TRIP REPORT: THE RIGHT TOOL FOR THE JOB
MicroUnity Designs New Chip for Media-Processing Applications

by Denise Caruso

c. 1994 Technology & Media


Not much is public about MicroUnity Systems Engineering. Other than a few recently published details about the juvenile delinquency of its cofounder, Chairman and CEO John Moussouris -- who happens to be one of the world's leading microprocessor designers -- and the fact that his old college roommate, Will Hearst, was his fist investor and board member, the company has disclosed very little, particularly about its very secret group of industry backers.

The speculation so far is a who's who of "convergence" companies -- everyone from TCI to US West, Microsoft, Silicon Graphics (which owns Moussouris' former company, chipmaker MIPS Computer Systems), Digital Equipment, Hewlett-Packard, General Instrument and more than a few others. No one at MicroUnity will confirm or deny, as they say, though a recent database search did yield a single factual nugget: The Federal Trade Commission cleared Liberty Media of antitrust concerns "in connection with the proposed purchase of assets or common stock" of MicroUnity. (What company just regained 100 percent control of Liberty Media? TCI? Just checking. . .)

I've been semi-tapped into MicroUnity for a while, but I was stunned during a recent visit at the progress the company has made on what it calls its "media processor," a new semiconductor architecture invented at MicroUnity that Moussouris says provides "the highest possible physical bandwidth that's both affordable and useful to the end user."

Translation: MicroUnity has patented several processes for producing small, inexpensive and extremely fast computer chips designed to be built into anything that communicates: cellular phones, handheld computers, cable headends, settop boxes, telco switches, media servers, wireless base stations, pagers, etc.

These media chips can compress and decompress, render, error-correct, encrypt and decrypt, and otherwise process -- in real time, not "while you wait" -- virtually any kind of digital signal, including video, voice, data, radio and interactive content. And this happens over virtually any kind of network, including cellular, PCS, the Internet and other computer networks, cable and telco.

Moussouris claims the chips will cost the same as an Intel 386 processor does today (about $45 in quantity), but supply more than 100 times the bandwidth of Intel's hot new Pentium chip.

As you might expect, Moussouris makes a good case that digital media's present shortcomings -- i.e., poor quality video and slow performance -- stem from using the wrong chip for the job.

That's because today's microprocessors were designed to calculate, not communicate. A communications processor in the digital age has to be able to take in any data from any source, process it and send it to any display device -- TV, telephone, computer, whatever -- without today's common problems of incompatibility, delays and/or quality degradation.

This kind of processing is done today, not particularly well, by adding six or so specialized function chips to a computer's main processor. MicroUnity does it -- better and faster, according to Moussouris -- by designing a general-purpose chip that eliminates the need for specialized functions.

MicroUnity's approach to designing and building the chip is remarkable. For one thing, the company built its own state-of-the-art foundry in one year for only $50 million, an extraordinary accomplishment. But what's equally fascinating is how MicroUnity is positioned in the larger world of network owners and content providers.

Moussouris believes the main reason network owners will want to use MicroUnity's chip is because "they don't want to roll trucks" to physically deliver new service capabilities to consumers. Once a MicroUnity chip is in a consumer device, such as a settop box or a telephone, network providers will be able to upgrade customers to new capabilities without having to ship out specially equipped boxes (as they must today) to deliver services such as pay-per-view or scrambled programming such as Disney Channel or Playboy Channel. "If it saves one truck roll, it saves $3 billion," Moussouris says.

That figure is not particularly meaningful, since there isn't (yet) just one cable company, for example, and he's counting 60 million cable households as a single entity. (And of course, what if the problem is one of MicroUnity's chips? You can't ship a chip over a wire.)

But his point is well taken: Once a sufficiently capable piece of hardware is in place, it becomes much easier and cheaper to upgrade the software that will run new consumer services. There is great benefit to being able to quickly prototype and test services without launching an expensive and risky new hardware platform. Gauging customer reaction to such offerings as mutliplayer videogames, for example, or 3DTV is no longer a question of polling customers -- it becomes a simple technical issue of network management -- tallying "hits" on the network as they happen, in much the same way that QVC's network tracks how many items are available and how quickly they're selling.

