WO2009136152A1 - Lighting - Google Patents

Abstract

A PAPI light (fig.1) is operable to emit a beam of light comprising red light in a first sector of the beam and white light in a second sector of the beam. The white light has a colour temperature greater than 3500K. The white light may have a colour temperature of substantially 4500K. The white light sector of the beam may be produced by a bulb, and the colour temperature of the light produced by the bulb can be adjusted using a filter, in some embodiments. The filter may be operable to raise the colour temperature of the light produced by the bulb.

Description

LIGHTING

The invention relates to lighting, and particularly, but not exclusively to aviation lighting. By 'aviation lighting' it is meant lighting for providing information and/or signals to pilots and other flight crew, both on the flight deck and externally to an aircraft.

The use of colour in aviation lighting is important for the efficient coding of signals and information. Provided that an observer can make use of colour signals, the use of colour enhances visual performance. In most countries, the colours of aviation lights that can be used are regulated. For example, Figure 8 shows the chromaticities ranges of lights recommended by the Civil Aviation Authority for use in an aerodrome in the UK.

Because of the widespread use of colour in aviation signals, there is a need for pilots and other flight crew to be able to discriminate and recognise all the different colours of lights used. This may not be possible for someone with colour deficient vision. Therefore, there is a need to set colour vision requirements for flight crew to ensure that flight crew are able to recognise all the light signals used in aviation. Humans with normal trichromatic (three colour) colour vision possess three distinct classes of cone photoreceptors. These contain short (S, 'blue'), middle (M, 'green') and long (L, 'red') wavelength sensitive photopigments with appropriate peak absorption wavelengths (λ_max) . Generally, short photoreceptors detect light ranging between 400 and 500nm (with λ_max at 420-440nm), middle photoreceptors detect light at around 450-630nm (with λ_max at 534-545nm) and long photoreceptors detect light at 500-700nm (with λ_max at 564-580nm). Variant L- and/or M-cone genes can cause significant shifts in λ_max and this in turn can cause large changes in chromatic sensitivity. In addition to λ_max changes, other factors such as the amount of pigment present in photoreceptors can also affect chromatic sensitivity. Red-green deficiency is the most common type and is caused by either the absence of or abnormal functioning of L- or M-cones. The corresponding condition is normally described as protan or deutan deficiency, respectively. Colour vision deficiency affects approximately 8% of men and less than 1% of women.

Colour is used extensively to code information in the aviation environment and pilots are normally expected to have good colour discrimination. Even when other cues are also available, the ability to use colour information increases redundancy and in some situations it improves considerably the level of visual performance that can be achieved. Some accidents have been linked to loss of colour vision. There is also some evidence to suggest that the likelihood of accidents is increased in pilots that are colour deficient. Other studies have shown that subjects with colour vision deficiencies make more errors and are slower in recognising aviation signals and colour coded instrument displays. Aviation accidents have high social and economical costs, especially if the accident involves large passenger planes. An important strategy in achieving high levels of safety in aviation is to build redundancy in equipment and the interpretation of signals and other information by pilots and other personnel. However, there are also a small number of tasks when there is no redundancy and the correct interpretation of colour signals becomes very important.

Occupational colour vision standards were introduced in aviation in 1919 by The Aeronautical Commission of the International Civil Air Navigation Authority. Effectively, these standards screen for any colour vision abnormality, and subjects found to have such abnormalities are excluded from flying. In Europe, the 38 members of the Joint Aviation Authority (JAA) have agreed to apply the same colour vision standards for flight crew. The current JAA colour vision regulations use the Ishihara pseudo- isochromatic test as a screening test for colour vision. The JAA use the first 15 plates of the 24-plate version of the Ishihara pseudo-isochromatic test, with no errors as the pass criteria. If this test is failed then either a lantern test of the Nagel anomaloscope test is used. The three lanterns recommended by the JAA are the Holmes-Wright Type A (United Kingdom), the Spectrolux (Switzerland) and the Beyne (France) . The subjects pass when they make no errors on the corresponding lantern test. For the Nagel anomaloscope "This test is considered passed if the colour match is trichromatic and the matching range is 4 scale units or less.... ". The tests currently employed by JAA member states and the corresponding pass/fail criteria are fully described in a report by the CAA (Civil Aviation Authority, Minimum colour requirements for flight crew - Part 1, Paper 2006/04, 2006a) .

In the USA, the Federal Aviation Authority (FAA) employs a different set of tests, namely the Dvorine plates as the standard test, followed by the Aviation Lights Test (ALT) if the subject fails the standard test. This can then be followed by the more practical Signal Light Gun Test (SLGT), usually carried out in an airport tower.

It can be seen that current colour vision requirements vary from country to country, even within the JAA member states. The correlation between the outcomes of different tests is poor and therefore it is not uncommon for pilot applicants to fail the colour vision assessment in one country and to pass in another. This is not completely unexpected given the large inter-subject variability, the different factors that can contribute to loss of chromatic sensitivity and the different characteristics of the various colour vision tests. The lack of standardisation often causes confusion amongst applicants and provides the opportunity to attempt several tests in order to pass one of the many colour vision standards.

