"INAL REPORT

<.0 .NO.4 4

controct No.FA A /BRD-127

II THE ROLE OF EXTERIOR LIGHTS

I IN

MID-AIR COLLISION PREVENTION

DDC

PROJECT NO. 110--512 I JIJI 7 iT1

prepared for JLJ \-:JDDC-IRA C

FEDERAL AVIATION AGENCY

Systems Research and Development Service

Cl-CD by

I tAPPLIED PSYCHOLOGY CORPORATION

4113 LEE HIGHWAY

-•" ARLINGTON 7. VIRGINIA

1 JULY 1962

II

IFinal Report No.

Contract No. FAA/BRD-127Contract Tasks 5, 6, 7, 13, 17; and

Portions of Tasks 1, 2, 3, 4, 12

Sm THE ROLE OF EXTERIOR LIGHTS

* IN

MID-AIR COLLISION PREVENTION

i Prepared for

Federal Aviation AgencySystems Research and Development Service

Washington 25, D. C.

I Project No. 110-512R

ITheodore H. ProjectorI Applied Psychology Corporation

4113 Lee HighwayArlington 7, Virginia

This report has been prepared by the Applied PsychologyCorporation for the Systems Research and DevelopmentService, Federal Aviation Agency, under Contract No.FAA/BRD-127. The contents of this report reflect theviews of the contractor, who is responsible for thefacts and the accuracy of the data presented herein.The contents do not necessarily reflect the officialpolicy of the SRDS or the FAA.

July 1962I- ~--

Applied Psychology Corporation, Arlington, VirginiaTHE ROLE OF EXTERIOR LIGHTS IN MID-AIR COLLISION PREVENTIONTheodore H. Projector, July 1962126 pp., includ. 5 ilas., 16 tables, and 82 refs., FinalReport No. 4(Contract No. FAA//RD-127)

ABSTRACT

This report summarizes that portion of a researchprogram on visual collision-avoidance techniques whichdeals with the design and use of navigation light sys-tems. The findings are examined from the viewpoint ofthe Civil Air Regulations, and a three-phase programfor improvement is outlined. In Phase I, it is indi-cated that genuine standardization is urgently needed,and can be achieved with minimum delay with a standardconsisting of red, green, and white steady-burningposition lights and a red, flashing anti-collisionlight. In Phase II, an intermediate-range program,the one major defect of the system suggested .rr PhaseI (the absence of left-right indication to the rear)is to be corrected by substitution of a two-color light,yellow and bluish white, for the white taillight. InPhase III, a long-range program, it is suggested thatinvestigation into the possibility of altitude codingin l1-iit systems be continued. In all three phases,additional detailed recommendations are made to insuremaxinim utilization of navigation lights.

- ILL -

TADLE OF CONTENTS

Page

Abstract *.......................... . ... ... l..List of Figures .. . . ... ....... ... .* * *List of Tables ......... .......... z

Acknowledgments ..... ..... . ........... ......... x.

I Introduction ......... . . . ........... I

II The Information Transmissible by NavigationLight Systems .............. ...... ......... 3

Presence, Location, and the Fixity-of-Bearing Criterion .................... 3

Sector Information ................ ... 5Other Information ....... .... ......... * 9

Range .... .... *...*.... ........... 9Speed and Identification ......... 9Attitude and Maneuver ........... 10Altitude . .a.... *...00060 . . . 10

Summary 0 . .. ..a0*eesOe *.............. *. 12

III Light Coding Techniques **................... 15Color Coding ............ *.... .......... 15Flash Coding ... ..... 0......... 0.....0. 16Flash-Frequency Coding ............... 9Flash-Pattern Coding .......... ..... 20Intensity Coding .... *.*............... 22Array Coding ...... .... *.*..* .......... 25Combination Coding *................... 27

IV Intensity, Efficiency, and Visibility ...... 31Visual Thresholds ..................... 32Adaptation, Fixation, and Signal Color. 33Flashing Lights o .................. o... 35Efficiency o.o.....*..0.*0.........0... 36

Conversion efficiency ............ 37Signal efficiency--intensitydistribution ........ aa.a.0...... 37

Signal efficiency--color .. *....*. 38Signal efficiency--flashinglights ........ ..... ... *00..000 3.

Signal efficiency--extendedsources e. a a, a ........ e.e.... .

Other Engineering Considerations ......

V The Atmosphere, Backscatter, and VFR ... ... 41Signal Color and the Atmosphere ....... 46Backscatter . . ... *• • • • •6 • • • a a • • • • • • • • * a 49

VI Lights in the Daytime ...................... 53

TABLE OF CONTENTS (cont.)

VII Lights and Backgrounds .....................

VIII Search Techniques and Cockpit Visibility .... 63

IX Rules-of-the-Road . ..... ..... ............ 67

X Procedures for Evaluatirg Aircraft NavigationLight Systems.. a... ...... ..... *,.. * * a 73

XI Navigation Light Systems and the Civil AirRegulations ......... .. a. a 77

The Need for Standardization . ....... 77Problems and Procedures in Standardi-zation * ... . 49 . a .a.. a .a 0 a 0 a & 0 0 .a * 0 a 79

Intensity and the CAR .................. 80Limitations of Visual CollisionAvoidance .. a .* a .a * . . ..a . .a.. . . . . 81

Cockpit visibility a........Pilot attentiveness ....... .......Sector information ................ 85Use of the fixity-of-bearing

criterion 8...... .a.a..... a6

High-speed aircraft .............. 87The limits of VIM ........ 0a....... 87

XII Possible Changes in thc Civil Air Regulations 91Phase I: Regulatory Action to AchieveGenuine Standardization ............... 91

Phase II: Regulatory Action to Improvethe Phase I Standard System............ 96

Converting the Standard 3-Sectorto a 4-Sector Quadrantal System... 96

Increasing the Minimum-IntensityRequirements .. ..... .............. 100

Position light intensity re-quirements... ...... ... ...... a. 101

Anti-collision light intensityrequirements ......... **....a. 112

Power required for increasedIntensity... ....... .. *0.....*. 115

Traffic Rules, Rules-of-the-Road,and Minimum VFR Conditions ....... 1us

Restricting possible collisioncourses Ua......a.*a....... 110

Raising VFR minimums .......... 120Increasing cockpit visibility.. 121Changing rules-of-the-road .... 122

TABLE OF CONTrNTS (Cont.)

