|
This article outlines the line of attack on the problem of all weather landing which has been adopted at B.L.E.U. and explains why this line has been chosen.
This is a long term attack and will not affect, for example, the facilities available in the next few years. During, that time the implementation of the P.I.C.A.O. recommendations by the manufacture and general use of I.L.S. in Civil Aviation should result in a considerable reduction of the difficulties of bad weather landing.
It is for this reason that the next major step, that of automatic landing, is taking precedence over the development of improved approach aids such as the narrow-beam centimetre system.
PRESENT LIMITATIONS
When landing an aircraft in good visibility, a pilot collects practically all the information needed regarding the aircraft's attitude, position, and motion with respect to the runway from his general view of objects outside the cockpit.
When visibility is bad, this view is replaced by a collection of instruments which define, within limitations, the aircraft attitude and. motion, and further instruments or aural signals which give the aircraft position with respect to an approach path defined by radio signals. This means that the pilot is presented with an extremely artificial set of data with which to perform a difficult manoeuvre.
The scope of present techniques can be summarised by saying that a skilful pilot using the S.C.S. 51 beam approach system on a good site can bring an aircraft to a position some 200-250 feet in height and some 1200 yards from the touch-down point, from which, provided that the ground can then be. seen, a successful landing can be made.
While, when the facilities become generally available, these techniques should result in a considerable reduction of delays and cancellations due to landing conditions, they will by no means render aircraft independent of these conditions.
To the number of occasions on which conditions are actually worse than the minimum usable, must be added those on which the forecasting is inaccurate, or the faith of the control officer or pilot is insufficiently great.
The above broad criticism of present techniques has been made as an introduction to the attack on the problem adopted at the Blind Landing Experimental Unit. Before outlining this attack, however, mention must be made of the idea, often discussed, of presenting the pilot with a 'picture' comparable with that seen when landing visually. The resolution and detail needed for such a picture is not obtainable from existing radio techniques; to obtain it would mean the use of wavelengths sufficiently short to enable reasonable aerial or lens systems to be used, but long enough to penetrate all' atmospheric conditions - two conditions which are possibly irreconcilable.
AUTOMATIC LANDING
Since, therefore, it seems necessary to develop further the less direct methods, the aim at B.L.E.U. is to evolve a system which will provide sufficiently accurate guidance right down to the runway, and to use this system to control the flight path of the aircraft automatically by feeding appropriate signals to an auto-pilot. In this way it is hoped finally to enable aircraft to land automatically under all conditions in which flying is possible.
If it is realised that the pilot's role in carrying out a 'blind' landing is to interconnect the data available from gyroscopes in the aircraft and from radio signals from the ground, and control the aircraft accordingly, it will be seen that to relieve him of this task by carrying out the interconnection by electro-mechanical means is a logical step. A pilot acts as a network which may amplify with variable gain, differentiate, integrate or ignore its input signals to suit the occasion. 'nether he can be replaced by a 'black box' operating according to fixed laws remains to be seen.
THE AUTO-PILOT
The major difficulty which arises in the use of an autopilot during approach and landing is that of obtaining sufficiently accurate flying at low airspeeds. The primary use of auto-pilots up to the present has been to relieve the pilot during long periods of straight and level flying at cruising speed.
As the speed falls, the control surfaces become less effective, and therefore the amount of control movement required for a given error must be increased in order to maintain the flying accuracy.
A considerable amount of information on this point in relation to automatic approach and landing has been obtained by using the Minneapolis-Honeywell 0.1 auto-pilot.
This auto-pilot is of the displacement type in which an attitude or flight path error is corrected by a control surface movement directly proportional to the error. The maximum 'gearing', that is the ratio of control movement to error amplitude, is limited to about unity; i.e. if the aircraft rolls 1° from the datum, 1° of aileron will be applied.
It was soon apparent that this gearing was not high enough to provide really accurate flying at low airspeeds, but the data obtained from a large number of automatic approaches formed the basis for the work at present proceeding, in which the C.1 has been replaced by the Smith's Mk. IX auto-pilot.
The latter uses a 'rate-rate' system in which the rate of application of control is proportional to the rate of deviation of the aircraft. With the Mk. IX auto-pilot it is possible to obtain higher gearings with consequent improvement in flying accuracy.
Early experiments with one of the first models of this auto-pilot revealed that a rather serious weakness had to be overcome before it could be used for automatic approach and landing. This weakness was the loss of roll datum which could occur during turns. Since the primary error signals are derived from rate of change detectors, the datum can only be retained if the system as a whole integrates with perfect accuracy, or if monitoring is employed. But the latter is in fact provided in the auto-pilot by pendulums, but during turns these cannot be used directly, and the accuracy with which the aircraft is brought back to straight and level flight after a turn is dependent on the integration accuracy. This is no great disadvantage in cruising flight, since the monitors immediately start to correct any wander that may have occurred. An automatic landing, however, may well have been preceded by orbiting the airfield in a continuous turn and by considerable manoeuvring during the approach. Apart from the obvious need to be on an even keel immediately prior to landing, loss of roll datum involves considerable sideslip at low airspeeds, thereby reducing the flying accuracy.
One of the most serious effects was the loss of datum caused by incorrect aileron trim. The latter means that a force is needed on the control surfaces in order to maintain the aircraft's attitude. During straight flight, this force can be provided by a signal from the pendulum monitor, but during a turn the necessary signal must come from the rate of roll gyro, giving rise to a loss of datum.
The incorporation of delayed positive feedback in the servo amplifier considerably reduced this effect, but this, together with other minor improvements, did not make the performance quite good enough for automatic approach.