Furthermore, subjects with minimal colour vision deficiencies often fail these normal trichromacy tests, and are therefore stopped from becoming pilots, although such subjects are often able to perform as well as normal trichromats when presented with the suprathreshold colour vision tasks used in aviation. In principle, such subjects should be allowed to fly. Although this is a well known observation, the only solution so far would be to adopt the arbitrary use of less demanding colour screening tests. Such tests may well pass some subjects with mild colour vision loss, but may also fail others, in an arbitrary way. This is not satisfactory, as the results provide no reliable information as to the minimum colour vision requirements that can be considered safe in an aviation environment. This leads most authorities to demand full trichomacy, as no adequate solutions exist to allow those with a 'safe' level of colour vision abnormality to fly.

There are also further considerations that justify the need to establish safe, minimum requirements for colour discrimination (when appropriate) and to avoid the easier alternative of requiring every applicant to have normal colour vision. The recent UK Disability Discrimination Act (2004) has to a certain extent exposed weaknesses in the current standards and procedures. Companies need to justify refusal to employ an applicant on the basis of his/her defective colour vision and this requires scientific evidence to demonstrate convincingly that the applicant will be not able to carry out necessary occupational tasks that involve colour vision with the accuracy and efficiency expected of normal trichromats . It is an object of the invention to alleviate some or all of the above difficulties .

According to a first aspect of the invention we provide a red aviation light in combination with a white aviation light, wherein the white aviation light comprises a bulb arranged to emit substantially white light having a colour temperature greater than 3500K.

Colour temperature of a light source is determined by comparing its chromaticity with a theoretical, heated black-body radiator. The temperature (in Kelvin) at which the heated black-body radiator matches the colour of the light source is that source's colour temperature. Standard airfield lamps usually comprise tungsten bulbs, for example the JF6.6A 100W/PK30D. These generally produce light with a colour temperature of approximately 2400K, i.e. reasonably 'cool' light, with large red and green components. A white light having a higher colour temperature will appear bluer to an observer. This can make such a modified white light easier to distinguish from other airfield lights, such as red lights, when viewed by a colour impaired observer.

According to another aspect of the invention there is provided a red aviation light in combination with a white aviation light, wherein the white aviation light comprises a bulb arranged to emit substantially white light having a colour temperature greater than 3000K, and preferably greater than 3500K.

The red and white aviation lights may be separately housed, and may, for example, comprise at least part of an aerodrome approach lighting system arranged adjacent a runway. Alternatively, the red and white aviation lights may comprise at least part of an aerodrome PAPI light system. In that case, the lights may comprise a common unit, and/or be housed in a single housing.

According to another aspect of the invention there is provided a PAPI light operable to emit a beam of light comprising red light in a first sector of the beam and white light in a second sector of the beam, wherein the white light has a colour temperature greater than 3000K, and preferably greater than 3500K.

The white light may have a colour temperature in the range 3500 to 25000K. Below 3500K, and certainly below 3000K, white light has a substantial red-green spectral response in the eye, which can make it difficult for a colour deficient observer to distinguish from the red light used in the PAPI light. Above 25000K the white light has a very large blue component, which can mean that the light appears substantially blue, rather than white. Furthermore, to produce light having a very high colour temperature, it can be necessary to filter out a large portion of the red-green optical power produced by a bulb. This filtering wastes energy, and also generates a significant amount of heat. In addition, the appearance of the light can become dimmed. In order to balance a requirement for a large blue component in the white light with a requirement to avoid excessive filtering, the white light may have a colour temperature in the range 4000 to 9600K, and preferably has a colour temperature of substantially 4500K.

The white light sector of the beam may be produced by a bulb, and the colour temperature of the light produced by the bulb may be adjusted using a filter. The filter is preferably operable to raise the colour temperature of the light produced by the bulb. The bulb may be a tungsten filament bulb, of the sort commonly used in aerodrome lighting. According to a further aspect of the invention there is provided a PAPI light operable to emit a beam of light comprising red light in a first sector of the beam and white light in a second sector of the beam, further comprising a filter operable to raise the colour temperature of at least the white light sector of the beam.

According to yet another aspect of the invention there is provided an aerodrome comprising at least one runway and a set of PAPI lights arranged adjacent the runway, wherein at least one, and preferably each, of the PAPI lights is in accordance with one of the aspects of the invention discussed above.

According to a further aspect of the invention there is provided a method of modifying a PAPI light that is operable to produce a beam of light comprising red light in a first sector of the beam and white light in a second sector of the beam, the method comprising placing a filter in the path of at least part of the white sector of the beam to raise the colour temperature of the beam.