XII. (Continued)Phase III: Long-Range Program toDevelop New and Improved Naviga-tion LUght Systems ........... ,. 123

References ... ,.., . .. ,o,,.,.,.,,.,.,...,... 127

LIST OF FIGURES

No. Title Pae

1 Illustration of use of quadrantal sectorcoding to distinguish non-threats frompossible threats......................... .... . 7

2 Range vs. atmospheric transmissivity forlight intensities of 20, 100A and 500candles, based on "practical illumina-tion threshold of 0.5 mile-candle.............. 23

3 Illumination from a 100-candle signal atvarious distances for selected values ofatmosphere transmlss4vity........ ........ •..•• 43

4 Night visual range of lights of various in.t=cn.Zilee as related to atmospheric condition.. 45

5 Illustration of rule requiring counterclock-wise rotation of sightline S, betweenaircraft ....... .... . .. -........... ........... 70

-Viii -

LIST OF TABLE:S

No. Title Page

I P'inimum Transmittance Limits for Two ColorSeries Specified in Federal Standard No. 3.... .7

2 Visual Properties of the Atmosphere;International VisibilityCode................. 42

3Atmospheric Selectivity.48

4 CAR Minimum Intensity Require=-•nts forPosition Lights ............................ 82

5 Minimum Flight Visibilities ...................

6 Approximate Costs of Navigation Light SystemEquipment ......... . . . .. ,. ......... 93

7 CAR Minimum Position Light Intensity in theHorizontal Plane ......... ,..... ......... ..... 102

8 CAR Minimum Position Light Intensities Aboveand Below Horizontal ............... .......... 103

9 CAR Maximum Position Light Intensities inOverlap Zones ... .. . ..... ... 104

10 CAR Minimum Intensity for Anti-CollisionLights .. ....... ....... . a . a .a .. .. 105

l1 Required Si,-nal Intensity for VariousCollision Courses .... ........ a...... . 107

12 Suggested Minimum Intensity FRequirements(Candles) for Aircraft Position Lights 0.... . 110

13 Sulrested Maximum Intensity Requirements,or Position Light Overlaps . . . 113

14 Suggested Minimum Effective Intensitiesfor Anti-Collision Liq.hts ........ 11... Ill

15 Power Consumption of Navigation Light SystemsMeetlnr Present CAR Requirements Compared withEstimated Power Consumption of Systems MeetingProposed Higher Intensity Requirements ........ 117

16 VFR Cruising Altitudes .................... 119

IIix -

ACKNOWLEDG MENTS

I The author wishes to acknowledpee his in-debtedness to the many persons whose cooperation,contributions, and advice made the study possible,and in particular, the original contract monitor,Dr. H. Richard var Saun, and the current contractmonitor, Eugene E. Pazera; Wayne D. Howell, Jack J.

q Shrager, Philir R. Marshall, James A. Sunkes andtheir associates at N4FEC; Dr. C. E. Leberknightand James Newton of the Kopp Glass Co.; Lew Mooreand Paul Greenlee of the Grimes Mfg. Co.; andF. C. Brec;,enridze and Charles A. Douglas of theNational Bureau of Standards.!

I

I

!I

I

I

THE ROLE OF EXTERIOR LIGHT'S IN

MID-AIR COLLISION PREVENTION

I. INTRODUCTION

The earliest navigation light system used on aircraftwas patterned after the system used in marine practice. Itconsisted of red and green wingtip position lights coveringazimuthal sectors from 00 (dead ahead) to i100 abeam, leftand right respectively, and a white light covering the re-maining sector to the rear. This system remained in usewithout change until about 1940. At that time, severalcommercial air lines because of several near collisions Inthe air and one actual collision on the ground, all involv-ing overtaking courses, requested that the Civil AeronauticsAdministration improve conspicuity to the rear. To improveconspicuity and to avoid confusion of the white taillightwith other lights, alternately flashing white and red tail-lights were adopted.

From this time on, one proposal followed another andvery soon, a number of navigation light systems, differingmore or less from the original system, were in use. Someadded other lights to the basic configuration, others changedcharacteristics of operation.

In the 1950's, a number of prcposals which departedmore radically from the original configuration were offered.New kinds of light sources were used, different kinds of in-formation were presented, and new ways of codifying informa-tion into light signals were tried. In 1955 Special CivilAir Regulation SR 392 was adopted to facilitate experimenta-tion with new systems. Under the regulation, up to 15 in-stallations of a given experimental system might be author-ized for a period of six months. Renewals of authorizationshave been granted regularly.

