20
$\begingroup$

While I was watching a cockpit video of an A330 landing in which the pilot was frenetically moving its sidestick, I wander what was the reaction time of this flight by wire system. Indeed, the time for transmiting the signal from the sidestick to flight computer, the time for computer to interpret all its inputs (pilot's input, probes,...) and to decide to act on flight control surfaces, the aircraft's reaction is not instantaneous.

Then, I realize that whatever the transmission system, there are delay between pilot's input and air control surfaces movement (material's elasticity, time for hydraulic fluid to transmit pressure, other mechanism I can't imagine).

Thus my question is: is there a maximal delay between pilot's input and flight control surface deflection to certify an aircraft?

If needed, for the FBW system, a direct law can be considered (no complex computation as flight control surface movement is proportional to input)

If needed, the question can be restricted to airliners flying under FAA and EASA jurisdictions.

EDIT: given the first feedback (comments, edits, answer), I want to highlight this question is not restricted to fly-by-wire (transmitting pilot's input through mechanical links may also induce delay)

EDIT: I think I didn't emphasize enough that this question is only about delay between pilot's input and control surface reaction. I understand that this delay is negligible compared to all other delay, but this is the one the question focus on.

$\endgroup$
5
  • 2
    $\begingroup$ Related: What kind of delay does the A320's fly-by-wire system add? $\endgroup$
    – user14897
    Sep 1, 2019 at 16:27
  • $\begingroup$ It's always seemed to me that the rigid body dynamics @Jimmy mentions tend to dwarf all of the other system delays, particularly at lower airspeeds. In other words, the control linkages could be instantaneous but you'd still see delays of as much as hundreds of milliseconds between stick movement and aircraft response, particularly at the low airspeeds reached on short final and during the flare. That's part of why landing can be such a dance. $\endgroup$
    – stevegt
    Sep 2, 2019 at 5:44
  • 1
    $\begingroup$ Peripheral: Not exactly what you are asking but of relevance to note that engine spool up times can be 6 seconds plus from throttles being "pushed to the wall". A much longer delay than would be useful in many emergency 'Go Around Power' situations. $\endgroup$ Sep 3, 2019 at 12:21
  • $\begingroup$ Remember that in the Airbus, controller input does not control flight surfaces. It merely gives the flight computer a very strong hint of what the pilot wants, but the flight computer decides if the command should go to the control surfaces or not. $\endgroup$
    – PcMan
    Nov 30, 2021 at 18:34
  • $\begingroup$ Very good question, and I can't give a better answer as the one already presented, but would like to add, that you should consider not only the (auto) pilot input -> processing/transmission -> effector (flight control) moves but (auto) pilot input -> processing/transmission -> effector command performed as full loop. Since e.g. engines, even though they respond to a change fast, do not reach a (thrust level) within milliseconds, rather with 1,2 seconds delay (in good conditions). $\endgroup$
    – Suijin
    Oct 29, 2023 at 17:18

7 Answers 7

23
$\begingroup$

Excessive phase lag is a direct contributor to Type I Pilot-Induced Oscillation (PIO). Phase lag comes from:

  • Rigid body dynamics of the aircraft (e.g. delay between elevator surface and pitch rate response)
  • Actuators (finite acceleration time between input and desired surface angle)
  • Structural compliance (e.g. cable friction)
  • Transport delay in signals
  • Finite computational bandwidth (e.g. loop closure bandwidth)

From NASA Report 4683, PIO susceptibility can be expressed assuming the pilot is compensatory; that is, the pilot input and the aircraft response would be exactly in phase, except for a constant time delay (across frequencies). This model is expressed as:

$$G(s)=\frac{K}{s}e^{-\tau_e s}$$

where $\tau_e$ is the effective time delay, or equivalently, phase rate as a function of frequency

From its research, it found that an effective time delay larger than 0.3 sec leads to PIO issues. Given a typical pilot time delay of 0.2 sec, this would imply an upper bound aircraft effective time delay of 0.1 sec at higher frequency (around 5 rad/s), end to end.