Embodiments of the invention will now be discussed on more detail, with reference to the accompanying drawings:

Figure 1 : The Precision Approach Path Indicator (PAPI) signal lights that are used to inform pilots of the correct glide path for landing,

Figure 2: Photograph of airport parking lights,

Figure 3: Frequency distributions of the Yellow-Blue (YB) and Red-Green (RG) chromatic thresholds obtained in 250 observers with 'normal' trichromatic vision, Figure 4: Data showing the 97.5 and 2.5% statistical limits that define the "standard" normal CAD test observer,

Figure 5: Chromatic thresholds for two colour vision deficient observers with minimal colour vision deficiency,

Figure 6: Chromatic thresholds for two colour vision deficient observers with severe colour vision deficiency,

Figure 7: Graph showing Red-Green (RG) and Yellow-Blue (YB) thresholds expressed in CAD Standard Normal units for the population of subjects tested as part of this study,

Figure 8: 1931 CIE-x,y colour space diagram showing the recommended chromaticity boundaries for the colours of light signals. The signal colours used in the laboratory set-up are also plotted, as is the effect of varying the current (i.e. the output intensity) for white light.

Figure 9: Schematic representation of PAPI simulator,

Figure 10: Graphs showing the CIE-x,y chromaticity coordinates of the Red and White PAPI lights under the effect of neutral density filtering and current setting of the lamp,

Figure 11: Schematic representation of the Position Approach Path Indicator (PAPI) simulator test (left) and PAPI Signal Lights Test (PSL) (right), Figure 12: The number of plates read correctly on the Ishihara test (24 plates) is compared to performance on the PAPI simulator test separately for normals, deutan and protan colour vision observers,

Figure 13: PAPI % correct scores plotted as a function of the number of plates read correctly on the Dvorine test for normals, deutan and protan observers,

Figure 14: The number of presentations identified correctly on the Aviation Light Test (ALT) compared to performance on the PAPI simulator test,

Figure 15: PAPI test scores plotted against an index of red-green chromatic sensitivity based on the Nagel anomaloscope range,

Figure 16: Graphs showing performance of normal, deutans and protan observers on the PAPI (standard white) versus CAD test sensitivity (1 /threshold)

Figure 17: Graphs showing comparisons between standard and modified PAPI white versus the CAD test sensitivity values,

Figure 18: Graphs showing R = W and W = R errors only made on the PSL versus the RG CAD sensitivity.

We investigated whether subjects with minimal colour vision loss were able to carry out the most demanding colour related tasks with the same accuracy as normal trichromats. Our investigation indicated that "normal" colour vision is not actually required to carry out the most demanding tasks. Furthermore, we have determined that most colour signals can be modified to allow colour deficient observers to interpret them more easily.

In our study we concluded that the precision approach path indicator (PAPI) lights and parking signal lights are the most colour critical tasks when no redundant information is available to carry out the task (as is usually the case). The PAPI lights provide the pilot with accurate glide slope information on final approach to landing. The geometry of the PAPI signal system is shown in Figure 1.

Four PAPI lights 1 are shown located adjacent an aerodrome runway 3. In this example, the lights are on the right of the runway. However, often PAPI lights are alternatively, or additionally, provided on the left of the runway. Each PAPI light 1 comprises a housing (not shown) comprising a red lamp and a white lamp arranged so that the PAPI light emits a beam of light having a first sector (generally the upper half of the beam) that is white and a second sector (generally the lower half of the beam) that is red. It will be appreciated that a single white bulb provided with an appropriate partial red filter could also be used. A PAPI light often contains more than one of each type of lamp, to ensure that there is some redundancy in case a lamp fails.

Each PAPI light 1 is located adjacent the runway 3 at a particular angle, so as to give a pilot approaching the runway information about the angle of his approach. Generally, a combination of two apparently white PAPI lights and two apparently red PAPI lights indicates a correct approach path angle, whilst two many reds indicate that the approach is too low and too many whites indicate that the approach height is too high. In some cases, only two PAPI lights are provided. In that case, a correct angle of approach is indicated by one red and one white light. It can be seen that it is very important for a pilot to be able to tell the different between an apparently white PAPI light and an apparently red PAPI light. Some colour deficient observers are able to do this, whilst others are not. The approach adopted in our investigation was to relate the accurate assessment of colour vision loss to the subject's ability to correctly identify the different PAPI light conditions, which we believe is the most critical, colour based task within an aviation environment, when the use of other than colour cues was minimized.

The PAPI lights are seen from large distances ( > 5kms) at night when both the angular subtense of each light and the angular separation between adjacent lights is very small. Adjacent lights tend to overlap visually and this is particularly obvious at night when the pupil size is large. Subjects with large higher order aberrations and increased light scatter in the eye will be disadvantaged at night. Although most subjects will have high visual acuity ( < 1 min arc) under photopic (well-lit) conditions, subjects with large higher order aberrations and scattered light may have very poor visual acuity under mesopic conditions when the pupil size is large. Visual acuity at low light levels in the mesopic (low lighting) range is not normally assessed. Partial overlap of adjacent lights makes the task of discriminating the number of red and white lights even more difficult. These additional factors explain why the PAPI task (even though it involves only two colours) is considered to be more critical than other colour based tasks.

The PAPI lights 1 are arranged in a horizontal line and installed at right angles to the runway 3 with the nearest light usually some 15m away from the edge. The lights are approximately 230cm in diameter with an inter- light separation of 9m. The unit nearest the runway is set higher than the required approach angle at 3°30\ with progressive reductions further out field of -20 minutes of arc; 3°10\ 2°50' and 2°30' (for a 3° approach) . Usually each unit contains three light projectors (in case one fails) . The light system has an intensity control for day and night use, with six luminous intensity settings, 100%, 80%, 30%, 10%, 3% and 1% (Civil Aviation Authority, Safety Regulation Group, Licensing of Aerodromes, Chapter 6, Report CAP 168, 2004) .