At the present time a variety of navigation light systemsare installed on aircraft. Some are fairly distinctive, otherspresent signals that are ambiguous--a visual signal in one sys-tem presents information that is different from that presentedby the same signal in another system. It may be asked whetherthe diversity of light systems and the ambiguity of some sig-nals add to the danger of collision. Many pilots contend thatthe information presented by coded signal lights is of no in-terest to them anyway--they are concerned only with the singleitem of information that an aircraft is there. Collisions arefar fewer at night than in the daytime, and no night collisionhas been blamed on the inadequacy of light signal information.The analysis of near-miss reports is very difficult and often

speculative, but here again, no clear picture can be obtainedof information inadequacy as a source of trouble.

If it is true that no information other than aircraftpresence is of value, then it is unreasonable for the CARto impose a regulation system on aircraft operators. Asingle light of the highest possible conspicuity is allthat should be required, and requirements for sector-codingshould be eliminated. Generally speaking, aircraft lightsare more visible at night than aircraft themselves in thedaytime, and maximum conspicuity would be attained if allavailable power were concentrated in a single signal visiblein all necessary directions. Lighting systems lend themselvesto transmitting information, but the need for and usefulnessof any given information must be shown before it can reason-ably be required on aircraft. If it can be shown that someinformation is useful enough to be re-quired, then it is es-sential that (a) the required lighting system present itclearly, and (b) no differing systems be permitted.

This report summarizes the effort of the contractor toinvestigate the characteristics of navigation light systemsand their usefulness in minimizing the risk of collision.In order to do this the various relevant characteristics havebeen analyzed and isolated. Some of the essential propertiesof such systems are relatively well understood--the relationbetween intensity and visibility, for example--and it isonly necessary to examine the literature to obtain this in-formation. Other properties, not so well understood, havehad to be investigated "xperimentally. This report dealswith the isolated aspects separately, then synthesizes theresults. This approach is considered the only effective wayto deal with one of the problems that has hitherto made itdifficult to make reasonable decisions about lighting systems.As proposed, the systems are usually in the form of completed"hardware" in which many separate characteristics are inex-tricably combined. It is felt that a rational evaluationof system effectiveness is possible only when each system isanalyzed in terms of the separate characteristics--for example,the information presented, the coding technique used to pre-sent it, its intensity distribution, its efficiency, weight,size, reliability, etc.

-2-

II. THE INFORMATION TRANSMISSIBLE BYNAVIGATION LIGHT SYSTEMS

Any aircraft navigation light, by virtue of being sightedand identified, transmits the most essential information: thepresence and location of the aircraft. Every lighting systemcurrently being used provides, in addition, sector informationof one kind or another which identifies an azimuthal sectorof the sighted aircraft, and consequently, to some degree, itsdirection of flight. Six additional types of information havebeen cited as possibly uleful if they could be coded into anavigation light system:•

1. Identification of aircraft by type or other

characteristic;

2. Distance;

3. Altitude;

4. Speed;

5. Attitude (pitch or roll);

6. Maneuver (change of course frcm straightand/or level).

Of these items of information, which are "essential,"which useful, and which of little or no value? If for sim-plicity it will be necessary to limit the amount of informa-tion presented, which items should be included In the system,and how precisely or in how much detail should the informationbe coded?

Presence, Location, and the Fixity-of-Bearing Criterion

As noted above, the first item, presence and location,is conveyed, without coding, as soon as a light signal issighted and identified. (Identification as an aircraft lightis important; as will be discussed later, aircri!' gts areoften confused with stars, tower and other obstruction lights,and ground lights.)

A preliminary discussion of information presentation maybe found In Projector & Robinson, 1958. A detailed analysis ofthe usefulness of coded information is contained in AppliedPsychology Corporation Technical Report No. 1, January 1961, onwhich the following discussion is largely based.

No particular coded information is involved in ex-tracting presence and location information from a sightedlight. Nevertheless it is possible to use such a signalto determine other useful information about an intruderaircraft, especially one observed for any length of time.The motion of the light relative to the background or toone's own field of view can help the pilot determine whetheror not a collision can be ruled out. It is particularlyhelpful to analyze the problem in terms of the "fixity-of-bearing" criterion: when two aircraft are flying straight,constant-speed courses (not necessarily level) toward acollision, the bearing of either aircraft remains constantin the field of view of the pilot of the other. The cri-terion is not usable if the conditions are not met or cannotbe assumed to be met, and thus is not applicable in terminalareas, where it is likely that a sighted aircraft will be ina maneuver; under any conditions, "fixity-of-bearing" must beused with reservations. Calvert (1958) has penetratinglyanalyzed this criterion, with important implications both forits limitations and for the approach to analyzing the useful-ness of any technique.

One commonly used premise underlying analyses of colli-sion probability is that there exists some required "warningtime," admittedly uncertain, and variously estimated bydifferent sources. Laufer (1955), in emphasizing the com-plexity of determining warning time, says that in "someexceptional cases a full minute or more may be required."He carries out his collision analysis for two warning times,25 and 50 seconds. Another source (Honeywell AeronauticalDivision, 1961) says, "Depending upon maneuverability of theaircraft, the desired minimum warning time generally acceptedis 10 to 20 seconds." Stone (1954), thinking in terms ofDC-7 aircraft, said "...we are now down to 15 seconds toavoid collisions." Projector & Robinson (1958), referringto Laufer (1955), said that the "required warning time prob-ably lies between 25 and 50 seconds." Many illuminatingengineers have pointed out (Larfer, 1955; Projector & Robin-son, 1958, for example) that the light intensities requiredto furnish the required warnirg times, as estimates, underthe full range of VFR conditions, were so high as to be im-practicable. It has thus been recognized that visual colli-sion avoidance, with presently available techniques and equip-ment, has serious limitations when closing speeds are high orflight vi3ibility is near the VFR minimum.

