the vert stab that came off of the airbus was NOT a maintenance issue. It was what I consider an extremely dangerous rudder authority setup in that type of aircraft.
Let me explain.
In a non fly-by-wire aircraft that use powered assist control surfaces (ie hydraulic actuated controls), by default you get a linear movement of a control surface with movement of the controls. No matter what's going on, if you push the rudders with 10 lbs of force, you get x inches of rudder pedal travel which results in x degrees of rudder movement. To help stability and give more consistent results for any given effort, pretty much all modern aircraft using powered controls will use mechanical or electro-mechanical systems to apply a control ratio bias curve based mostly on aircraft speed. For example, at low speed, a rudder is probably less effective than it is at high speed, so the designer might make a rudder move 10 deg with 10 lbs of force at low speed, and apply a ratio at high speed that results in maybe 5 deg of rudder movement from that same 10 lbs. That's about as far as this sort of thing goes with non fly-by-wire systems, and pilots can very easily understand this and instinctively compensate while flying.
With a fly by wire system, the designer can do all sorts of things beyond merely damping out control movement. They can add a deadband for example, meaning that the rudder won't move at all until the pilot puts more than a certain amount of pressure on the pedals. Then after that, the pedal pressure increase may result in dramatically different actual rudder movements based on anything from airspeed and altitude to whether the gear is down or angle of attack. They can even make it harder to push the pedals, so in some conditions 10 lbs of pressure would give you 6 inches of travel, and in other conditions that same 10 lbs of pressure would give you only 3 inches of travel.
Add up all these various configurations, and the chance of making a really serious error in programming the FBW system add up.
It turned out that in the case of the airbus, there was a HUGE deadband set up in the rudders. That meant that the pilot could progressively input more and more pressure, and NOTHING would happen. Then as he crossed the pressure threshold, the rudder would move. Imagine your AH stick setup increasing and decreasing your joystick deadband during a dogfight, and you can see why I think this is a terrible idea.
On top of that, in the climb regime they were in, the FBW gains were tweaked to give large rudder control movements to relatively small increases in pressure once they got outside of the deadband. Well, as they found out in the F-16, humans aren't very good at calibrating our feet and hands for precise levels of pressure output. So putting 20 vs. 25 pounds of effort into the rudder pedals is pretty much trial and error. In the case of the airbus, when the pilot put in some rudder, nothing happened until he broke out from the deadband, and then with only a relatively small amount of extra pressure on the pedals, the rudder swung nearly full scale. This was obviously too much pressure and the nose swung far in the other direction. From the flight data recorder, you can see that the pilot made very rapidly increasing rudder inputs showing that he was completely out of sync of these huge unintentional rudder inputs.
As the aircraft sideslip angle increased, the huge deadband and twitchy rudder response resulted in the rudder being deflected very rapidly into the windstream farther than the vert stab could structurally handle.
I'm convinced that the large deadband and increased input gain significantly contributed not only to the accident, but it also contributed to a complete inability for any normal pilot to ever be able to instinctively fly that particular aircraft, leading to a breakdown in emergency procedure training and accompanying inapproprate performance during an actual upset event. Supposedly the designers didn't imagine that the rudder would ever be swapped full deflection side to side after a maximum rudder-commanded sideslip angle was achieved, however the goofed up rudder control setup virtually guaranteed that it would happen if the pilot ever felt the need to work the rudders hard in that stage of flight, because of the large deadband and twitchy rudder response after getting out of the deadband.
This sort of thing has resulted in one F-22 crash, and the F-22 system is deliberately set up to give the pilot the greatest degree of control.
I'm not an airbus basher by any means, but I'm convinced that in this case, the fly by wire control laws directly contributed to the structural failure of the vertical stab. The safety margin built into an airbus by limiting aircraft response to control inputs comes with a price, and that price is paid whenever a "hole" like this gets poked in the flight control laws by a real world situation that was not anticipated.
Personally, I want the plane to respond predictably every time. If that means that I must compensate for controls becoming more effective as speed increases (ie. no automatic speed compensation), then in my mind that's better than the flight controls being automatically adjusted behind the scenes to the extent that occurs in airbus aircraft. I can instinctively compensate for controls being more or less effective as my speed changes, but I can't memorize 200 deadband and gain charts and try to guess how effective my controls will be at any particular speed and flight condition. In a regular aircraft, even a fly by wire plane with permissive flight control laws, I only need to remember and compensate for basic aerodynamic rules that every pilot learns from the beginning. Don't haul on the stick at high speed. At low speed the controls lose effectiveness. Rudder is more effective than ailerons at low speeds and high angle of attack. Etc. These become instinctive compensations after as little as 20 hours of flight time, especially if ground school adequately covers aerodynamics and effects of flight conditions on flight control inputs. But trying to anticipate a computer with a complex and limiting set of flight control laws, deadband, and input/output gain curves... Not possible, and IMHO dangerous.