The units 1 direct a beam of light, red in the lower half and while in the upper half, towards the approach. The different elevation angles give a combination of red and white for an on-slope signal, all-red if the aircraft is too low, and all-white if it is too high. The chromaticities of the lights should follow the specification for Aerodrome Ground Lighting (AGL) . The light intensity of the white signal is required to be no less than twice and no more than 6.5 times as bright as the red signal. The recommended intensities for the white and red light are 85000cd and 12750cd, respectively, at the maximum of their light intensity distribution (CAA 2004) . Generally the white light has a colour temperature of around 2400K, and the red light has chromaticity co-ordinates of approximately (0.685, 0.315) as viewed on a 1931 CIE-x,y colour space diagram (see Figure 8) .

There are many other colour signals that are used in the aviation environment to enhance conspicuity, code information and group objects of interest together i.e. , there is a need to be able to see and recognise red, green, yellow, white and blue lights. There are also signals that involve flashing lights (e.g., runway guard lights, road-holding position lights, etc). Good overall colour vision ability is therefore needed, but most other uses of colour, often as redundant information, are less demanding and also less safety critical.

Other colour critical lights are, for example, the red and white lights making up a runway approach lighting system. Usually, in such a system, an approach to the runway is indicated mainly by a row of white lights, with the row of lights changing to red lights as the pilot gets closer to the runway. The area marked by the red lights indicates to the pilot a region where it may no longer be possible to abort a landing and/or safe to land.

Another example of a two colour codes is provided by the navigation lights mounted on the wing tips of a plane, with a red light on the left wingtip, and a green light on the right. These lights give other pilots and indication of the course and direction of another plane. If a pilot cannot distinguish between the red and green lights it will be more difficult for him to tell whether another plane is approaching or moving away.

Similarly, aerodrome parking lights are also colour critical simply because no other redundant cues are available (see Figure 2) . The red 5 and green 7 parking lights are used in airports to indicate to a pilot the correct line for parking the aircraft at a stand. A change in the colour of the lights signals the correct path; it is not necessary to identify the colours of the lights, but the ability to recognise a difference. However, the colour difference between the lights is very large and the lights subtend a large visual angle at the eye. Consequently the colour discrimination task is less demanding than that presented by the PAPI lights.

The PAPI signal system, on the other hand, offers no redundancy - there is no other unique cue to help the pilot recognise the red and white light signals reliably in order to determine visually whether the plane is on the correct approach path for landing.

The PAPI task is a simple two colour code (and the white and red lights generate both red-green (RG) and yellow-blue (YB) colour signals in the eye) . Red-green colour deficient observers (protanomalous or deutanomalous observers - the most common type of colour deficient observers) will continue to have full use of their YB channel. It has been suggested that even dichromats (subjects with severe colour vision loss) may be able to interpret correctly differences between two colours, at least under some conditions. Whilst it is likely that some colour deficient subjects will be able to distinguish between the PAPI lights, it is also essential to ensure that the subjects recognise and name all four lights as red when too low and as white when too high. Furthermore, the recognition of the red and white lights in PAPI is not always an easy task since atmospheric scatter, and the use of reduced lamp current settings at night to dim the lights, can shift the white signal toward the yellow region of the spectrum locus (see Figure 10) . This often causes problems for colour normal observers and may cause even greater problems for colour deficient observers.

Taking into account these considerations, we have developed a PAPI simulator and a PAPI Signal Lights test that can be used under controlled laboratory conditions. Our aim was to determine which types of colour deficient observers were able to carry out the PAPI lights task correctly (i.e. how severe a colour deficiency needs to be before someone is unsafe to fly) , and whether the amount of colour deficient observers that pass the PAPI lights task (i.e. correctly identify all the lights) can be increased.

The simulators reproduce both the photometric and the angular subtense of the real lights under demanding viewing conditions when the lights are viewed against a dark background. The aim was to correlate the measured loss of chromatic sensitivity on the colour assessment and diagnosis (CAD) test. The CAD test has been described in an earlier CAA report (mentioned above, CAA, 2006a) , which the reader is directed to read now. In summary, the CAD test is implemented on a calibrated visual display and consists of coloured stimuli of precise chromaticity and saturation that are presented moving along each of the diagonal directions of a square foreground region made up of dynamic luminance contrast (LC) noise. The subject's task is to report the direction of motion of the colour- defined stimulus by pressing one of four appropriate buttons. The CAD test has a number of advantages over conventional tests both in terms of isolation of colour signals as well as sensitivity and accuracy.