Calvert's (1958) analysis shows, however, that there areother and more profound limitations. His analysis, althoughlimited to the fixity-of-bearing criterion, has much broaderimplications, which apply generally to all avoidance techniquescurrently In use. Calvert bases his dpproaoh on how well apilot can estimate the probability of collision and, in the

-4-

event he undertakes an avoidance maneuver, how assured hecan be that the maneuver he selects will eliminate or atleast reduce. the probability of collision. The analysisshows that the uncertainties inherent in the fixily-of-bearing criterion are so great that the pilot often cannotuse it effectively. In many situations, including somewith moderate-speed aircraft, the information he needs touse the fixity criterion properly is unavailable or inade-quate. If he does undertake an avoidance maneuver withinadequate information, he cannot tel' what effect it willhave on the probability of collision. Once he has begunthe maneuver, he is committed, but he no longer has thefixity criterion, nor can he know wnen to end the maneuver.Since the uncertainties increase with distance, very earlywarning is sometimes of little or no help to him.

Because of the limitations on when it may be applied atall, and the inherent uncertainties when it is applicable, thefixity-of-bearing criterion, it seems evident, will not sufficeas a visual collisio i-avoidance technique. It is often usefulfor roughly determining that an aircraft is not on a collisioncourse; in other cases it is not applicable at all, or cannotbe relied on.

Sector Information

All systems now in use present sector information. Posi-tion lights, as called for in the CAR, identify three azimuthalsectors around the aircraft: the red wingtip light identifiesthe sector from dead ahead around left to 1100; the green wing-tip light identifies a similar sector on the right side, thewhite tail light identifies the r-emaining 1400 to the rear.Some systems provide quadrantal coding, and others divide theazimuthal plane into six sectors. In some cases the sectorsare sharply defined, in others the boundaries are vague.

It has always been supposed that sector information isessential and can be used by the pilot to determine collisionprobability, although how he might do so has not been clear.When carefully analyzed, it turns out to be impossible for apilot to make such a determination except in very crude terms.The possibility of collision can sometimes be ruled out withthe aid of sector information, but in large numbers of casesit is not possible for a pilot to use sector information todetermine collision hazards.

To show why this is so, it is useful first to indicatehow sector information can be used for crude screening. Opti-mally a sector-coded lig-- system should accurately identifyfour quadrantal sectors in two modes, left vs. right and frontvs. back. With such a system, no collision impends if:

-- r -

1. The left side of the intruder aircraft is sightedon the left;

2. the right side of the intruder aircraft is sightedon the right;

3. the rear of the intruder aircraft is sighted to therear.

An example of the use of these rules is illustrated inFig. 1. If an observing pilot sights an aircraft to his leftIn one of two aspects, A or B, as shown in the figure, he mayuse quadrantal sector information to determine that aircraftB is not a collision threat by rule I above, since he seesits left aspect to his left. Aircraft A, on the other hand,showing its right aspect on the observer's left, may be athreat, and will require further attention.

It will be noted that the position light system calledfor by the CAR, with taillight coverage that is symmetricalto the rear, does not provide information that would distin-guish these two cases.

Thisscreening process is uncertain if either aircraftis turning (although in some cases, the turn may result inan indicated sector change early enough to warn the pilot).

The quadrantal sector system can be used to obtain otherimportant information, but again only broadly and imprecisely.In evaluating a collision threat, the pilot finds it helpfulto know the closing speed of his own aircraft and the intruder.If it is low, and the sighting distance reasonably large, hehas time to continue observing in an effort to evaluate moreprecisely. If the closing speed is high, he must decide morequickly. A quickly made decision is subject to greater uncer-tainty and much higher probability of an unnecessary avoidancemaneuver, especially if the pilot acts conservatively. Forexample, if a pilot sees an intruder ahead of him and identi-fies the intruder's light signal as the forward-aspect signal,he must assume a relatively high closing speed. If in thesame situation he identifies the rear sector of the intruder,the may assume a lower closing speed, and consequently more

time to decide. Because of the uncertainty of some of hisinformation, neither determination can be very precise. Whilehe knows his own speed and direction precisely, he does not ingeneral know with any exactness the speed, relative bearing,or distance to the intruder, all of which would be needed, inaddition to some difficult computations, to determine the timehe has available before possible collision.

Another situation in which quadrantal sector coding mightprovide additional information is one, for example, in which

- 6-

Fig. 1. Illustration of use of quadrantal sector codingto distinguish non-threats from possible threats. Observer,0, sees B~s left on his left and classifies him as a non-threat. A shows his right aspect on Ols left and Is classi-fied as a possible threat.

-7-

the intruder is sighted on the left, slightly ahead of a90-degree bearirg. If in such a sighting the right rear-ward quadrant of the intruder is signaled the observingpilot mny consider two possibilities: (al the high proba-bility that the intruder is on a diverging course and thereis no danger of collision, (b) the smaller probability thatthe Intruder is on a converging course. Even if it were thelatter, the convergence would necessarily be relatively slow,since the angle of convergence can be no greater than theangle that the intruder sight line makes with the left abeam

f direction. If the pilot is satisfied that the intruder isnot very close to him, he can reasonably decide to continue

;L observing for a while in order to improve the precision ofhis determination of the existence of a collision hazard.

In principle, the kind of collision-probability analysisdescribed above can be sharpened by a light system that de-fines more than four sectors. One might design an 8-sectorsystem which divides each quadrant into two 45-degree sectors.The rules for categorizing threats and nonthreats could thenbe extended. However, there are limitations on the gainsthat might accrue:

1. Added complexity of the signal system would makeanalysis more difficult, and more subject to error and un-certainty.

2. Engineering coding techniques to identify as manyas eight sectors would be difficult. For example, colorcoding is limited to four ,,.)lors if reliable distinctivenessIs to be achieved.

