The situation with the stations robot arm, meanwhile, took an unexpected turn Wednesday and Thursday as ground engineers continued troubleshooting a problem with the shoulder joint of the crane-like device.

Delivered to the station in April, the 57.7-foot (17.5-meter) arm and its shoulder joint have been working as advertised when its prime control and electronics units are being used to operate it. But over the past several weeks, the shoulder joint has failed to operate properly when back-up units were being pressed into service.

Ground engineers thought the problem might be a faulty back-up electronics unit. So diagnostic software designed to pinpoint the cause of the problem was beamed up to the station Wednesday.

Later that day, the shoulder joint once again failed when engineers began testing the arms back-up systems. But then the problem disappeared, and the shoulder joint performed as expected, during 12 additional tests that were run Wednesday and Thursday.

Perplexed engineers now are trying to figure out how to explain the sudden recovery. And Curry said U.S. and Canadian engineers want to understand the situation before launching Atlantis and its astronauts on the airlock installation mission.

"When we get these situations that occur where a failure manifests itself for some period of time and then all sudden just goes away magically well, nobody likes magic in the flight control world," he said. "We want to understand whats causing the problem."

Considered the heart of Canadas $900 million station contribution, the robot arm must be working properly in order to attach the 12-ton airlock to the station. The shuttles shorter robot arm is not long enough to do the job.

In order to operate the crane, computer commands are sent from a robotic workstation inside the station to a computer control box located near the elbow of the arm. Those commands then are relayed to an electronics unit on the arm joint that operators are attempting to move.

Engineers now think that the shoulder problem likely cropped up because commands for some reason were not being routed properly between the back-up computer control box and an associated joint electronics unit.

But they still are trying to figure out both the root cause of the problem and why it suddenly disappeared.

The plan now is to develop a software patch that would mask the communications problem between the back-up computer control box and its linked electronics unit, enabling the arm to operate even if the trouble recurred.

A second software patch also is being developed to enable back-up systems to operate the arm even if the shoulder joint were to fail. Under normal conditions, a failure of any of the arms seven joints would cause it to automatically shut down a situation NASA wants to avoid during the airlock installation job.

NASA, meanwhile, is facing a mid-July deadline for launching the Atlantis mission.

The flight cannot be launched between July 17 and August 4. The station during that time will be flying in an orbit that would expose a docked shuttle to high temperatures that could foul spaceship systems.

Sistership Discovery now is scheduled to launch Aug. 5 on a mission to ferry a new crew to the station and then return to Earth with its current residents, outpost commander Yuri Usachev and flight engineers Susan Helms and Jim Voss.

An Atlantis launch delay beyond July 16 likely would prompt NASA to reverse the order of the two upcoming shuttle flights, with the airlock installation mission being postponed until September.