Figure 3 shows the distribution of YB and RG chromatic thresholds obtained in the 250 normal trichromatic subjects. The mean, standard deviation (SD) and median are shown. Figure 4 shows the statistical limits for the 'standard normal' (SN) observer on the CAD test plotted in the 1931 CIE-x, y colour chart (Rodriguez-Carmona et al. , 2005; Rodriguez-Carmona, 2006) . The variability in both RG and YB thresholds is shown by the grey shaded ellipse, which represents the region of the CIE chart where we expect to find 95% of normal trichromats. The 2.5% and 97.5% limits define the boundaries of this region. The median chromatic discrimination threshold ellipse (SO) is also plotted. The median threshold value is important since it represents the Standard Normal (SN) observer. A subject's thresholds can then be expressed in SN units and this makes it possible to assess the severity of colour vision loss, i.e. an observer with a RG threshold of 2 SN units requires twice the colour signal strength that is needed by the average standard CAD observer. The RG protanopic (P) and deuteranopic (D) confusion bands are shown respectively by circles and a solid line, while the YB tritanopic (T) confusion band is shown by crosses. Figure 4 is an extremely useful representation in that it provides a CAD test template for the SN observer. Any subject's results provide instant diagnosis of either normal or deficient colour vision when plotted on this template.

The CAD test is able to accurately determine the type and severity of colour vision loss whereas this is not possible with other known tests (e.g. the Ishihara test or the Nagel anomaloscope). For example, figure 5 shows two mildly deuteranopic subjects, which may not have been identified as deuteranomalous in other tests (the black contour shows data for an average trichromat) .

The CAD test identifies the type of deficiency involved by the elongation of the subject's results either along the deuteranopic (Figure 6, left) or protanopic (Figure 6, right) confusion bands. In the case of absolute minimum deuteranomalous deficiencies the distribution of the thresholds is a shown in Figure 5. In the case of minimum protanomalous deficiencies the thresholds are much larger and extend sufficiently along in the protanopic direction to be able to diagnose minimum protanomaly with no ambiguity.

The severity of red-green and yellow-blue loss of colour vision is proportional to the colour signal strength needed for threshold detection. For example, subjects in Figure 6 show more severe loss (i.e. higher thresholds or lower chromatic sensitivity) than the subjects shown in Figure 5. The severity of colour vision loss can be quantified with respect to the standard normal observer (Figure 3 and 4) . Chromatic sensitivity varies greatly within colour deficient observers from complete absence of red-green discrimination, in the case of dichromats, to almost normal sensitivity in subjects with thresholds not much larger than 2 SN units. Figure 7 shows the subject's RG threshold in SN CAD units along the abscissa, plotted against the YB threshold along the ordinate in 450 observers. The results show that the RG thresholds vary almost continuously from very close to 'normal' to extreme values which can be 25 times larger than the standard normal threshold. The YB thresholds, on the other hand, vary very little as expected in the absence of yellow- blue loss or acquired deficiency. Interestingly, the RG thresholds show some correlation with YB thresholds in normal trichromats, suggesting that subjects with high RG chromatic sensitivity are also likely to exhibit high YB sensitivity. The loss of sensitivity (when expressed in Standard Normal (CAD) units (SN units)) is greater in protanomalous than deuteranomalous observers (Figure 7) .

Both the ambient light adaptation level and the size of the coloured stimulus can affect chromatic sensitivity. In general, as the light level is reduced and/or the stimulus size is decreased the RG and YB thresholds increase. The YB thresholds is affected most at lower light levels. Both background luminance and stimulus size have been optimised for the CAD test so that no significant improvement in chromatic sensitivity results by increasing either the light level or stimulus size. Any small variations in either light level or stimulus size will not therefore alter significantly the subject's RG and YB thresholds. However, older subjects are likely to show more rapid effects as the light level is reduced simply because the retinal illuminance in these subjects is already low as a result of small pupil sizes and increased pre-receptoral absorption of blue light.

To determine which colour deficient observers can accurately perform aviation tasks requiring colour vision a PAPI simulator test and a PAPI

Signal Lights test (PSL) were designed and constructed specifically for this investigation. A full assessment of colour vision using all these tests takes between 1.5 to 2 hours per subject. The order the different test are carried out varied randomly and the testing took place in three different rooms, allowing the subject to take short breaks in between tests. 182 subjects were examined in this investigation: 65 normal trichromats and 117 subjects with deutan- and protan-like colour deficiencies. The age of the subjects ranged from 15 to 55 years (mean 30.2 years, median 27 years) .

A schematic of the laboratory set-up developed to simulate the PAPI signal system is shown in Figure 9. The PAPI light simulator 10 is a four-channel optical system which uses an airfield halogen lamp 11 (JF6.6A100W/PK30d) as a single light source. Light is emitted from slits 15 in a lamp housing 13 to produce two beams. Those beams are then split into two further beams using two beam splitters (BS) so as to generate four beams 17 of substantially equal strength. Each beam of light 17 passes through two motorised filter wheels 19; colour (CW) and neutral density (NDW) wheels. The CW has six different filters: red, blue, green, yellow, standard white (having a colour temperature of ~2400K), and modified white (having a colour temperature of ~3900K). Each NDW has neutral density filters with optical density values of, respectively, 0.0, 0.3, 0.6, 1.0, 1.3 and 1.6, to simulate the varying intensities of the real PAPI lights.