The other benefit is clearly for content providers who have been shy about committing quality properties to substandard digital media technologies. If a MicroUnity media processor can exploit every bit-per-second of available bandwidth, and it becomes widely used by distributors such as cable companies, hardware designers and telcos, Moussouris believes the net result is the creation of a reference platform that guarantees a known level of performance, whether it's an interactive media experience or speech processing on a cellular phone.

But these benefits clearly put the cart before the horse. What MicroUnity has to do first is take on the monumental task of unseating at least two powerful incumbents -- Intel and Silicon Graphics -- both of which already have developed significant customer bases both in digital media and the traditional computing communities. It also must go up against Motorola, the market leader for chips built into devices such as pagers and cellular phones.

The goal is ubiquity far beyond what the telephone network has today. In five years, says Moussouris, he wants to be able to open up any telco switch, digital device or TV set and find MicroUnity chips in 50 percent of them. How he intends to accomplish this formidable goal is simple: First get a phalanx of powerful partners, then sell like hell into every equipment-providing industry he can, from consumer electronics to computer and telephone manufacturers. MicroUnity has 18 patents filed, 50 in progress and they're all available for license on a non-exclusive basis, "for a fair price," to all comers. "We want to help other companies make $100 for every dollar we make," he says.

It's an ambitious strategy, made even more so by the possibility that it just might work. Clearly even today's best technology is barely keeping up with the demands that digital media puts on it, and the trend is toward more -- more speed, more power, more bandwidth. If MicroUnity delivers on the technology end of its bargain, it's hard to imagine that the industry won't respond.


HOT TECHNOLOGY ALMOST OUT THE DOOR
Sarnoff Needs Funds for 3DTV, Other Imaging Projects

The David Sarnoff Research Center in Princeton, N.J., for many years the research arm of the RCA Corp., is still doing what it has done for more than 50 years -- pioneering work in imaging and image manipulation. A recent visit to the lab revealed an incredible array of technologies with great potential for next-generation entertainment products, as well as for medicine and business.

The biggest difference between today and 50 years ago, though, is that the lab is now hungry for money. Hit by the same scarcity of funding for pure research as other venerable institutions such as AT&T Bell Labs and Bellcore, Sarnoff -- now owned by SRI International -- is looking for corporate partners to help push its work out of the lab and into the market. Any company in the digital media industry with money to invest would do themselves a service by looking into the possibilities.

Though it's impossible to detail them here, the projects mentioned below have great commercial potential. (For example: All the underlying research for the Hughes DirecTV system, the direct broadcast satellite system, was done at Sarnoff.)

First and most significant is the 3DTV initiative headed by Dennis Matthies, director of Sarnoff's Display & Imaging Systems Lab. Matthies claims that by using a digital signal rather than analog, a standard TV set (attached to an appropriately equipped settop box) can display 3D information that can be viewed with a simple, comfortable pair of glasses -- not the big, bulky electronic goggles used to view virtual reality or the flimsy, cardboard-and-plastic lenses that most 3D programs use today.

The 3D effect is amazingly realistic (Senior Editor Mary Fallon and I kept leaping backward during a demo where a ball was swung toward the camera) and the possibilities for a wide variety of applications -- home shopping, videogames, music videos, cartoons and video teleconferencing -- were immediately apparent.

Seems that a big cable or telephone company looking for a way to be "futuristic" and get a programming edge in the new world of digital TV would be interested in getting its hands on this technology.

Though he said he couldn't easily put a price tag on what it would cost to finish the project, Matthies says the cost of the whole package -- camera development, transmission standards and displays -- "costs significantly less than one feature film." (We're assuming he doesn't mean a Jim Cameron gargantua.)

In addition, the lab is flexible in terms of business arrangements for all its projects, not just 3DTV. Costs vary depending on whether the lab is doing contract development, has a royalty arrangement or equity position, is participating in a joint venture or if the end result is a separate, spin-off company.

Sarnoff also is developing an impressive array of dazzling video manipulation technologies. Among them are scene manipulation, in which people or objects can be eliminated from video images without a trace; video manipulation, which can eliminate jitter from video scenes (this has many applications in the security business); super-resolution, which can take a series of low-resolution video images and integrate them into a single frame, making it possible to clarify details in an image that otherwise would be impossible to distinguish (such as license plate numbers and facial detail); and video indexing, an oddly effective "virtual" compression technology that takes a panned video image and transforms it into a single panoramic still. Again, cost to commercialize depends on the individual business arrangement.

