic modeling that makes use of light and shading effects to give photographic realism to digital designs. That photo gives non-engineering types—who often have the approval power—a mental image of just how the finished product will look. With those photorealistic images, designs can be sold before they're made.

"The engineer is getting a little bit more exposure than he'd had because his work product is leveraged more," Baker said.

If this is the present, so very different from 50 years ago, what will the engineering future look like? Can we expect to see a new cell phone model every week, or be presented with ever-more-dazzling choices of telephones at Target?

Maybe, said Baker.

But don't worry. He doesn't expect consumer choice to get quite so far out of control. In other words, look for the next big new cell phone next month, rather than next week.