After passing through the filters each beam (channel) is focussed using a lens 21 into an optical fibre head which is attached to a viewing panel 23 so as to simulate PAPI lights 25. During the calibration procedure the luminous intensity of each beam was measured with each filter in place so as to account for the actual absorption of each filter.

The angular subtense of each light was 1.36' at a viewing distance of 4m. Beyond lkm the angular subtense of the real PAPI lights approaches the diffraction limit of the eye. The size of each light on the retina remains relatively unchanged as the approach distance is increased but the light flux captured from each light is decreased. On approach the PAPI lights are first seen as a small continuous line until the angular separation between adjacent lights is resolved by the eye (typically less than 2' taking into consideration pupil size and optical aberrations. In order to reproduce the geometry of the real PAPI lights in the laboratory for a viewing distance of 4m, the adjacent lights (centre to centre) were separated by ~6.5mm. This corresponds to an angular separation of 5.5' which translates to an approach distance of 5.54km in the case of the real PAPI lights. This design therefore requires the pilot to locate and recognise the white and red PAPI lights from 5.54km when the size of the image of each light on the retina is determined by the Point Spread Function (PSF) of the eye. We have not chosen a larger approach distance in order to minimise the effects of higher order aberrations and increased scatter in the eye have on the retinal images of the lights. When the pupil size is large, the higher order aberrations in the eye can be quite large and this causes the PSF to broaden and the visual acuity to decrease, The light distribution in adjacent PAPI lights can overlap significantly and this in turn makes it more difficult for the subject to process the colour of each light. Since a larger approach distance would produce even more overlap, a distance of 5.54km that is considered to be safe was selected for the study.

The optical fibre heads 22 form a line located at the centre of a black plate 23 which provides a dark uniform surround (see Figure 9) . The whole system is encased and ventilated by two fans on either end to prevent rises in temperature from the lamp. The intensities of the red and white lights were adjusted using ND filters so that the simulated PAPI lights appear as intense as the real PAPI when viewed in the dark from a distance of 5.54km. In addition, the intensities of the coloured lights also varied randomly by ± 0.3 OD with respect to the nominal values to eliminate the detection of brightness cues. The effect of the different intensity settings and ND filters was investigated to establish the extent to which the chromaticities of the while and red lights change with the lamp current setting and/or the use of ND filters (see Figure 10). The results show that the ND filters cause only small changes in the chromaticity (i.e. observed colour) coordinates of the white and even less so for the red light. Changes in lamp current cause larger changes in the chromaticity of the white light, but in spite of these changes the white remains within the 'variable white' area indicated on the CIE diagram as appropriate for AGL (see Figures 8 and 10) . In the case of real PAPI lights, other factors such as atmospheric absorption can also affect the chromaticity of the white, with very little effect on the red.

The four horizontal lights are presented for 3 seconds and the subject's task is to simply report the number of red lights in the display. There are five possible combinations of reds and white lights which are presented randomly (Figure 11, left) . Observers are told to respond using the following names: one, two, three, four or zero when carrying out the PAPI simulator test. Prior to the test observers were allowed to dark adapt to the low mesopic surround and then were given a practice run. A low power lamp was placed behind the test equipment so as to provide low mesopic conditions of ambient illumination. The black, immediate surround around the PAPI lights were dark (i.e. , mean luminance ~ 0.005 cd/m²). Subjects were encouraged to respond only after an auditory cue signalled the end of the 3 second viewing period. The PAPI test was carried out twice, once with the standard white (~2400K) and once with a modified white light (higher colour temperature of ~3900K), discussed in more detail below.

The PAPI Signals Light test (PSL) schematically demonstrated in the right of Figure 11 uses the same equipment as the PAPI lights test, but this time makes use of the other colour filters besides the red and two whites, as well as the optical density filters. In this test six possible colours are presented (standard white (W) , modified white (CC) , red (R), green (G) , blue (B) and yellow (Y)) . The chromatic properties of the lights lie within the boundaries for the recommended signal light for AGL (CAA, 2004) as shown in Figure 8. The solid lines shown in Figure 8 represent the boundaries of the chromaticities of coloured lights allowed in CAA lighting, whilst the dotted boundary shows the allowed chromaticities of white light. The PSL addresses the issue of correct colour naming when all lights have the same chromaticity as opposed to the ability to distinguish and categorise some of the four lights as red and the others as white on the bases of some perceived differences between the lights. The PSL tests whether the applicant can recognise and name reds as 'red' and whites as 'white' for the same conditions and geometry as the PAPI lights, but when all the lights are of the same colour. The condition when all four PAPI lights have the same colour to indicate "far too low" (all reds) or "far too high" (all whites) is clearly very important. Observers were instructed to report the colour of the lights as either red, green, yellow, blue or white.

There were two whites, the standard white as produced by the lamp, with a 'cool' colour temperature of approximately 2400K and a modified white produced by raising the colour temperature of the standard white by 200 MIREDS (to approximately 3900K) . Prior to the test observers were given a practice run. All the colours were shown to the subject and named by the examiner during the practice run and the subject was allowed to review any of the lights and to ask the examiner to confirm their colour. The results for the PAPI and PSL are recorded as percentage correct response. The colour vision of the subjects was examined using five different colour vision tests as well as the PAPI and PSL simulator tests, and the results are shown in Figures 12 to 18. Results from each of the five tests were then examined in relation to the subject's performance on the PAPI to establish which test yields the best prediction of PAPI performance. Performance on the PAPI test is computed as number of correct presentations out of a total of 60 presentations.