3. The remaining and unavoidable imprecisions of thetechnique (as described above) seriously limit the usefulnessof such additional information.

Because of these limitations, it might be preferable tosupply additional information in altogether different forms.

Thus, it is felt that a four-element quadrantal systemrprovides crude but readily useful information, and should beconsidered the minimum for an effective sector-coded system.Gains in precision afforded by finer sectoring are likely tobe offset by the difficulties of design and use, so that thequadrantal system may be considered optimal.

Sector Information can be supplemented by additionalvisual cues apart from those coded into the light system.These Include the intruder's motion relative to the back-ground and relative to the observing pilot's frame of refer-ence (the fixity-of-bearing criterion mentioned earlier),altitude difference as suggested by the angle of elevation

-8-

of the sight line to the intruder, and both distance andchange of distance as estimated, however crudely, by ob-serving the apparent intensity of the observed lights, etc.To utilize sector-coding information and supplemental cueswith maximum efficiency, pilots must be trained to extractthe essential information quickly and accurately, and theymust be thoroughly aware of the limitations and uncertain-ties of this Information.

In terminal areas, the information is likely to be ofbetter quality than on airways, since speeds may be presumedgenerally low, even for high-performance aircraft, and thevisual cues richer and more precise.

Other Information

Additional items of information that might be coded intoa navigation light system are: (a) range; (b) speed; (c) iden-tification of aircraft or aircraft type; (d) attitude (pitchor roll); (e) maneuver; and (f) altitude.

Range. As noted previously, secondary visual cues, suchas mo =on relative to the background, or apparent intensityof lights, can help pilots make crude estimates of range.Reliable estimates based on apparent intensity would requirestandardizing both the intensity of various models of the sameequipment and the distribution of intensity around the aircraft.Intensity distributions of most lights vary markedly with angleof elevation, and some vary in azimuthal distribution. "thepilot is unable to estimate precisely the angle within thedistribution. Also, as discussed below, the attenuation ofintensity by the atmosphere varies widely over the range oftransmissivities possible in VPR flight. Therefore, apparentintensity as a source of range information is unreliable. How-ever, within limited conditions, pilots can estimate rangefairly well, especially after some training. This was shownin ground-to-air and air-to-air tests of pilot ability to esti-mate range (Applied Psychology Corporation, 1962c). Lights canbe coded to aid in range estimation by array coding or Intensivcoding, but this is either infeasible or unreliable, as will beshown in the discussion of coding techniques. Range coding ina lighting system therefore appears impractical.

Speed and identification. It would be useful to a pilotto kno ;ne Speea or a s.gn~ed aircraft, but only for estimatingthe approximate order of magnitude of his decision time. Pre-cise analysis would require precise knowledge of speed, plussector information and a fair-I-T-Tficult computation. Sincethe complete operation requires far more information than couldconceivably be coded into lights, and, as well, difficult compu-tation, such precise analysis seems infeasible. Nevertheless,it is both useful and feasible to attempt a rough bracketing of

"-9-

closing speed. Sector coding permits this to a limited extent,Knowledge of speed would improve the estimate. A light codethat would indicate speed therefore seems worth considering.This could be in the form of an automatic device that wouldcode a light signal from an airspeed indicator. A secondtechnique might Identify aircraft type and thereby indicatea probable speed bracket. Por example, identification bytype night Identify helicopters and small aircraft as a slow-speed group, larger propeller-driven aircraft as a medium-speed group, and high-performance Jet aircraft as a high-speed group.

Because of its limited usefulness and residual uncertaintyprecise speed coding is probably not worthwhile. It may bethat the limited information potential of lighting systems willmake other types of coding preferable.

Attitude and maneuver. Attitude and maneuver are groupedtoge~ier because trey lnclcate some change in the course ofthe sighted aircraft from straight and level flight. Pitchinformation is not considered useful or feasible, since, ex-cept for extreme dive or climb, variation of pitch Is too smallto be reliably indicated by lights and is not related in asimple way to change of course. Turns and banks are feasiblefor coding Into a light system, and it would be useful for apilot to know that a sighted aircraft is engaged in or aboutto engage in such a maneuver. To a more or less limited ex-tent, some existing systems provide such information. Whenboth wingtip position lights are visible, and separable, thepilot has a fairly good Indication of a banking maneuver.However, in a well-defined quadrantal system, quadrant cutoffsare relatively sharp and the overlap zones between quadrantstoo small to provide this information in other than very limiteddirections. Information as to maneuver--left or right turn,climb or descent--Is generally useful, and should be consideredas one of the kinds of information that could be coded intolight systems.

Altitude. The remaining item of information that may becoded tnto Ae light system is altitude. It has been suggestedthat a light system indicating angle of elevation constitutes akind or altitude-coded system. However, in the absence of rangeinformation, the triangulation required to convert angle of ele-vation to altitude difference Is impossible. Thus the Indica-tions of an angle-of-elevation system, "same altitude," "above,"or "below" are unresolvably imprecise.

A true altitude-indicating system, altimeter derived, doesoffer information which may be of considerable value. A detaileddiscussion of various types of altitude-codeC navigation lightsyatems, their probable advantages, and the practical problemsto be solved is contained in Applied Psychology Corporation

-10-

Technical Report No. 1 (1961a). They are summarized below. 1

Altitude-coded light systems can identify predeterminedaltitude segments such as 500, 1000, or 2000 feet. Con-ceivably the system segmentation could vary with altitude;for example, 500-foot segments could be used at low altitude,1000-foot segments at medium altitudes, and 2000-foot segmentsat high altitudes. The altimeter transducer may be set toproduce, for example, a light signal to identify a 1000-footsegment at 12,000 feet whenever the indicated altitude isanywhere from 11,500 to 12,500 feet.