$\endgroup$
4
  • 1
    $\begingroup$ The exact time depends on flight speed. Slow flight will tolerate a larger delay since aerodynamic eigenfrequencies go up with speed. At a higher frequency the same delay means more phase lag. $\endgroup$ Sep 2, 2019 at 5:45
  • $\begingroup$ @PeterKämpf MIL-F-8785C specifies the desired eigenfrequencies for the rigid-body modes. With the exception of SP, they are speed-independent (but flight phase dependent). The NASA paper went through both fighters and transport category aircraft; it's pretty wide ranging. $\endgroup$
    – JZYL
    Sep 2, 2019 at 6:01
  • 2
    $\begingroup$ That must have been poor research. Human reaction delay is well studied and on the order of a quarter second. With training and good health it may be less, but never below 0.1 seconds. $\endgroup$ Sep 2, 2019 at 11:51
  • $\begingroup$ @PeterKämpf We can narrow down the effective vehicle time delay further by including pilot delay. Amended my answer. $\endgroup$
    – JZYL
    Sep 2, 2019 at 17:31
9
$\begingroup$

This is a classic problem in control system theory. The condition to be avoided at all costs is the case where the pilot's control actions get out of phase with the movements of the plane, so the sidestick-action makes the oscillations worse instead of damping them out.

The two ways that could happen are 1) if there are significant processing time delays in the control system connected to the sidestick and 2) if there are significant delays in the pilot's reactions.

As pointed out above, the control system time lags are tiny compared to the time constants of the plane's responses to aileron movement, etc. and the significant time lag in the overall system consisting of plane + pilot + computer control system is in the PILOT, not the control system.

This gives rise to something called PIO or pilot-induced oscillation, where the response time lag of the pilot pushes the whole system into divergent oscillation- as for example in the case of a pilot porpoising a plane down the runway after bouncing off the runway on his or her initial touchdown.

I do not know if computerized flight control systems contain subroutines that prevent PIO but perhaps Peter Kaempf knows!

$\endgroup$
6
  • $\begingroup$ answer has been corrected- NN $\endgroup$ Sep 2, 2019 at 1:28
  • 2
    $\begingroup$ The best way to prevent PIO is to take the pilot's reaction time out of the loop. The second best is to keep the pilot sober and not tired to shorten that delay. I don't have access to FCS source code but have a difficult time to imagine how else to compensate for the delay from human reaction time. $\endgroup$ Sep 2, 2019 at 6:10
  • $\begingroup$ Good answer but rather than “As pointed out above...” it’s best to refer to the specific answer, even better link to it. As the voting system may mean the answer you’re referring to isn’t always above yours. $\endgroup$
    – Notts90
    Sep 2, 2019 at 10:15
  • 2
    $\begingroup$ @PeterKämpf: One way in which a skilled pilot can prevent PIO is by noticing when it's occurring and taking action to de-excite it — such as (seemingly paradoxically) by deliberately slowing down their control responses and/or reducing their amplitude in order move them out of sync with the oscillatory frequency and to reduce the loop gain below unity. Or in human terms, "stay calm, keep steady, take it slow and easy." In principle, an FCS could also actively detect PIO and adjust the control response to de-excite it — although in practice, the best option may be simply to alert the pilot. $\endgroup$ Sep 2, 2019 at 10:41
  • 1
    $\begingroup$ @IlmariKaronen: Correct – in many cases it would had been best to just keep the stick centered. But try to realize what is going on and have the nerve to respond accordingly in a stressful situation – I cannot blame pilots who were trapped in a PIO. Your desire is to get out of it, so you apply what you have learned. Even if it makes matters worse. The insight comes much later. $\endgroup$ Sep 2, 2019 at 11:47
3
$\begingroup$

There is quite some experience in this in Level D simulators, which have computer generated responses that must match those of the original aircraft, within tight tolerances.

A couple of decades ago, the gold standard for Unix real time host computers was 30 Hz. So 30 times per second, all of the following was computed:

  • Surface deflection from stick input, including cable stretch, oil flow simulation etc.
  • Aerodynamic hinge moments on the surface.
  • Hydraulic hinge moments exerted by the actuators.
  • Aerodynamic forces amd moments on the aeroplane.
  • Inertial response of the aeroplane.
  • Visual system response.
  • Motion system response.
  • All other system states and responses.

With an update rate of 30 Hz the standard was deemed acceptable for Level D zero flight time training, which implies a time delay of 1 frame = 0.0333 sec. So we know that this is fast enough: frequency rate 30 Hz, time delay 0.0333 sec.

As an aside, for present day computers this iteration rate is something to smile at, the code that ran @ 30Hz on a state of the art realtime unix machine runs @ 3000Hz on a Macbook Pro now.

$\endgroup$
3
  • 4
    $\begingroup$ Computation time isn't the only time delay. More pronounced time delays include transport delay in signals and confirmation delays (should they exist). $\endgroup$
    – JZYL
    Sep 1, 2019 at 17:00
  • 1
    $\begingroup$ @Jimmy Indeed. 30 Hz in the simulator computer was fast enough to include all signal and other delays that occur in the aeroplane though. $\endgroup$
    – Koyovis
    Sep 1, 2019 at 17:10
  • 1
    $\begingroup$ @Jimmy As Niels points out, the largest delay is the pilot response. $\endgroup$ Sep 2, 2019 at 5:49
3
$\begingroup$

Is there a maximal delay between pilot's input and flight control surface deflection to certify an aircraft?