The results summarised in Figure 12 show that normal trichromats can make errors on the PAPI and also on the Ishihara test (i.e. , five subjects produce one error, one subject produces two errors and the one other subject produces three errors) . The rest of the normal subjects score

100% correct on this test. Results for deutan colour deficient observers reveal that all subjects with scores > 70% (i.e. , 16 or more correct plates out of 24 on the Ishihara 24-plate test) pass the PAPI with a score of

100% correct. Results for protan observers show that four subjects with scores greater than 40% pass the PAPI test. Overall the results show very poor correlation between the subjects' performance on the Ishihara and the PAPI test scores. Many of the subjects can pass the PAPI test scores that range from 0 to 95% correct on the Ishihara test.

Comparisons of data from the Dvorine plate test with the PAPI simulator show similar results to those obtained with the Ishihara test (Figure 13) . Three normals obtain less than 100% on the Dvorine test (but pass the PAPI with no errors) . Deutan and protan colour deficient subjects need more than 65 and 50%, respectively, on the Dvorine plate test to achieve 100% on PAPI. Since the prediction of the class of deficiency involved is poor with both Ishihara and Dvorine tests, it is difficult to know which of the two limits one should apply to any colour deficient subject. Figure 14 plots the PAPI scores against the subjects' performance on the ALT test. All normals secure 100#% score on the ALT but not on the PAPI test. Results for deutan observers show that quite a significant percentage of subjects fail the PAPI but score 100% correct on ALT. Results for protans show that almost all protans (with one exception) score less than 80% correct on the ALT, but many pass the PAPI test.

Figure 15 compares PAPI scores with a measure of red-green sensitivity based on the Nagel anomaloscope range. Only a few deutan and protan observers pass the PAPI with Nagel sensitivity > 0.6 (deutan) and > 0.4 (protan) . The Nagel anomaloscope test is excellent at distinguishing between deutan and protan like deficiencies, but fails to quantify reliably the severity of colour vision loss and does not test for yellow-blue deficiency.

PAPI test scores show in Figure 16 plotted against the corresponding CAD based measure of RG sensitivity. The top section shows the performance in normal trichromats. Most normal trichromats perform 100% correct in the identification of the red and white lights. However, 7 out of the 65 normals tested made some errors. This could be due to lack of attention and/or reduced visual acuity at low light levels caused by increased higher order aberrations which the pupil size is large. The errors were found to be distributed with equal probability amongst the five conditions. The blue dotted line in Figure 16 shows the 95% confidence interval. The lower sections compare the performance on the PAPI test with the corresponding CAD measure of RG sensitivity for subjects with deutan- and protan-like deficiencies. The RG sensitivity limits beyond which deutans and protans perform and PAPI test as well as normal trichromats are 0.17 (RG threshold ~ 6 SN units) and 0.085 (RG threshold ~ 12 SN units) , respectively. These limits are indicated by dotted vertical lines in Figure 16. Figure 17 shows the significant benefit of replacing the standard white with the colour-corrected (CC) white in the PAPI test. All groups performed significantly better with the colour-corrected white. All, but one normal trichromat, scored 100% correct on the PAPI. Figure 17 also shows the massive overall improvement observed in the deutan group.

All deutans obtain 80% correct or higher and many score 100% correct with the CC white. The improvement is however very small amongst protanomalous observers, particularly amongst subjects with severe deficiency (i.e. those with RG sensitivity less than 0.05 units) .

The PSL test was introduced to examine the condition when all four PAPI lights have the same colour. In this condition the subjects can no longer make use of perceived differences between two adjacent colours presented simultaneously without being able to name the correct colour. The results in Figure 18 below show that subjects do not often confuse reds with whites or whites with reds. Four protan subjects with severe loss of chromatic sensitivity (i.e. , CAD RG sensitivity less than 0.05 units) make W = R and R = W errors and six deutan subjects with CAD RG sensitivity less than 0.15 units made errors on W = R.

From the above it can be seen that using the colour corrected white light there is a significant improvement in many subjects' ability to distinguish between the red and white PAPI lights. As discussed above, white light produces both yellow-blue and red-green colour response signals in the eye, such that an observer with red-green colour deficiency can find it difficult to tell red and white light apart. Therefore, by filtering the white light to increase its yellow-blue component and/or reduce its red green component, we can make it easier for a colour deficient observer to distinguish red and white lights in crucial signals like the PAPI lights. For a light to be noticeably better to distinguish from red than existing commonly used tungsten lamps, the colour temperature of the light should be greater than about 2900K to 3000K, and preferably greater than 3200K. Such a colour corrected light can be achieved by filtering light from a tungsten bulb, as discussed above. Alternatively, a replacement white lamp could be provided which produces light having a suitable colour temperature.