Any practical system would be cyclic, the code systemrepeating after a given number of consecutive segments. If,for example, 1000-foot segments were identified in a 5-segmentcycle, the indications for 6000, ilO00, 16,000 feet, etc.,would be identical.

The advantage of altitude coding is that it identifiesa fixed segment of the atmosphere. Unlike other informationavailable to the pilot, it is not indeterminate because otherinformation, such as range, is lacking. No matter what thesighting range, the quality of the information is the same.An avoidance maneuver can oe undertaken at any time, requiringonly that the pilot attain a fixed altitude separation. Thepilot continues, during his maneuver and at its end, to haveinformation enabling him to determine when avoidance has beensafely effected. While some ambiguity is possible in a cyclicsystem, it leads only to an occasional unnecessary avoidance.It does not, as in the case of other techniques, result in asituation in which the uncertainties include the possibilitythat the avoidance maneuver has not been successful.

One possible advantage of altitude coding is that it mayaid in estimating range. It was shown that estimating alti-tude difference from an angle-of-elevation observation is anindeterminate triangulation problem. But if the altitude dif-ference is given by altitude-coded lights, the range is nolonger indeterminate. Even though the altitude informationis still approximate, it may significantly improve the accuracyof range estimates over what they would be without any altitudeinformation.

Designing altitude-coded systems presents a number ofother problems.

1 The results of studies of the usefulness of altitude-coded navigation lights are contained in Technical ReportsNos. 8 and 16 (Applied Psychology Corporation, 1962b, f).

- 11 -

Inaccuracies are unavoidable. In addition to the in-herent inaccuracies of presently available altimeters, whichare a serious problem in all methods of collision avoidancerelying wholly or partly on altitude separation, there arebound to be errors in transducing altimeter information intothe coded light signal.

Since only altitude segments are identified, there issome uncertainty about precisely where in a segment a sightedaircraft is located. The top of one segment Is at ae samealtitude as the bottom of the next higher segment, and a one-segment separation may not be a certain indicator of safeseparation.

Rules-of-the-road would have to be designed to utilizethe altitude information and would have to be tailored tothe particular system used. It would be necessary, for exam-ple, to insure that two aircraft on the same altitude did notengage in the same altitude avoidance maneuver.

In a complete navigation light system, it would be possibleto include other information in addition to altitude information.One comioination might be altitude and sector information.

Summary

Four kinds of information codable into navigation lightsystems may be of value in avoiding collision. They are:

1. Sector information. An optimum system should identifyfour quadrantal ecT-ors" riy precisely. While sector infor-mation is nr4:t as useful as has been supposed, it does have someutility for roughly categorizing aircraft into "threats" and"nonthreats." It is also useful in terminal areas in deter-mining the course and future location of sighted aircraft.

2. Altitude information. Altitude information mightenable a ploft to Improve considerably his ability to control fthe outcome of a collision threat, in many cases with a degree

of certainty not attainable with other information. It hasother possible advantages that suggest it is well worth inves-tigating.

3. Speed information, Such information would be usefulto some extent. Iz is oub tful, however, that it would beadvantageous to identify more than two or three categories ofspeed, for example, slow, medium, fast. The speed code couldbe derived from the airspeed indicator or could identify air-craft type by maximum cruising speed.

4. Maneuver Information. Indicating a change of course,either turn or change Br aIt-tude, would be helpful in warning

- 12 -

a pilot that his judgment based on other information or cuesis subject to change.

In addition to the information obtained from coded lights,the pilot can obtain useful information from any sighted air-craft light. This includes, of course, the presence and sightbearing of the aircraft. It also includes the use of thefixity-of-bearing criterion and, through secondary visual cuessuch as movement relative to background, some capability ofdetermining the course of the aircraft and the existence of a"threat."

A system that would code all four types of informationwould be complex, difficult to use, and probably impracticableon the basis of engineering, economic, or human capabilityconsiderations. There is, however, a strong possibility thatthe present "standard" system calling for sector coding couldbe modified to supply quadrantal sector coding, and that itwould be worthwhile to supplement this with at least one ofthe three remaining information types. Of these, altitudecoding seems most promising.

i l l l ll l i Fl l l l ll l l*

III. LIGHT CODING TECHNIQUES

Five more-or-less distinct methods of coding informa-tion into light signals may be defined: (a) color coding,(b) flash coding, c) flash-frequency coding (d) flash-pattern coding,, (e) intensity coding, and (f5 array coding.

These may be combined in a great variety of ways, sothat in principle the isxforiation-conveling capability ofnavigation light systems is very large.

Each technique has its own distinctive characteristics,often complexly related to the over-all problem of visualcollision avoidance. In assessing the usefulness of a light-coding technique several factors must be considered: (a)the speed and (b5 the accuracy with which the code can beread; (c) the amount of information that can be coded; and(d) the effect of the code on other aspects of operation.

Color Coding

The earliest technique applied to navigation light sys-tems, and still the most widely used, Is color coding asprescribed for use in the current CAR position light system.These colors are codified and their precise definitionsgiven under "Aviation" colors in Federal Standard No. 3,1951.

Three-color coding systems are in widespread use inother fields as well as aviation (highway traffic signals,for example) and there is little doubt the 3-color systemsprovide a high degree of reading accuracy and speed, andfairly good efficiency with standard light sources and fil-ters.

