Stanford University

News Service



Dawn Levy, News Service (650) 725-1944; e-mail:

Theory begets practice

"So I told him -- are you ready for this? -- I told him . . . hello? HELLO???"

Famous last words. Your colleague, who for the past five minutes has been leading up to a punchline that may make you or break your business, just drove around a downtown corner and behind a tall concrete-and-steel skyscraper at warp speed. Too bad: Her cell phone connection just couldn't navigate that turn with her, and you're left sitting at your computer with your fingers poised in midpeck above the keyboard and a lifeless receiver squinched between your cranium and your clavicle.

If you've ever been frustrated about cell phone calls being dropped mid-conversation, or being unable to place one to begin with, you're not alone. But few of us are familiar with the intimate details of the underlying hookups. We'd really prefer not to have to think about how our widgets work; we just want them to work. That's where some very sharp, mathematics-minded gadgeteers come in.

One of those digital wizards is Margaret Wright, head of Lucent Technologies' scientific-computing research department, a Stanford computer science Ph.D. and past president of the Society for Industrial and Applied Mathematics. Wright delivered this year's Forsythe Lecture, sponsored by Stanford's Computer Science Department on March 13 in the Gates Computer Science Building. The annual speech is named in honor of George Forsythe, the first chairman of Stanford's Computer Science Department, and his wife, Sandra, a noted computer science educator and textbook author.

Wright told an attentive audience of about 70 people that she constantly draws on the theoretical foundation she picked up at Stanford to work on problems the rest of us would rather let somebody else solve. A good example is a computer-based antenna-installation system that Bell Labs, the research division of Lucent Technologies, has devised to take the holes out of cell-phone coverage by minimizing those dreaded dead spots that can cost you your connection and make doing business a tricky proposition.

Figuring out how best to configure the grid of antennas that transmit signals to and from cell phones is pretty tricky business itself. "I've developed a new respect for cell phones," Wright said.

First, the electromagnetic waves cell phones send through the air can be reflected by hills or tall buildings. Second, "when an antenna tries to tune in to a phone, there are all the other phones emitting signals as well," Wright said. Those signals interfere with one another, like ripples at the surface of a swimming pool. It can get so bad that you can't make the call you're trying to make, or you can lose your call in mid-sentence.

That's not all, said Wright. "All the phones are actually adjusting their power all the time, depending on what the antenna is telling them. And traffic is moving, so things are always changing." Finally, there's a trade-off between the number of calls an antenna grid can accommodate and its "coverage," or ability to cover large areas without dead spots.

Wherever antennas are installed is pretty much where they're going to stay. In fact, because of such things as limitations on property rights, installers may have little or no say regarding the actual locations of the antennas to begin with. What they can control is the angles at which the antennas are tilted and their power output. This traditionally has been done via so-called "drive tests": Crews of installers literally cruise around in vans and check for dead spots by placing numerous calls and seeing how many get blocked. This is not only haphazard but expensive, because drive tests can go on for weeks.

Lucent installs wireless systems all over the world. As actual nuts-and-bolts hardware costs have fallen, the company has focused on reducing the cost of installing the grid of antennas that pass signals from phone to phone by eliminating or minimizing drive tests. The question was, How could that be done while still providing customers with the contractually guaranteed level of performance?

A Bell Labs team, several of whose members are Stanford computer science alumni, came up with an analytical software tool that allows computers to predict, with more accuracy than that of even expert human installers, which antenna tilt and power levels will ensure maximum capacity with the minimum number of dropped calls. The mathematical model they came up with, Wright said, employed fast, reliable numerical techniques and clever approximations, and was adaptable enough to accommodate the inevitable fine-tuning that programmers perform on the model to make it work better.

In many cases, installers are not equally concerned about being able to make calls from every spot within reach of the antenna grid. The bulk of cell-phone calls are placed from cars, which of course are found not in cow pastures or on craggy mountain slopes but on established streets and roads. The software package has been enhanced to optimize antenna placement for streets and roads scanned into a computer system from existing roadmaps.

The Bell Labs team was helped immensely by its ability to apply the kind of rigorous numerical analysis taught at Stanford, Wright said. Yet another Stanford-alumni-rich team, she continued, was able to apply its comfort level with deep theory to a quite different problem. This latter team designed a device, now in operation, that provides secure links between office computers and the approximately 30 million Americans who want access to them from their home computers.

"If your machine is connected to your office network and to the Internet, we all know that people can try to break in," Wright said. "Bell Labs is a favorite target for hackers -- I think we're on their top five list of places to break into -- so this is very much on our people's minds."

Key to acceptance of a security-enhancing solution was a near-zero need for monkeying around with connections or software settings. "Neither people at home -- who tend not to want to do anything -- nor people at the office wanted to have to go through a whole lot of administration."

The solution -- a small, brick-shaped piece of hardware, made from parts costing perhaps $80 -- provides telecommuters with a secure "encrypted tunnel" from home to office, said Wright. It was important that "you just plug it in and it goes."



By Bruce Goldman

© Stanford University. All Rights Reserved. Stanford, CA 94305. (650) 723-2300. Terms of Use  |  Copyright Complaints