Launch of the shuttle Discovery is on hold until late next week at the earliest, NASA officials said today, to give engineers time to fully troubleshoot the fuel level sensor problem that grounded the spaceplane Wednesday.

The launch team at the Kennedy Space Center had been maintaining Discovery in an extended countdown "hold" while NASA managers debated their options. But earlier today, deputy shuttle program manager Wayne Hale said the work needed to find and fix the problem that grounded Discovery will take several days at least and that launch is out of the question until late in the week at the earliest.

And that assumes the problem can, in fact, be resolved by then.

Discovery's crew, meanwhile - commander Eileen Collins, pilot James Kelly, flight engineer Stephen Robinson, Andrew Thomas, Wendy Lawrence, Charles Camarda and Japanese astronaut Soichi Noguchi - was cleared to fly back to the Johnson Space Center to continue training and to await further developments.

The shuttle must get off the ground by July 31 or the flight most likely will slip into September. The launch window is defined by a complex combination of factors, including the shuttle's abiilty to reach the international space station; the nature of the station's orbit; and a NASA-imposed requirement to launch in daylight and to jettison the external tank half a world away with enough light to see potential foam insulation loss.

NASA managers are studying the possibility of extending the current launch to Aug. 4, but that would require giving up the lighting required to photograph the tank with a camera mounted in an umbilical cavity in the belly of the shuttle where propellant lines enter the engine compartment. That imagery considered a high-priority item and it's not clear that senior managers would be willing to give those pictures up.

Assuming the launch window is not extended, NASA would have find the problem, fix it, verify the fix and close up Discovery's aft compartment by around July 25 to preserve the final three days of the launch period. If the problem is resolved faster, of course, the countdown could be restarted earlier.

Early today, engineers unloaded hydrogen and oxygen used by the shuttle's electrical generators before opening the ship's aft compartment to gain access to the fuel sensor electronics that route sensor data to Discovery's flight computers.

Discovery was grounded Wednesday two-and-a-half hours before blastoff when a routine computerized test revealed problems with one of the four hydrogen fuel engine cutoff - ECO - sensors in the shuttle's external fuel tank.

The sensors are part of a backup system intended to make sure the ship's engines don't shut down too early or run too long, draining the tank dry with potentially catastrophic results. All four sensors are required for a countdown to proceed. Here's a recap of the problem to date:

ECO SENSOR BACKGROUND

NASA has encountered a string of problems in recent weeks with the ECO sensor system in Discovery, glitches that have proven to be surprisingly difficult to resolve. The problems began during a tanking test in April when ECO sensors 3 and 4 operated intermittently.

Engineers removed an electronic controller, called a point sensor box, from Discovery and replaced wiring to the two sensors in question (wiring to sensors 1 and 2 wasn't touched). But the controller checked out OK and troubleshooters were unable to trace the cause of the problem A point sensor box from the shuttle Atlantis was installed and a second fueling test was conducted.

This time around, the sensors worked normally. But during additional post-test troubleshooting, the replacement sensor controller box malfunctioned. It was replaced by one taken from the shuttle Endeavour. NASA already had decided to replace Discovery's tank to address ice debris issues. With a fresh controller, replacement cabling, a new tank and solid test results, NASA managers decided to treat the sensor issue as an "unexplained anomaly" that presumably had been fixed.

But during Wednesday's countdown, the No. 2 low-level hydrogen sensor failed to switch from "wet" to "dry" during a test in which computers send signals to simulate a dry tank. When the tank was drained, the other three sensors changed from wet to dry as expected. The No. 2 sensor remained "wet" for another three hours before switching back to "dry."

The engineering teams assessing the problem are reviewing the history of the sensors, cabling and control electronics, changes made in the fuel tank since Columbia's ill-fated flight in 2003, changes made to Discovery since it last flew in 2002 and any procedural changes that might possibly have an impact.

At launch pad 39B, meanwhile, engineers plan to verify the cabling leading from the point sensor box to various connectors and ultimately, into the tank through a complex umbilical connection.

If a problem is found in the point sensor box, NASA would still be faced with explaining the sensor problems encountered during the tanking test in April, the original "unexplained anomaly" that NASA managers accepted going into Wednesday's countdown.

Many engineers believe the problem may be related to one or more changes made in the wake of the Columbia disaster, but at this point, that remains speculation.

Additional coverage for subscribers:
VIDEO: THURSDAY'S UPDATE NEWS CONFERENCE DIAL-UP | BROADBAND
AUDIO: LISTEN TO THE NEWS CONFERENCE MP3 FILE

VIDEO: POST-SCRUB NEWS CONFERENCE PLAY
VIDEO: WEDNESDAY'S LAUNCH IS SCRUBBED PLAY
VIDEO: ASTRONAUTS DEPART QUARTERS FOR LAUNCH PAD PLAY
VIDEO: CREW DONS LAUNCH SPACESUITS PART 1 | PART 2
VIDEO: ASTRONAUTS GATHER FOR PRE-LAUNCH SNACK PLAY
VIDEO: PAD'S ROTATING SERVICE STRUCTURE ROLLED BACK PLAY
MORE: SHUTTLE RETURN TO FLIGHT VIDEO LISTING
SUBSCRIBE NOW