A color-coded system with more than three colors pre-sents difficulty. Federal Standard No. 3 defines five Avia-tion colors: red, yellow, green, blue, and white, but thesa,colors

... are intended for high-intensity long-rangesignal lights in which the primary considerationis that the light be seen, the secondary consid-eration is that its color be identified. Ifaviation colors are used In situations requiringpositive color identification, relatively fewcolors are used at a time. For example, aviationwhite is not intended to be distinguishable fromaviation yellow unless it is in Juxtaposition withit.

1 Combination codes will be discussed as a seventh type of

coding. . .

The Federal Standard goes on to define a second category of"Identification" colors: red, yellow, green, and lunar-white (a bluish white color). In order to insure reliabledistinctiveness the specitficationlimits for this series ofcolors are much narrower than for the Aviation series. Itis thus implicit that the Federal Standard Is written withthe idia that a four-color system is the maximum for reli-able distinctiveness, and even here, color specificationtolerances are small, with a consequent reduction in lumi-nous efficiency.

The differe.nce Ir. the specifications for the two seriesof colors is iYlicated by the minimum transmittance limitsfor comparable glassware, as shown In Table 1.

MoNicholas (1936), who extensively Investigated thebest colors for use In a railway sig•al system, found thata set with color names "red, orange-yellow, white, green,bluc, and purple" was the optimum set of six. For variousreasons, Including the requirement for reasonable tolerancesin specifying color filters, the need for high efficiency inproducing specified colors, the unsuitability of blue andpurple for distant signals, and the residual confusabilitybetween adjacent colors in WaNicnolas' 6-color system, itIs ev4 dent that six colors are too many for aircraft navi-gation light systems.

Hill (1947) experimented on the distinctiveness of 73widely distributed colors. T1'e results indicate the reli-ability with which his subjects identified the test colorsas red, green-blue, orange-yellow, and white, when the sig-nals were presented at low levels of intensity. The dataare helpful in establishing boundaries for color signals inaircraft navigation light systems.

Three-, four-, five-, and six-color systems are discussedin an Annex to Recommendations by the Technical Comnittee onthe Colors of Signal Lights of the CIE (1955) and in subse-quent formal recommendations (CIE, 1959). It Is suggestedthat a three-color system be considered maximum for generalapplication. A four-color system Is described and specified,but it is suggested that the yellow and bluish white will notbe adequately distinctive for distant viewing at low illumi-nation. The four colors recommended are red, green, yellow,and bluish white, more or less as described In Federal Stand-ard No. 3.

Another problem In color coding is color deficiency Inpilots. Part 29 of the CAR requires norma color vision fora first-class medical certificate. Pilots with second- andthird-class certificates are required to be able to distinguishaviation colors, red, green, and white. A few such pilots may

X6

Table 1

Minimum Transmittance Limits for Two

Color Series Specified in Federal Standard No. 3

Minimum TransmittanceColor Aviation .. entirication

Series Series

Red .175 .048

Green .200 .048

Yellow .500 .o00

- 17 -

have difficulty in distinguishing the colors of a four-colorsystem@

It may be noted here that color distinction can be con-siderably more precise than suggested above if colors areseer. not In isolation but close together in space or time.For example, yellow and bluish white may be somewhat con-fusable if either in seen alone, but if both colors are seenside by side, or if they are seen in rapid succession, theyare readily distinguished. Under such conditions, systemswith four or even more colors are entirely practicable. Itshould also be noted that the precision of color distinctionin any color system is poor when signal lights are at or nearthreshold. With any color system, but particularly with high-er order systems, there is a region near threshold where eventhough signal lights may be detected color distinction may beimpossible or uncertain. 1

In summary, three-color systems provide good color dis-tinction, are quickly and accurately read if not too nearthreshold, and can be designed with fairly good efficiency.(Aviation red and green glassware with transmittances of theorder of 20% are readily a,ailable for use with incandescentlamps.) If four colors are required in isolation, colorcoding is less distinctive, and it may be necessary to re-strict the tolerances of the colors to minimize confusion.

Flash Coding

Light signal flashes may impart information as well asincrease the efficiency of utilization of electrical energyto produce visibility. The distinction between flashing andsteA4y lights Is used in one version of the navigation lightsystem described in the current CAR, wherein the "anti-col-

1s8, .on" light emits an oumidirectional flashing red signalwhile the left wingtip position light emits a steady redsignal identifying the left forward sector.

Since lights may be flashed in a variety of wlays, thistechniq.e '.' versatile. Flashes may be distinctive becauseof their on-off ratio, frequency of flashes, grouping offlashes by varying successive dark intervals, or because ofthe use of sequences of long- and short-duration flashes(dots and dashes).

1 An excellent detailed critical review of research workon color-identification systems may be found in Breckenridge,1960. A discussion of the properties of signal color systemsand their uses and specification may be found in Breckenridge,1962.

-18- -

The fact that a light is flashing can be determinedfairly qulckly, usually by the time two or three flasheshave been observed. Flash patterns can usually be identi-fied by the time two or three cycles have been completed.Flash characteristics are not as rapidly identified as iscolov at intensity levels well above threshold. On theother hand, color identification near threshold is oftenvery difficult or impossible, when the flashing character-Istic is still identifiable. At threshold the flashingcharacteristic is not distincttve, because even steadylights appear to twinkle or flash.

