Narrative:

Climbing at 500 FPM through FL345 for FL350, the aircraft suddenly and inexplicably rolled left to about 45 degrees angle of bank. First officer (PF) disconnected the 'B' autoplt and regained wings level and leveled off at FL350. I asked ATC about the aircraft 6 mi ahead on TCASII at FL350, and they confirmed it was a B767. I notified them of our left roll, and said I suspected wake turbulence. First officer then reattempted the 'B' autoplt. This time it rolled slowly right. He disconnected it, then tried the 'a' system. That autoplt also rolled slowly right. After disconnecting that also, he hand flew it the remainder of the flight. Our lead flight attendant told me one of our flight attendants had an internal back injury from a cart, but was ok to continue. I notified flight control and maintenance. Callback conversation with reporter revealed the following information: the reporter stated the first uncommanded roll was caused by the autoplt 'B' channel initially but the rolling to the right was 'a' and 'B' channels. The reporter said at first it was believed to be a wake turbulence problem but this proved wrong when the aircraft was manually flown. The reporter said it was strictly an autoplt aileron problem and not a yaw damper malfunction. The reporter stated maintenance replaced the 'a' and 'B' autoplt roll servos and flight control computers. The reporter said he later flew the airplane and autoplt performance was normal. Supplemental information from acn 474465: just leveled off at our cruise altitude of FL350. Autoplt had been engaged for 10-15 mins. Suddenly, aircraft rolled (uncommanded) quickly to 45 degrees of bank to the left. ATC reported another aircraft (B767) level at FL350 6 mi ahead of us. Thought the uncommanded roll might have been caused by wake turbulence from the B767 ahead of us. We re-engaged the autoplt, and it began a slow roll to the right. Disengaged the autoplt and hand flew the aircraft to our destination. Had no more uncommanded rolls while hand flying the aircraft. Still not sure whether it was a wake turbulence problem or an autoplt problem. Callback conversation with reporter revealed the following information: the 'G' in B737-300G stands for 'glass.' this is the standard production model from boeing. Autoplt 'B' was in use when the initial hard roll occurred. Both autoplts 'a' and 'B' exhibited roll at a lesser rate after the initial problem. The air carrier did a flight recorder readout and found no evidence of rudder input associated with the roll. Air carrier maintenance replaced a number of components and the aircraft was given a flight test. The autoplt failed the flight test. The manufacturer was called for help and further component replacement occurred. A second flight test occurred and operation was normal. The aircraft was OTS approximately 5 days. The reporter does not know what components were replaced or what the final fix was.

Google
 

Original NASA ASRS Text

Title: A B737-300 IN CLB WITH AUTOPLT ENGAGED AT FL345 EXPERIENCED AN UNCOMMANDED L ROLL TO 45 DEGS AND REGAINED CTL MANUALLY WHEN THE AUTOPLT WAS DISCONNECTED.

Narrative: CLBING AT 500 FPM THROUGH FL345 FOR FL350, THE ACFT SUDDENLY AND INEXPLICABLY ROLLED L TO ABOUT 45 DEGS ANGLE OF BANK. FO (PF) DISCONNECTED THE 'B' AUTOPLT AND REGAINED WINGS LEVEL AND LEVELED OFF AT FL350. I ASKED ATC ABOUT THE ACFT 6 MI AHEAD ON TCASII AT FL350, AND THEY CONFIRMED IT WAS A B767. I NOTIFIED THEM OF OUR L ROLL, AND SAID I SUSPECTED WAKE TURB. FO THEN REATTEMPTED THE 'B' AUTOPLT. THIS TIME IT ROLLED SLOWLY R. HE DISCONNECTED IT, THEN TRIED THE 'A' SYS. THAT AUTOPLT ALSO ROLLED SLOWLY R. AFTER DISCONNECTING THAT ALSO, HE HAND FLEW IT THE REMAINDER OF THE FLT. OUR LEAD FLT ATTENDANT TOLD ME ONE OF OUR FLT ATTENDANTS HAD AN INTERNAL BACK INJURY FROM A CART, BUT WAS OK TO CONTINUE. I NOTIFIED FLT CTL AND MAINT. CALLBACK CONVERSATION WITH RPTR REVEALED THE FOLLOWING INFO: THE RPTR STATED THE FIRST UNCOMMANDED ROLL WAS CAUSED BY THE AUTOPLT 'B' CHANNEL INITIALLY BUT THE ROLLING TO THE R WAS 'A' AND 'B' CHANNELS. THE RPTR SAID AT FIRST IT WAS BELIEVED TO BE A WAKE TURB PROB BUT THIS PROVED WRONG WHEN THE ACFT WAS MANUALLY FLOWN. THE RPTR SAID IT WAS STRICTLY AN AUTOPLT AILERON PROB AND NOT A YAW DAMPER MALFUNCTION. THE RPTR STATED MAINT REPLACED THE 'A' AND 'B' AUTOPLT ROLL SERVOS AND FLT CTL COMPUTERS. THE RPTR SAID HE LATER FLEW THE AIRPLANE AND AUTOPLT PERFORMANCE WAS NORMAL. SUPPLEMENTAL INFO FROM ACN 474465: JUST LEVELED OFF AT OUR CRUISE ALT OF FL350. AUTOPLT HAD BEEN ENGAGED FOR 10-15 MINS. SUDDENLY, ACFT ROLLED (UNCOMMANDED) QUICKLY TO 45 DEGS OF BANK TO THE L. ATC RPTED ANOTHER ACFT (B767) LEVEL AT FL350 6 MI AHEAD OF US. THOUGHT THE UNCOMMANDED ROLL MIGHT HAVE BEEN CAUSED BY WAKE TURB FROM THE B767 AHEAD OF US. WE RE-ENGAGED THE AUTOPLT, AND IT BEGAN A SLOW ROLL TO THE R. DISENGAGED THE AUTOPLT AND HAND FLEW THE ACFT TO OUR DEST. HAD NO MORE UNCOMMANDED ROLLS WHILE HAND FLYING THE ACFT. STILL NOT SURE WHETHER IT WAS A WAKE TURB PROB OR AN AUTOPLT PROB. CALLBACK CONVERSATION WITH RPTR REVEALED THE FOLLOWING INFO: THE 'G' IN B737-300G STANDS FOR 'GLASS.' THIS IS THE STANDARD PRODUCTION MODEL FROM BOEING. AUTOPLT 'B' WAS IN USE WHEN THE INITIAL HARD ROLL OCCURRED. BOTH AUTOPLTS 'A' AND 'B' EXHIBITED ROLL AT A LESSER RATE AFTER THE INITIAL PROB. THE ACR DID A FLT RECORDER READOUT AND FOUND NO EVIDENCE OF RUDDER INPUT ASSOCIATED WITH THE ROLL. ACR MAINT REPLACED A NUMBER OF COMPONENTS AND THE ACFT WAS GIVEN A FLT TEST. THE AUTOPLT FAILED THE FLT TEST. THE MANUFACTURER WAS CALLED FOR HELP AND FURTHER COMPONENT REPLACEMENT OCCURRED. A SECOND FLT TEST OCCURRED AND OP WAS NORMAL. THE ACFT WAS OTS APPROX 5 DAYS. THE RPTR DOES NOT KNOW WHAT COMPONENTS WERE REPLACED OR WHAT THE FINAL FIX WAS.

Data retrieved from NASA's ASRS site as of July 2007 and automatically converted to unabbreviated mixed upper/lowercase text. This report is for informational purposes with no guarantee of accuracy. See NASA's ASRS site for official report.