Rwy20 wrote:
Here, you clearly misunderstood (or are deliberately misquoting?) what I wrote.
I would never deliberately misquote anybody, really! And especially not on a friendly forum like this. I guess that I rather misinterpreted what you wrote, possibly based on actual behavior of pilots I flew with who I saw doing similar mindless things with their autopilots.
When instructing, especially students who I know will mostly fly privately/single pilot later, I deviate from the syllabus a lot and try to tell them as much as possible about autopilot/flight director usage, operating modes and caveats. Because this is what they will be doing 95% of their future flying time: operating the autopilot. Of course this is difficult do to with the typical training aircraft which either has no autopilot at all or a very basic one. But one can point them in the right direction.
Rwy20 wrote:
If you get into a situation where your plane does something which you don’t understand, whilst on autopilot, then you shouldn’t throw your automation out of the window altogether.
Again, with a caveat. When your autopilot starts doing strange things, this might as well be caused by faulty electronics and hardware of the autopilot itself, in which case selecting a different mode might not change anything but wastes precious time instead better used for getting control of the aircraft back. On the planes I fly, one of the few memory items in the abnormal and emergency checklist is: “Autopilot malfunction: 1. Control Wheel: Grip 2. Autopilot: Disengage”. And not: Autopilot: Select heading mode…
On many of the plane/AP configurations I have seen, it is:
2: A/P disengage & trim interrupt … press and hold
because one of the possible faulty electronics and hardware problems could be a trim runaway, and the trim motor is only de-energized as long as the red button is held down.
huv wrote:
because one of the possible faulty electronics and hardware problems could be a trim runaway, and the trim motor is only de-energized as long as the red button is held down.
Yes, ouf course. Cessna’s manual only says “PRESS”, but it was demonstrated to me more then once in the simulator that if you don’t actually hold it during the time you need to locate the pitch-trim CB (who remembers where that one is placed…) it will keep trimming and trimming…
what_next wrote:
Autopilot malfunction
This is the bit where maybe you are diverging from other’s understandings.
If, for example, I press “Activate Approach” and the aircraft turns right rather than left, I don’t consider it an autopilot malfunction. I assume that something has gone wrong in the programming, so I fall back to HDG and ALT, get it going in generally the right direction, then sort out the programming.
I do have a real autopilot malfunction story, though. 
I was PF P2 and the Captain was making a PA. As he spoke, the autopilot pitched up and rolled rapidly into a 45° bank – pretty much entered a barrel roll. I disengaged and gently rolled it level, then turned back on course and gently descended to altitude and continued to hand fly (as one would.) What really impressed me was that if you had made a recording of that PA, you would absolutely not know that anything had happened, let alone that the aircraft had entered quite a violent deviation; not the slightest hesitation or intake of breath. That’s cool captaincy for you!
Timothy wrote:
If, for example, I press “Activate Approach” and the aircraft turns right rather than left, I don’t consider it an autopilot malfunction.
With ours, this can be normal operation when the autopilot “thinks” that the intercept heading is too shallow. At other times it will start to climb towards the glideslope. These are not programming or operating errors made by the pilot, but errors in the internal programming of the autopilot itself which I treat as a malfuncion.
Well, to use that interpretation as the basis for “one of the stupidest things you have read on the internet” is remarkable.
Going back to the headline of the thread, I think the traps seen in the use of autopilots come in two levels:
What I mean is that vertical problems get serious very easily, in one of three ways.
The most common in my experience is climbing in VS mode with a non-turbo’ed piston engine and not realising that the A/P slowly trims all the way back trying to maintain climb rate, eventually stalling the aircraft with full power, trim all the way back and the pilot completely out of the loop. I have seen that coming many times with others, and I did it myself once in a Bonanza at 2 a.m., very unpleasant. The problem is best solved by having an A/P with IAS mode (or FLC mode or whatever it is called) and always using that mode for climbing.
The second vertical problem is when the pilot tries to “help” the autopilot with elevator. The autopilot does not know pilot input from aerodynamic input, and trims against the input, i.e. against what the pilot wants. When the fight is over, the aircraft will be out of trim. A VIP passenger in a Dassault jet broke his neck hitting the cabin ceiling in negative g’s following this scenario several years ago. Apparently even some bizjet pilots are unfamiliar with the concept that only one pilot should be flying at any time, so if you want to take control, always take the autopilot out of the loop first.
The third vertical problem is descending too far. I have both heard of, and seen, instances where the pilot thinks he is selecting 1000 fpm vertical speed, but actually has preselected 1000 as level-out altitude for the descent. Of course this is mostly a problem on the systems where the same displays and buttons are used for VS and altitude, like the KAP xxx and some KFC systems.
Horizontal problems seem to be mostly about understanding the NAV mode. Selecting HDG may work well, especially if the pilot has the good habit of always having the hdg bug aligned with where he wants to go. Even if he initially overpowers the autopilot in NAV mode, no catastrophe is imminent, unlike when overpowering the elevator/trim-servos.
I remember reading (in an old issue of Flying magazine, probably an article by Richard Collins) that one-axis autopilots seemed to improve the safety for the average GA pilot. In contrast, autopilots capable of controlling elevator and/or trim, i.e. having vertical modes, seemed to reduce safety on a statistical level. Of course, some contemporary autopilots with FLC and envelope protection may alter that equation, but what I have seen pilots do in a number of different acft/ap configs does support that verticals modes bite the hardest.
The most common in my experience is climbing in VS mode with a non-turbo’ed piston engine and not realising that the A/P slowly trims all the way back trying to maintain climb rate, eventually stalling the aircraft with full power, trim all the way back and the pilot completely out of the loop.
That’s one of the big advantages of both the GFC700 autopilots and the DFC90/100 series. They both have an IAS mode for climb and their envelope protection (although speed based only) will prevent the airplane for, stalling (or yell at you when hand flying/flight director)
The second vertical problem is when the pilot tries to “help” the autopilot with elevator
Now that is a really bad idea. Either the A/P flies the airplane or both. I only saw that with beginners who don’t trust the A/P and are scared to let go the yoke.
Cessna’s manual only says “PRESS”, but it was demonstrated to me more then once in the simulator that if you don’t actually hold it during the time you need to locate the pitch-trim CB (who remembers where that one is placed…) it will keep trimming and trimming…
This is likely implementation dependent (for sure I would expect a Collins Proline to be smarter than a King autopilot) but on the King ones the red disconnect button has two contacts: one is a signal to the autopilot software, and the other carries the current to all the servo clutch solenoids. The software could obviously have crashed, so removing power from the clutches disconnects all axes and enables manual flight, while reaching for the autopilot Master switch! (which also removes power from the clutches). So the red button has to be held down continuously. When it’s all sorted, you probably should pull the autopilot system CBs (as I have done a few times after a burnt-out KFC225 servo filled the cockpit with a nice smell 
) Details here.
The most common in my experience is climbing in VS mode with a non-turbo’ed piston engine and not realising that the A/P slowly trims all the way back trying to maintain climb rate, eventually stalling the aircraft with full power, trim all the way back and the pilot completely out of the loop. I have seen that coming many times with others, and I did it myself once in a Bonanza at 2 a.m., very unpleasant. The problem is best solved by having an A/P with IAS mode
Yes; that one has caught me out a few times when climbing to ~ FL200 and busy elsewhere. But you should just get the stall warner and that will focus your attention fast
Doesn’t the IAS climb mode only postpone the issue? One starts a typical (IFR profile) climb with say 120kt IAS (for CHT management) and this then bleeds off, to some 85kt IAS at FL200. So what constant-IAS value would you climb at? At say 90kt the CHT will go too high and you have no forward visibility at low level which is not good. And at say 120kt, the IAS mode will reach the stop at something like FL100.
Obviously a turbo engine will be better and the above scenario will get pushed out before it happens, but it will still happen. So one has to pay attention.
In contrast, autopilots capable of controlling elevator and/or trim, i.e. having vertical modes, seemed to reduce safety on a statistical level.
I would think that depends on what the comparison baseline is. If you took 100 fresh PPLs from the C152 scene and gave them an SR22, without a doubt you would find that an SR22 is really dangerous, and a G1000 makes it even worse
So there are so many factors e.g. a good size group which gets into a plane whose systems they don’t understand. To me it seems clear that an AP which can fly an ILS improves safety considerably for a pilot who knows how to use it.
It would be a good discussion whether it is right to argue that a system degrades safety when in fact it is because many users don’t understand it. We have a training system in GA which for the most part has no “type rating” and people can get into planes which they don’t understand. But that makes it very difficult to get data on safety which is meaningful to a newcomer to the scene who is diligent.
Peter wrote:
Doesn’t the IAS climb mode only postpone the issue?
I would think not, although I have never climbed that high in a non-turbo’ed acft using IAS/FLC mode.
If you forget adjusting IAS as you climb, your problem will only be that you loose climb performance (eventually level out prematurely), not that you stall at full power and spiral out of the sky, out of control. As long as you select an IAS above Vy, nothing really bad should happen.
