It depends on your research question Andreas, and whether you are looking for an academic or practical answer. In practice, for an Airbus, the main determinant is probably turbulence/air movement: apart from very minor changes required to account for longitudinal stability, the aircraft could be set up and trimmed at top of descent in a 3 deg glideslope and, in completely smooth air, the only required flight control inputs would be elevator to adjust for configuration changes (gear, flap & associated power). For academic purposes, and in a simulator, you could introduce a deviation and see how long/what amplitude inputs are required to return to glideslope/centreline. A typical way to do this might be to set the aircraft up at 1500' AGL, on speed, on glideslope, offset 500' from centreline and go from there. If you are looking at control inputs, in addition to controlling for deviations from flight path you would need to measure/control for deviations from approach/landing speed (suggest if exceeding +/- 2kt desired), and deviation of touchdown point from desired (suggest anything outside 1000 +/- 500 ft long, +/- 25 ft lateral). 

Hi Andreas!

I fly Airbus A320 and I do hope to be helpful in your research.

Well, flying an ILS manually with the Airbus is quite easy because the aircraft is autotrimmed. This means that any input you give to the sidestick, has a reflection on the trim wheel, thus, it stabilises the aircraft into the new attitude.

Theoretically, you can impose a vertical speed (or, better, a descent angle) and then the aircraft will descent flying the imposed angle. The corrections made by the pilots occur, as Robyn said, only when there is a change in the configuration: lowering the gears increases the resistance and modifies the CG, thus a small pitch up correction is required, extending the flaps creates a ballooning effect that requires a pitch down correction.

If you have a standard ILS approach, you will fly a -3° path. At the approach speed, this correspond more or less to -700 ft/min. Airbus said that any gust of disturbance is automatically corrected by the system.

Well, this is theory...

Practically the system is sometimes lazy and pilots should intervene to correct the path to compensate vertical and horizontal wind variations and thus, maintaining the correct flying path.

How I do this? In the Airbus' fly-by-wire it is important to apply small corrections, then wait few second to let the system to operate and then check if the imposed target is reached. Any input you give to the sidestick is a "demand" (left/right inputs are roll demands, forward/backward inputs are pitch demands). Since it is digital, this "demands" can be asked by small inputs lasting from half to 4 seconds (the latter during the rotation). This is true when the aircraft is flying in the so called "normal law".

If the systems are degraded, it can happen that you have to fly in "direct law". in this mode, flight controls are powerful, because are connected in a quite direct mode to the inputs given to the sidestick.

It is extremely important to avoid the overcontrol. This can happen in gusting conditions. The gust hit the aircraft and move her on one or more axis: both the pilot and the system feel this movement and react. The reaction of the pilot is quite immediate while the one of the system is delayed of about one second or more.The system correct the gust, but the pilot did it before. So the result is that the aircraft has too much corrected.both the pilot and the system realised it and both apply the correction. This will start a sequence of diverging oscillations around the target. To stop them, the pilot should force himself to leave the system to operate, thus to leave the sidestick in the neutral position. This is very dangerous, especially during the flare (look to the accident happened at Lufthansa a couple of years ago: they damage the wing during the go around, because they hit the ground with the wing tip).

Another important clue to fly a good ILS manually is the management of the thrust.

In normal conditions you can fly a nice ILS without the help of the autothrottle by setting a thrust of about 50% N1 (on CFM56 engines). Once the descent angle and the thrust are set, speed and vertical speed are obtained.

If you need any other information, feel free to contact me!

Best regards!

Marco

Lots of interesting work out there - NASA and NRC Canada have done a lot of workload studies in the past. Are you looking at the frequency domain or time domain? There is a generally accepted view that 0.2 Hz control inputs and below are used to modify flight path vector (0.1 in some studies but in my view it is dependant on the aircraft handling qualities and rate response), whereas 0.2 and above are associated with maintaining the desired attitude (compensation for turbulence / handling qualities). Of course control strategies above the cut off frequency of the control system achieve nothing (in flight test taken as when the control input / aircraft response phase difference exceeds 180 deg) and are often accosted with either panic, impending loss of control or application of control strategies inherited from previous more manoeuvrable types (fighter pilot in second career flying heavy metal). Good luck!