It's been a while since I've been so impressed by the depth and range of technological expertise in a single location. Now that the infrastructure visionaries have everyone's attention, it's time they turned their attention toward new and interesting ways to create what's going to be flowing through all those nice fat fiber-optic and coax pipes. The work being done at Sarnoff has great potential to move that process forward.


VIRTUAL REALITY MOVES OUT OF THE LABS
Seattle Startups Show Commercial Promise

For some enigmatic reason, Seattle is becoming a kind of fertility zone for new virtual reality (VR) companies. A recent trip focused on two VR startups, both of which seem to have a surprisingly high degree of commercial potential.

The first is Worldesign, brainchild of president Bob Jacobson, who in 1992 left the Human Interface Technology (HIT) Lab for VR research at the University of Washington to start Worldesign. Jacobson's intent is to move VR technology out of the lab and into the market. While he's clearly not the first to have this idea, he might actually have a chance to realize it.

After an early, unsuccessful focus on designing software tool kits to build 3D (or 3-space, as the VR folks call it) worlds, Worldesign's investors asked Jacobson to try something new. His response was to move toward what he knew best: computer-aided design, combined with geographical information systems, a fast-growing software genre that provides highly accurate digital descriptions of topography for customers such as utility companies and architects.

Worldesign's RealityWorks Worldbuilding Software has been designed to combine these disparate types of simulation data to create navigable, 3-space environments that planners can use to visualize and test designs before building them.

Finding resistance in the professional community to wearing the clunky and weird head-mounted displays traditionally associated with VR, Worldesign also designed a new kind of display called a Virtual Environment Theater, or VET. A standard-sized VET, which seats 5 to 10 people and takes up about 1,000 square feet, costs $380,000.

Jacobson says the company's tightened focus is starting to pay off. Corporate clients such as Bechtel, Puget Power and the Idaho National Energy Labs are lining up to create, view and "inhabit" spaces ranging from building designs to simulations of hazardous environments, and consumer applications aren't far behind. An Atlantic City mogul, for instance, is hot on the idea of setting up a chain of VETs that could be installed in casinos, malls and storefronts around the globe.

Consumer appeal is the raison d'être for Zombie, the second Seattle-area VR development house with a commercial edge. Zombie has already designed the hardware, software and head-mounted display for a VR game system that a large videogame maker will release in 1995.

Zombie's niche will be games that utilize the head-mounted display technology that Jacobson's professional customers reject -- yet another painful indicator that the videogame generation is far more willing to be "plugged in" than most of us ever hope to be.

Founded by former Sarnoff researchers Mark Long and Joanna Alexander -- he's the business guy, she's the technical guru, by the way -- Zombie got its name from countless VR conferences where, as Long puts it, someone always stands up and says, "This technology is going to make zombies of our children!"

That kind of amiable perversity is key to Zombie's corporate charm. Long, a former VR specialist with the U.S. Army, nearly cackles while describing one of the company's games, Ice & Fire, by the same team that designed the outrageously successful Tetris. He says designer Vladimir Polhilko isn't a programmer, he's a cognitive psychologist -- thus the game is aimed dead-center at the kinds of activities that people tend to find psychologically addictive. (It's not a new design ploy, but most people don't cop to it with such glee.)

Other Zombie titles in development include Powerball, which Long calls "Rollerball in a giant sphere" (think hamsters) and which he swears will not make you throw up (a common VR problem); a "Myst-like" title for Activision; and a "dark, dark, dark" game called Zero Population Growth.

Though its investors (including Nick Nicholas, former co-chairman of Time Warner) must have to practice a bit of Zen letting-go at Zombie's irreverence, what makes it so refreshing is Long's and Alexander's intent to put creativity before money. "All we want to do is work with the very best people," says Long.

It takes chutzpah to launch a VR-only game company in what sure looks like a glutted videogame platform market. But the bigger question for Zombie, and to a lesser extent Worldesign, is still whether the very best people and technology available today can create VR that's real enough to be accepted by their chosen markets.

In fact, in some ways VR seems to be sitting in the same limbo zone as CD-ROM -- the technology is sufficiently impressive that people are afraid not to pay attention to it, but it's far from reaching its potential. Let's hope that VR companies like Zombie and Worldesign aren't mistaking a clear vision of the future for commercial demand today.