Honeywell suggested that the reason for the rapid servo failures is rapid motor brush wear, which is caused by a different material being used in the brushes which could cause them to wear out in as little as 60 hours. I pointed out that the last roll servo failed in much less airborne time than that; he said that Honeywell believe the brush wear could be even faster if there is noise on the servo analog input, causing the motor to oscillate rapidly and wear out the brushes. He proposed checking the cable shielding.
So I tested this hypothesis, by making a T-adaptor which goes inline with the 14-pole roll servo connector, and via a screened cable connects to a scope. I monitored both the main input and the 'reference' input. Due to heavy rain I was unable to do the test with the engine running (would need to be outdoors, and I needed mains power for the oscilloscope).
I immediately saw a waveform closely approximating a sinewave superimposed onto the main input signal, 140mV P-P at 1.4kHz. It isn't always there; during power up tests the DC voltage switches between 0V and +2V, with some larger transients; during this time the sinewave sometimes appears, its appearance being synchronised with particular phases of the power-up test sequence. When the tests are complete the servo input reverts to 0V, and at that point the the sinewave is present and remains so. If the autopilot is engaged (by pressing AP) there is no change in the DC voltage, and the sinewave remains present too. Clearly this sinewave is coming out of the KFC225 computer because there was nothing significant happening in the rest of the aircraft when it appeared or disappeared. There is no AC signal on the input 'reference' signal; just some DC shifts during the power-up tests.
An examination of the KFC225 schematics shows that the sinewave is intentional and is the product of a lowpass filter which is used to convert a PWM signal from within the KFC225 into the "DC" control voltage which goes to the servo.
There were no transients on the servo control voltage, and I checked this with all avionics (including DME) operating.
The failed servo was opened up and the motor was found to be working perfectly, so a problem with brushes is not the explanation in this case.
30 April 2004