Dear Jamie, Robyn, Marco, Simon!

Thank you for your answers. It’s correct that flying an Airbus, especially the A320 is rather simple, especially in theory and in perfect weather conditions. However, ILS flight path deviations measured in a flight sim experiment with A340 pilots in their familiar aircraft tell a completely different story. The very extreme cases are AF447, Asiana 214, but I have already seen remarkable variability in manual flight performance in typical A340 crews, as well as I assume the same effects in many long-haul crews.

I did a sim experiment measuring ILS deviations below 3,000 ft AGL. We have seen significant effects of practice on the manual flight performance and that stuff should be published soon. Here I consider flight performance as the result of the human operator and an Airbus in normal law without autopilot and without flight director but with available autothrust (nearly raw data).  

The next step of this analysis is about pilots’ stick inputs. Can I find systematic patterns in these inputs? Can I match such patterns to different control strategies? (At the moment) it’s not about time or frequency domain. I’m averaging inputs over an interval of several minutes to get a more summarizing picture of what a pilot has done.       

Great discussion already going on... Just a couple of additional thoughts on purely manually controlled flight. In essence, the pilot inputs are dependent on an interaction of three control circuits in accordance with the control and performance concept in instrument flying. Pilot control inputs are made to affect a known pitch, power, and heading change on the control instruments (ADI for pitch and bank input, Engine Instruments for power inputs), which constitutes the first control loop of input, resultant effect, and feedback and in fully manual flight depends greatly on speed and control surface effectiveness, while in automation assisted flight (such as flight control computer, auto-throttle, etc...) will be much simpler (because, for example, the flight control computer will handle stick inputs as desired pitch or bank angle requests rather than actual control surface deflection inputs). The pilot will continue to manipulate controls until the desired control instruments match expectations, and will then continue corrections to hold those inputs.

Then in a second control loop, the effect of the pitch,bank, and power inputs is assessed on VVI, ASI, and HSI to meet desired/planned numerical results for the anticipated flight condition (i.e., the planned ILS glidepath, course, and approach speed). This second loop just establishes that control inputs of the first loop had desired performance results and will feedback into the first loop to make known/distinct corrections if necessary. What this second loop doesn't do yet is correction for the actual ILS, which is part of a third loop that feeds back into the second control loop.

In the third loop, resulting deviations to ILS glidepath and course are assessed, determining the effect of desired/planned vertical speed, heading, and airspeed. Deviations from course or glidepath are translated into required corrections for heading and vertical speed and are fed back into the second loop as new desired/planned numerical values for those. Since that will result in a deviation within the second loop, it will in turn be translated back into the first loop to establish known pitch, power, and bank inputs to meet the newly desired VVI and Hdg values.

When making corrections, the pilot should in general follow a dampened oscillation curve; i.e., teaching manual ILS flying, I would recommend halfing of corrections: For example, if a certain hdg drifted us of course and a hdg change of six degrees corrects us back to course, I would recommend to take half of that correction out and leave three degrees correction in as the newly desired hdg to hold course; if that in turn slowly drifts us to the other side, I would take the three degrees out until back on course and then re-establish half of it (1.5deg) as new on-course hdg. Similar for required VVI updates...Also, when making the required control inputs, the pilot has to be aware of and make known corrections; e.g., for above small value hdg changes, I would recommend a bank angle equal to the hdg change amount, which keep the amount of time that input has to be held roughly the same for all corrections...

