Rewarding Failure

According to public records, the "custom" design of the Breda streetcars contained large and small errors that did not become apparent until the cars were test run on the rails in San Francisco. Some flaws were fixed, after expensive modifications. To address the weight problem that generates neighborhood-disturbing rumbles and eats Muni's track, for example, Breda is retrofitting its 38-ton cars with "softer-riding" suspensions.

Breda officials declined to be interviewed for this article. In an interview last week, Cruz said he was unaware that the Breda cars are too long, but that the Bredas' weight problems were being addressed. Attempts to obtain more detailed comment this week from him were unsuccessful.

But if Muni keeps using the Italian rail cars, San Francisco will have to learn to live with some Breda problems. Muni can do little to further reduce the soul-jarring whine that emanates from the heart of the Breda motor.

And there is no way to amputate two feet of length from a Breda streetcar. And so there is no way that $472 million of streetcars will ever fix the subway problem they were supposedly designed to solve.

Ten years ago, Muni planners dreamed of transit utopia. Sleek carbon-steel street-cars would speed through the underground in multicar trains that would come and go automatically, guided by omniscient central control.

Today, however, lone or paired streetcars straggle, very occasionally, into packed subway stations, and waiting thousands crush one another in the rush to load the steel cans.

What went wrong?
Simply put, Muni tried to swallow more than it could digest. Along with building an entirely new fleet of streetcars, Muni set about revolutionizing its signal system by acquiring Alcatel's automatic train control software. If it worked, the control system would be a wonder. In practice, it has been a disaster.

In the Alcatel system, a cable is laid down the middle of Muni's underground tracks. Each streetcar is equipped with a transponder (basically a radio that sends and receives signals from the cable). Using these signals, software running on a simple personal computer keeps track of and -- in theory -- drives trains in the subway from afar. The driver becomes peripheral, even as computer peripherals become the driver.

This supermodern technology was meant to replace the 20-year-old electromechanical relay signaling system -- a system that controlled tunnel traffic lights, which were obeyed by train operators. The relay system had worked just fine in the Muni Metro, and could have been upgraded relatively cheaply. Instead, Muni chose to graft Alcatel's expensive wares onto Muni's humancentric signal system -- which turned out to be a very bad idea, indeed.

There are a number of technical reasons why the Alcatel signal system has never worked properly. Public records show that the system contained design flaws, and that there were manufacturing flaws in some of the equipment used. But it is also now clear that the Alcatel system installed in San Francisco had never been proven in practice, anywhere in the world. To this day, the system remains an experimental product.

So why did Muni buy an experimental guidance system -- especially when the agency's own proposal explicitly states that the system had to have been operated successfully elsewhere?

Booz Allen Hamilton is responsible for "selling" Muni on Alcatel. After designing Muni's bid requirements for a train control system, Booz Allen insisted that only Alcatel Canada Inc. could do the job. This insistence, at times, bore at least the appearance of bias.

Alcatel's competitors were pre-emptorily disqualified because they would not promise there would be no disruption when the new signal system came online. AEG Westinghouse's bid, for example, was thrown out because the company would not swear that its software could simultaneously control the Boeing and Breda cars during the changeover from old to new signals.

Alcatel, on the other hand, solemnly vowed that the transition would be painless. But Alcatel would not guarantee that its software could keep track of 40 trains at one time, another bid requirement. At this point Alcatel could have been disqualified, too. But rather than toss Alcatel, Booz Allen wrote a report -- after bids were opened -- that recommended an easing of the 40-trains-at-once standard.

Another requirement for getting the job involved past performance: The software had to have been successfully proved in service for two years. Booz Allen Hamilton insisted that a train control system Alcatel was installing in London, England, for the Docklands Development Corp.'s new light railway met this requirement.

Muni and Booz Allen failed to mention to federal officials one unusual fact: The Docklands Railway was almost a toy railroad, when compared to the Muni Metro. At the time, the Docklands system used a driverless train moving a maximum of only 5,000 passengers an hour. At rush hour, Muni has to move 20 times that many riders.

On the basis of Booz Allen's lobbying for Alcatel, the Federal Transit Administration allowed Muni to disqualify AEG Westinghouse's bid and to give the contract to Alcatel on a "sole-source" basis -- that is, on the contention that no other responsible company could provide the system.