Originally Posted by mm43
but one has to wonder whether it is the visual search for reaction clues, rather than real SS feedback that leads to the "stirring" technique.
That was the core of PF's (Bonin's) initial control problem that night, that and the fact that his initial lateral control correction was far too large.
In the clag and at night, the only feedback he had was PFD movement and 'seat of the pants'. By the time he saw movement on the PFD, the roll rate was significant. Since he was now in Alt 2 law, neutralizing the stick did not stop the roll rate (It was now a 'conventional' aircraft in roll response with somewhat higher responsiveness to stick input.) He had to then make a corresponding opposing control motion to stop the roll.
When you consider the lags that naturally occur in the display, in the human response to that display, in the controls themselves as they are ordered to a new position, and in the airframe attitude as it responds to the control inputs, you have a classical aircraft control stability situation. It can become unstable if certain parameters are exceeded. This is the basis of an earlier comment I made on an initial roll PIO looking like a real possibility based on my examination of the roll data. If you will remember, the third AF447 BEA report simulations addressed the pitch axis performance only and left the lateral channel to later.
Originally Posted by 3rd BEA Interim Report
However, in view of the complexity of such a simulation, it was agreed that, initially, the simulation would be confined to the longitudinal axis, without introducing turbulence.
A number of years ago, the USN Training command introduced a new visual simulator enhancement to their existing A-4 Skyhawk simulators. I was fortunate to have the opportunity to try out the new simulator variant (Mid '70s). The visual display took its cues from the simulator and navigated around a data base representing the local training area. Because computers of that day were relatively slow, there was a perceptible delay (~ 0.3 seconds) in updating aircraft attitude in the display. The delay was especially visible along the roll axis. This delay was vaguely disquieting while flying visually from the external references. One of the features was the ability to fly formation with another simulated aircraft. This did not work well.
When I attempted to fly formation, the lag cause an over response (visual result delayed) which caused me to apply a strong opposing control input (visual result delayed) which caused me to continue to respond in a manner that caused the oscillation to rapidly build.
The loss of roll stability was abrupt and was caused by my change in control strategy (higher gain) as I began to fly formation. After ~4 half cycles of oscillation I switched to the attitude indicator and the oscillation stopped. For those non-flyers reading, the situation was akin to tripping over something, beginning to fall, and looking desperately for something solid to hold onto.
It would be interesting to hear from others that have experience PIO in any of its many forms. It would also be interesting learn how much lag there is in updating the PFD display in the A330.