This use of a second intermediate control loop is in contrast to automated control inputs such as flight director bars, which cut out the requirements to establish new headings and VVIs and directly translate ILS deviation inputs into bank and pitch steering commands. Therefore, a FD will do dynamic control input corrections while manual flying requires incremental, but discrete inputs that are held for a certain amount of time. The biggest mistake students new to manual instrument flying will make is to try dynamically doing corrections - i.e. chasing the needles...

I hope that perspective from the training side may help as input.

Why not use the DFDR data to analyse the research question?

Or record the data during a simulator session or two to determine what measurement you want to use.  The extraneous variables of wind, turbulence, pilot experience and pilot skill level, and perhaps physical condition may all be factors.  I was a 320 check inspector for 18 years and often saw pilots over controlling by using the side stick almost like a cake mixer in motion when manually flying, even in almost calm wind conditions and I put that down to lack of real understanding of the system and of lack of manual flying practice.  The 320 ILS approach is really easy to fly in normal law with full FD guidance.  Now, if you measured this same set of events using raw data ILS with FDs off and Track / FPA mode selected  and no AT or AP and in alternate or direct law modes  there will be a very different set of results.Then try it with an engine out. Chasing the needles is the least of the problems, chasing the fly by wire and dealing with control lags becomes the problem and pilot induced oscillations are not uncommon.

Great additional inputs on control lag and response and possible PIOs... I agree, especially for formation flying (but I guess, also for instrument) I would often find myself recommending to 'think fast but act restraint' (i.e., doing half the input  and giving it double as much time of what a student may think is required, precisely to overcome those over-control tendencies...) Greg, from your inputs it seems, heavier aircraft type application is (as would be expected) much of an exaggeration of what can be found/identified in fighter type aircraft as basic root causes...

That exaggeration would probably be due to the greater lag time for feedback in the control loop in transport aircraft compared to fighter aircraft, or due to longer control coupling moment arms and to the greater inertia in much heavier aircraft  that causes an apparent control response delay. It also may have been due to the EFIS and flight computers lagging a little when the FMGC was not operating in normal law. It always felt awkward in direct law.

I suppose that by purpose design the man machine interface provides much less feeling of separation from the machine in very high performance aircraft. I sometimes felt out of the loop or at least well behind it when operating in direct law in an Airbus. It always felt like I had to be over controlling, particularly in pitch control.  As we only flew in direct law in the simulator some of it could have been due to simulator artificial response and simulator lag as test pilots often told me the aircraft actually flew much better than the simulators when in direct law.  I never flew the aircraft in direct law so would have to accept that as an expert opinion.

Your advice of think fast but act restrained was good advice as I often compared the fly by wire to flying helicopters, just gentle pressures and no full fist grasping of the side stick, For what its worth, I used to rest my hand almost flat on the console surrounding the base of the side stick and then just pressure the stick rather than making greater inputs unless I wanted to make a large change of direction. It made for smoother and more accurate instrument approach flying.

Can i add general points on simulaion so that one may start thinking differently?

Simulation only allows one to estimate the performance of an existing system under some projected set of operation conditions.

many simulation studies have been composed of adhoc models, adhoc implementation and just run the simulator once to obtain the answer, attitude.

This attitude, which neglects the important issue of how a properly conceived and coded model should be used, has doubtless led to that erroneous conclusions have been drawn in many simulation studies.

Stochastic simulation models produce output that is itself random, and must therefore be treated as only an estimate of the true characteristics of the model.

This is one of the disadvantages of simulation.

moreover, well-defined set of objectives at the beginning of the simulation may not be possible. 

There are some reports to review that describe approaches to looking at pilot control strategy. It can generally be cast as a feedback control system whether a continuous tracking task or a discrete maneuver. See the following for some examples. Most important, look at it as a set of multi-axis nested loops that are separated by their crossover frequencies. E.g., for a hovering helicopter, inner-loop consisting of attitude, middle loop consisting of velocity, and possibly outer loop with a position command. Such a classical control approach is more meaningful to a pilot than is a modern control theory approach. And it can be more easily identified in system terms.