A further modification was therefore carried out involving the use of the sideslip pendulum monitor. Normally this pendulum is used to correct the rudder position so that a correctly correlated turn is ensured over a range of airspeeds.
The auto-pilot can be set so that, at the approach airspeed, an accurate turn is produced provided that the roll datum is maintained. If sideslip develops it is almost certainly due to loss of this datum, and the pendulum signal can be injected into the aileron channel as a correction.
The auto-pilot is therefore 'tuned' for the approach airspeed, at cruising speeds the sideslip monitor applies the necessary rudder correction, while for the approach, the signal from this monitor is switched to the ailerons to correct any loss of roll datum.
With this change incorporated, the performance was entirely satisfactory, and the following figures have been obtained. from the recording of one hundred approaches.
Arithmetical average error from runway C/L, 21 yards, from S.C.S. 51 null line, 15 yards, assuming the error distribution to be Gaussian, the probable overall error with respect to the runway C/L was 16 yards, made up of a
probable error in 51 definition of the C/L of 11 yards and a probable error of the aircraft control of 12 yards. All these approaches were made on the Woodbridge emergency landing strip.
POSITION-FIXING
The broad requirement set down earlier for a system giving accurate guidance down to the runway is not met by any existing system. The overlapping field pattern systems such as B.A.S. and S.C.S.51 suffer from distortion due to reflections from buildings, etc., and are therefore not sufficiently accurate for landing guidance. Even if the guidance provided by a system such as S.C.S.51 were sufficiently precise, it is doubtful whether an aircraft could follow a unique path in the vertical plane at heights below 200 ft.
A further disadvantage is that those systems which provide elevation guidance (S.C.S.51 and G.C.A.) do so by defining a fixed path. This means that varying windspeeds lead to widely different rates of descent for a given aircraft, and consequent variations in engine power settings and hold-off conditions.
The main requirement of a system giving approach and landing guidance in azimuth is straightforward, a signal proportional to the deviation from the runway centre line being required. This deviation should preferably be given in the form of the distance from the centre line rather than as an angular measurement with respect to some point on the ground.
The problem in azimuth is therefore to provide this signal by a system which is unaffected by peculiarities of the site. For landing, this could probably best be achieved. by the use of the inductive field from 'leader' cables, but to provide guidance during the approach in this way would present grave installation and cost difficulties.
Experiments are being carried out in which the deviation from the centre line is determined by a narrow beam ground radar located at the far end of the runway. This equipment provides measurements of azimuth, range and rate of change of range. The first two of these enable the ratio of two tones transmitted over a V.H.F. channel to be made proportional to the distance from the centre line, while the range and rate measurements are combined to provide a 'time-to-go' signal, the use of which is described below.
In the vertical plane, the situation is somewhat more complicated. In order to overcome the disadvantages of existing systems, mentioned above, an attempt has been made to evolve a technique in which the rate of descent is controlled so that it is constant during the approach (10-15 feet per second), but is reduced to a safe value (about 2 feet per second) for touchdown. Emphasis has been placed initially on the latter aspect of the problem, attention has been directed to the use of a radio altimeter since it is almost certainly impossible to obtain the accuracy required from 'a 'field-pattern' system.
Experiments have been carried out using a considerably modified Type 4 Altimeter (2400 Mc/s. carrier, 80 Mc/s. sweep). This altimeter has been used to provide an indication proportional to height and rate of change of height; by controlling the aircraft so that H + k dH/dt is zero, the path followed is exponential. A considerable number of blind touchdowns have been successfully made. The manual landing analyses mentioned earlier show that an exponential path is followed under visual conditions. The 'time-constant' under these conditions is surprisingly short (about 1.5 seconds); using the altimeter, the most satisfactory value was found to be about 4.5 seconds.
During the approach it is planned to combine the 'time-to-go' signal with the desired rate of descent and hence define the required height.
Ground contours, however, constitute a major difficulty; non-linear control, a combination of barometric and radio height measurement, and contour correction in the 'time-to-go' signal, have been suggested as possible methods of overcoming this difficulty.
CONTROL
The main problem in controlling the aircraft automatically from the radio signals is that of providing a satisfactory stabilising term, particularly in azimuth. A measurement of the rate of approach to the beam must be made, and to obtain this in azimuth, use has been made both of the aircraft heading with respect to the runway Q.D.M. and of the differential of the beam error signal. It is not easy to obtain a satisfactory rate term from existing beam systems, and the best results so far have been obtained using heading stabilisation. The latter has the disadvantage, however, that a cross-wind results in the aircraft following a track which makes an angle (a fraction of the drift angle) with the runway centre line. While this can be dealt with under steady cross-wind conditions by slowly changing the heading datum, the fact that the cross-wind frequently varies with height makes this effect a serious difficulty. More attention is now being given to obtaining a smooth differential in azimuth, as fundamentally this provides the better answer. To date, the most satisfactory stabilising term in pitch has been the pitch attitude of the aircraft with respect to a vertical datum; in this case, the need for a smooth differential of the beam signal is not so acute.
A full account of the control equations used and. results obtained is given in R.A.E. Tech. Note No. BL.3 - Mk. IX Automatic Pilot Approach Tests of the S.C.S.51 Beam System.
RECORDING
In order to study quantitatively the effects occurring during flight experiments, and particularly for obtaining landing data, it has been necessary to develop fairly elaborate recording techniques in which movements of control surfaces, attitude changes, engine conditions, radio and auto-pilot signals, are among the many quantities measured continuously and afterwards analysed.