Cape Canaveral, Fla., 14 July 2005—The space shuttle Discovery, which was due to launch at 3:51 p.m. Eastern Daylight Time yesterday, will remain on the ground until at least Sunday, though it is more likely to launch several days after that, NASA says. Yesterday, at about 1:30 p.m., as the Discovery's seven-person crew sat strapped into their seats, ready to launch, an automatic test of a sensor in the liquid-hydrogen fuel tank showed that the sensor had stopped working. The sensor is one of four "low-level" sensors that watch out for an abnormally low amount of fuel in the tank. They would never normally come into play during a flight, as NASA budgets for a generous fuel margin. The purpose of the sensors is to give the shuttle's computers enough time to shut down the spacecraft's three main engines before the last drop of fuel is exhausted. NASA doesn't know exactly what would happen to an engine that ran out of fuel while operating, says Wayne Hale, deputy shuttle program manager for NASA. "We never tested it," says Hale, but he believes it would cause "serious damage."

NASA spotted the sensor problem when a simulation signal was sent to the electronics box that controls the sensor. Immersed in the hydrogen tank, all four sensors were reading "wet," but the simulation signal commanded them to temporarily read "dry." However, one sensor continued to read "wet." It took about 5 minutes of discussion for launch controllers to scrub the launch after that, says Hale.

Just where the problem lies, however, is an open question, and answering it will dictate how much longer the Discovery will remain on the ground. Early suspicion has fallen on the sensor's electronics control box, which is mounted in the aft section of the shuttle. This box provides power to the sensors, adjusts its output signal so it is suitable for digestion by the rest of the Discovery's instrumentation system, and also houses the sensor simulation circuitry.

A similar box failed on an earlier tank test in April. That box was removed, examined and retested—and passed with flying colors. The problem with the box is "still unexplained today," says Steve Poulos, manager of the orbiter projects office. Problems with equipment that are intermittent—here one second, gone the next—are notoriously difficult to diagnose. But a possible clue emerged when later, unrelated routine testing of shuttle components revealed that a particular batch of transistors that happened to be used in the control box's simulation circuitry were subpar. NASA subsequently removed the suspect transistors from all its control boxes except for one—the control box installed on Discovery.

Shuttle engineers are being careful, though, not to leap to conclusions, and until they positively identify the problem they will have to examine the entire chain of circuits connected to the faulty sensor, from the sensor itself, through all the connecting wiring up to the shuttle's computers. Poulos, for one, believes the problem is somewhere in the wiring and not with the control box, saying it looks to him like an "open circuit" is the cause—in short, a broken wire or connection. The propellants in the external tank were emptied last night, and once again the sensor failed. It read "wet" when the fuel level had fallen to the point when it should have indicated "dry," bolstering the theory that something other than the transistors in the control box's simulation test circuitry was responsible.

In any case, NASA officials hope the problem can be solved without having to open up the external fuel tank itself. Although it is possible to open the tank at the launchpad, this has never been done. NASA would instead probably roll the shuttle back from its launchpad to the cavernous Vehicle Assembly Building, some 5 kilometers distant. This procedure would delay the launch of the shuttle by about two weeks.

For prior shuttle coverage see:

Space Shuttle Launch Called Off [13 July 2005]

Shuttle Still Go For Launch [12 July 2005]

This story was updated at 4:15 p.m.