Literally, no. The FAA's only pronouncements about latency are about ADS-B.

To measure what you're asking about, a temporal delay is too simplistic. You need something like the system's band-limited impulse response, or its temporal equivalent of a modulation transfer function. And not just from stick deflection to surface deflection, but all the way to rate of change of (say) roll rate. FAA doesn't even try to enforce numbers on the output of that process, never mind the intricacies leading up to it.

If an aircraft's control latency in some respect was dangerously large, the test pilots (or the flight simulators!) would notice it well before certification forms were sent to the FAA.

$\endgroup$
3
$\begingroup$

In the early 1980’s, the Flight Research Department at Arvin/Calspan ATC investigated the issues associated with the digital flight control systems, commonly known as Fly-By-Wire (FBW), which were becoming more common in aircraft such as the F-16, F-18, and Tornado. In that day, computer processing power was a mere fraction of what is available today, and the problem was compounded by complex control systems. The researchers were able to induce latency into the flight control system which was in addition to the inherent latency of the test aircraft, and have test pilots evaulate controllability issues in high stress maneuvers such as landing and target tracking. Their results were published in a paper entitled “Effect of Control System Delays on Fighter Flying Qualities”, papers presented at the Flight Mechanics Panel Symposium on Criteria for Handling Qualities of Military Aircraft held in Fort Worth, US, 19-22 April 1982 AGARD-CP-333 (Advisory Group for Aerospace Research and Development) AD A118596

In brief, what they found was at beyond to 130 msec delay in longitudinal control, and 120 msec in lateral control, pilot perception of controllability was reduced. For every 25 msec extra, the Cooper-Harper Pilot Rating score (1-10) increased by 1.

In flight simulator, we have to ensure that the INDUCED delay in the simulation in not more than the Specification. For military aircraft, that is generally 100 msec for fighter and helicopters, and 150 msec for transport aircraft. The FAA has looser standards, 1550 to 300 msec, depending on the level of the simulator.

$\endgroup$
2
$\begingroup$

For civilian certification there are no specific requirements for certification in the FAA Part 23/25 or in the EASA CS 23/25. But obviously they require aircraft not to be prone to PIOs, even though there is no specific section addressing the issue. As @Jimmy mentioned above time delays in the control system are the main reason for type I PIOs. So designers’ objective should be minimize those time delays as much as possible.

On the other hand military requirements goes a little bit more in detail in terms of certification requirements. Aircrafts are rated as Level 1, 2, and 3 based on the time delays of 0.1, 0.2, and 0.25 seconds in the control system. Obviously, Level 1 being the best.

There is also a requirement in the same manual (Flying Qualities of Piloted Aircrafts) to define time delay in terms of phase lag. And it classifies it according to flight phases, such as takeoff and landing, cruise etc. It starts from 15 degrees and goes up to 60 degrees of phase lag for Level 1, 2, and 3 requirements.

$\endgroup$
0
$\begingroup$

The technical term used is latency i.e. the propagation (or transport) delay between the input (pilot control) and the output (control surface movement). The aircraft designer (or Original Equipment Manufacturer) determines the acceptable latency.

The acceptable latency depends on the type of aircraft i.e. Airlines, General Aviation, or Hobby aircraft, flight control dynamics of the particular aircraft, the systems through which the resultant signal is produced (Pilot Control Sensors -> Flight Control Computer/Mechanical Linkages -> Actuation Unit -> Surface movement), and the critically of the signal (ex: control surface actuation).

For airlines like Airbus (A330), or Boeing (B787), the latency between the pilot control inputs and the flight control surface actuation usually ranges between 50 to 100 msec.

$\endgroup$
2
  • $\begingroup$ Do you have any docs for further readings (I'm curious as you present a latency than can double from one case to another, and how it is decomposed between the different steps you describe) ? $\endgroup$
    – Manu H
    Sep 4, 2019 at 4:58
  • $\begingroup$ The design requirement document of the particular aircraft will contain the latency requirements and this document cannot be disclosed. As specified in the answer the acceptable latency depends on the various aspects, and it is different for different aircraft. Therefore, it is mentioned in a range, and it is not quite big from the complete propagation delay perspective, even though it is double. $\endgroup$
    – ToUsIf
    Sep 4, 2019 at 8:55

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .