Autopilot correction times...

[Andy Rylance]

On autopilots on current commercial aircraft what is the time delay between the autopilot noting an adjustment is necessary to be made to either a heading/speed/level change and the control surfaces making that adjustment? Also what is the tolerance on speed/altitude/heading (i.e. percentage difference than that required) before the change kicks in immediately.

I mean is it something like:

Change noticed from current setting required.
Wait (how long? is it based on percentage of adjustment needed - further out from required amount the quicker the autopilot makes the adjustment)
Move control surfaces to adjust back to altitude/speed required.

I ask this because in turbulence if the "wait" is too short then it potentially makes for a more uncomfortable ride if the aircraft is forever chasing as close to the speed/level/heading as required. I am no expert so just interested in how the autopilot works out what to do and how quickly.

Hope that makes sense...

Andy

[TopBunk]

AR

Without ggiving a definitive answer, it wil likely depend on the phase of flight and the mode of the autopliot/flight director.

On an ILS the response will be much quicker (in general) than tracking a radial.

It probably depends on the phase of flight and the Required Nav Accuracy. In the terminal area with a Required accuracy of <1nm the response wil likely be quicker than when in an area where it is 5nm!

[Cardinal]

In the A320's fixed-thrust mode Open Climb the aircraft will happily allow itself to get 20 knots fast or slow. In a mode such as V/S where the FMGS has flexibility in both pitch and power it does a much better job of maintaining it's targets. In the "soft" Alt Crz mode it stays within 20 feet of the selected altitude. Heading is easy, and deviations are imperceptible. On an ILS the needles are glued in place, airspeed within 5 knots unless it's gusty.

[411A]

...airspeed within 5 knots unless it's gusty.
Indeed.
This is where some pilots can't get a grip on reality.
Example.
Descending on the glidepath, one of my First Officers mentioned...'Look at the commanded speed, ref+40 at 500 agl.'
What he did not realise (but does now) is that the autothrust on the type is extremely accurate...sixteen accelerometers (eight each A/T channel) compare, and adjust thrust accordingly....very accurately.
Alpha mode included.

Of course, the old type is....the superb Lockheed L1011.
Often immated, hardly ever equaled....even today.

A design from nearly 40 years ago...still works good for the folks that fly this truly superb aeroplane.
CATIIIB, included.:)

Sorry, back to regular scheduled programming...:E

[Dual ground]

There is also, on some autopilots at least, a Turbulence mode. When selected this effectively de-sensitizes the system so that it isn't constantly chasing altitude settings. This smooths out the ride somewhat, but is, I think, mainly to prevent excess stresses been placed on the airframe.

[safetypee]

Optimising autopilot control laws appropriate to the aircraft type, stage of flight, and within safety requirements is a complex design and development task.

The control system is usually programmed to provide a ‘g’ or pitch rate response (often modified by speed and altitude) to satisfy a particular error. Usually there is an error dead-band to prevent the system chasing small errors and creating instability. At the trigger value the response has to be sufficiently quick to satisfy the task and/or pilot perception, yet not too fast to be uncomfortable, and remain within, ‘g’ or rate limits for safety. In addition the aircraft has to remain at or near the correct trimmed condition. Advanced systems might cross couple pitch modes with power or consider ‘energy management’ routines, inertial flight path, etc.

The system will have ‘muscle’ limits so as not to overstress the controls/aircraft and also to provide a safety margin. These can change with task or phase of flight.
In reality these limits are far short of what a pilot could apply, thus when ‘muscle’ limited the system will have to re trim the aircraft.

The trim in particular should be rate limited as it has a powerful effect. Most modern systems are required to be fail-safe, e.g. sensed muscle demand too high, insufficient trim, or trim in wrong direction will disconnect all of the auto system. Similarly the trim will have error dead-band and perhaps a duty cycle – a time period between activity or reversal, to prevent oscillatory instability.

Thus ‘how long a time period to correct an error’ … as above it depends on many variables.
Some of the important aspects about auto flight systems are that crews should not expect pilot like response or necessarily pilot like operation, most do not have the necessary intelligence – perception, knowledge, and experience.

Respect auto flight systems. Don’t expect a super human or quick response in demanding situations. Don’t engage the system with parameter errors (unless the mode is target driven) or particularly with an out-of-trim force, autos can and will bite. Treat them like any other crew member, except that they may not communicate with you as clearly, and unlike a human they cannot indicate that they are struggling before they quit.
Autopilots are like dutiful first officers, but they never learn.

[411A]

...and unlike a human they cannot indicate that they are struggling before they quit.


Depends...on the specific type.
On the L1011, it can, if you know where to look.

[sharpshooter41]

@411A
Descending on the glidepath, one of my First Officers mentioned...'Look at the commanded speed, ref+40 at 500 agl.'

Sir, for my sake can you please elaborate the conditions in which you were descending on the glidepath and had ref+40 at 500 AGL.

And what does your manual say about being 40 kts higher at 500 AGL????

[Tight Slot]

I was thinking that too! Vref + 40 at 500' aal calls for a go around on any type I've flown. Even ground speed mini on my posh 332 wont give me that much of a "buffer"