It is important to ensure that the light remains white, and does not actually appear blue to an observer, in order to comply with airfield regulations. For this reason, the colour temperature of a white aviation light should be in the range 3500K to 25,00OK. A light with a colour temperature above 25,00OK can take on a noticeably bluer appearance.

Furthermore, in order to filter a tungsten light having a native colour temperature of about 2400K to achieve an apparent colour temperature of, say, 26,00OK a large proportion of the light output by the lamp needs to be filtered out, reducing the optical power of the output light, and considerably dimming the light. This is a significant waste of electricity, and also results in problems with providing cooling for the filters, which heat up.

We propose that a sensible range of colour temperatures suitable for use in white aviation lights is 4000K to 9600K. This balances the need for a large blue component to the light with the need not to filter the light too much, which would reduce in dimmer output light. However, as discussed above, even a light with a colour temperature as low as 3900K results in a significant increase in the number of colour deficient observers correctly identifying all the conditions of the PAPI lights. Therefore, we propose to select light having colour temperatures at the lower end of the above range, for example lights having colour temperatures in the range 4000 to 6500K (which is approximately the colour temperature of average daylight) , in order to produce a minimal impact of the brightness of the light by reducing the need for powerful filters. Preferably, the light should have a colour temperature of 4200K to 4900K, for example, 4500K.

As discussed, if normal white airfield lights were replaced with colour corrected white lights, with a colour temperature greater than 3500K, for example 4500K, this would improve the visibility of the PAPI signal for normal trichromats as well as colour deficient observers. If such colour corrected lights were used, aerodrome safety would be increased, as the PAPI signal would generally be more visible. Furthermore, some colour vision impaired subjects, who currently are excluded from becoming flight crew, would be able to fly safely, as they would be able to perform the critical colour-dependent tasks required in aviation.

This principle can also be extended to other colours of light. For example, when deciding which combinations of lights to be used in a particular signal (e.g. the airport parking lights shown in Figure 2, or the navigation lights provided on the wings of an aircraft) , consideration could be given to how easy those lights are for a colour deficient observer to distinguish. However, the colours used in aerodrome lighting are largely standard. Therefore, it is not practical to alter those colours. Instead, it might be desirable to provide filters to modify the existing colours, so that they remain similar to trichromats, but are easier for colour deficient observers to distinguish. For example, the green light might be filtered to make it appear slightly bluer, so that it is easier for a person with a red-green colour deficiency to distinguish from a red light. In general, it is not possible to improve on the red lights already used in aviation - they cannot be made to appear 'more red' . However, in principle, it may be possible to increase the blue component of a red light, whilst reducing the clue component of a green light. It will be appreciated that this principle of modifying white light can be extended to situations other than aviation lighting. For example, it might be useful when considering what colour of lights to use in other signals, such as road or rail signals. It also might be useful in selecting colours for use in, for example, advertisements.

Claims

1. A PAPI light operable to emit a beam of light comprising red light in a first sector of the beam and white light in a second sector of the beam, wherein the white light has a colour temperature greater than 3500K.

2. A PAPI light in accordance with claim 1 wherein the white light has a colour temperature in the range 3500 to 25000K.

3. A PAPI light in accordance with claim 1 or claim 2 wherein the white light has a colour temperature in the range 4000 to 9600K.

4. A PAPI light in accordance with any one of claims 1 to 3 wherein the white light has a colour temperature of substantially 4500K.

5. A PAPI light in accordance with any preceding claim wherein the white light sector of the beam is produced by a bulb, and wherein the colour temperature of the light produced by the bulb is adjusted using a filter.

6. A PAPI light in accordance with claim 5 wherein the filter is operable to raise the colour temperature of the light produced by the bulb.

7. A PAPI light in accordance with claim 6 wherein the bulb is a tungsten filament bulb.

8. An aerodrome comprising at least one runway and a set of PAPI lights arranged adjacent the runway, wherein at least one of the PAPI lights is in accordance with any one of claims 1 to 7.

9. An aerodrome in accordance with claim 8 wherein each of the PAPI lights is in accordance with one of claims 1 to 7.

10. A red aviation light in combination with a white aviation light, wherein the white aviation light comprises a bulb arranged to emit substantially white light having a colour temperature greater than 3500K.

11. A red aviation light in combination with a white aviation light as claimed in claim 10, wherein the red and white aviation lights comprise at least part of an aerodrome approach lighting system.

12. A red aviation light in combination with a white aviation light as claimed in claim 10, wherein the red and white aviation lights are provided as a unit in a common housing.

13. A red aviation light in combination with a white aviation light as claimed in claim 10 or claim 12 wherein the red and white aviation lights comprise at least part of an aerodrome PAPI light system.

14. A method of modifying a PAPI light that is operable to produce a beam of light comprising red light in a first sector of the beam and white light in a second sector of the beam, the method comprising placing a filter in the path of at least part of the white sector of the beam to raise the colour temperature of the beam.

15. A PAPI light substantially as described herein, with reference to the accompanying drawings.

16. An aerodrome substantially as described herein, with reference to the accompanying drawings.

17. A red aviation light in combination with a white aviation light substantially as described herein, with reference to the accompanying drawings.

18. A method substantially as described herein, with reference to the accompanying drawings.