The dark interval between light flashes, especiallyif of appreciable duration, can be troublesome. Occasion-ally It may delay pickup of the intruder aircraft, if thepilot's center of attention happens to pass the target whilethe signal light is in the dark phase. The flashing charac-teristic is also an impediment in the use of the fixity-of-bearing criterion (Applied Psychology Corporation, 1961a).The inherent uncertainty in a pilot's ability to detectfixity directly affects the precision of his determination.There is little data available on the threshold detecta-bility of movement for steady signal lights. In ideallaboratory situations movement rates of the order of 1minute of angle per second are detectable (Leibowitz, 1955),but detection thresholds in operational situations are un-doubtedly much larger for steady signals, and larger stillfor flashing signals.

Flash-Frequency Coding

Different flash frequencies may be used to code informa-tion into a navigation light system. Two frequencies are usedin a system proposed by the Lytle Engineering and ManufacturingCompany1 , and three frequencies in the "Honeywell-Atkins Maxi-mum Safety Light" (Honeywell Aero Div., 1961). In both, flashfrequencies Identify azimuthal sectors around the aircraft.

Sperling (1961) has shown in a closely controlled labor-atory experiment that observers can discriminate well amongthree flash rates--40, 80, and 160 flashes per minute--if thesignal is not too close to threshold. At and near threshold,discriminabillty was poor and there was a tendency to inter-pret both the lower frequencies as the highest. Robinson (1959),conductini a flight test of several navigation light systems,

1 According to information supplied to Bureau of Aeronautics,

Department of the Navy, Czot. 1957 by Lytle Engineering and Manu-facturing Company.

- 19 -

including the Honeywell-Atkins, found that occasionally the80 PPM frequency was read as either 40 or 160. In particulaiwhen the presentation of 80 FPM was preceded by either theslow or fast frequency, the error tended toward reading theuAMum frequency as the extreme frequency at the other endof the scale. Signal intensities were well above threshold.It should be noted that the observer's task was interpretinginformation coded into the signals, so that errors includederrors in interpretation as well as errors in reading theflash frequency. Gebhard (1948), after some preliminary ob-servations, suggested that only 15 discriminable steps existbetween 30 and 1800 PPM, but in later work (Gebhard, 1949)found only six identifiable rates between 120 and 2100 PPM.The entire upper range of these studies must be consideredunavailable for navigation light systems, because of engi-neering difficulties and also because they are in the rangeof flicker fusion for the intensity levels important in col-lision avoidance. Cohen and Dinnerstein (1958) concludedthat no more than five discriminable rates may be used betweei15 and 720 FPM.

It seems unlikely that more than three, or at the mostfour, frequencies could be used in navigation light systems,•and that even this small number might result In reading errorsof some consequence.

The time required to "read" frequencies in such systemswill vary with the frequency, and at the lowest may be appre-ciable. At a rate of 40 flashes per minute, for example,using the criterion that at least 3 flashes will be requiredfor discrimination, it will take nearly 5 seconds to come toa decision with confidence. However, it may be possible toidentify 40 FPM (in a system consisting of three frequencies,

oO, 80, and 160 FPM) by eliminating the higher frequencieswell before the third flias at 40 FPM is seen.

Plash-Pattern Coding

The most versatile technique, flash-pattern coding, isalmost unavoidable if large amounts of information must becoded into a light system. Three proposed navigation lightsystems flash in such a way that successive dark intervalsvary in duration, the difference indicating sector. Projector(1959a) has proposed a system In which a dot-dash code identi-fies the left side of the aircraft and a dash-dot code theright side.

The use of various sets of dot-dash signals, a standardtechnique for communicating informvation in military and otherapplicatior~s, may be considered for navigation light systems.One version is the "continuous" presentation, which omits thedistinctively long dark interval between groups, with all dark

-20 -

intervals uniformly short in duration. An example would bea dot followed by a dash, with this sequence repeated imme-diately, without separation. Continuous presentations reducethe number of distinctive codes which are practicable--thereis no difference, for example, between dot-dash and dash-dot--but have the advantage of eliminating the long dark interval.Another variation is the use of groups consisting of two ormore subgroups. For example, a group may consist of a singledot, a long interval, and then a dot-dash, etc.

Flash pattern coding has been investigated by Cook andBeazley (1962). Three sets of signals were tested: 12 Morsecode letters, 12 continuoui signals, and 12 signals consistingof various arrangements of dots alone, including several ofthe subgroup type. The 12 Morse signals all consisted of 2or 3 elements. Two of these were repeated in the set of 12dots-only signals. The three sets of signals were evaluatedboth as sets of 12 signals and as Individual signals, in termsof the time required to learn them, the accuracy with whichthey were read, and the time it took to read them. The contin-uous set of signals was found inferior to either of the othersets in regard to all three criteria. Apparently the pausebetween signal groups is helpful to the observers in readingthe codes. Eleven of the Morse code stgnals and 6 of the dots-only signals were 4udged a: constitutinp effective sets for asystem.

While there are as yet no experimental data on which tobase further selection of useful codes, some additional con-siderations may be helpful. In the test, the Interflash anddot duration was .15 seconds, and the dash and intergroup(or Intersubgroup) durations were .45 seconds. On this basis,the longest complete cycle in the Morse code set, three dashes,was 1.80 seconds. Two of the selected dots-only signals re-quired 2.'0 seconds pe.- cycle. The two next-largest durationswere 2.10 seconds. It is thus apparent that the cycle timesfor the Morse code set are appreciably shorter than those forthe dots-only set.

It also appears likely that the Morse code set, especiallyif restricted to signals containing both dots and dashes (re-ducing the selected number to nine) may be more distinctiveagainst city-light backgrounds. Many lights in such backgroundsflash either regularly or erratically. It is unlikely, however,that any background lights will be seen to flash in a regularcode consisting of both dots and dashes.

The Morse code set (with only those symbols consisting ofdots and dashes) thus appears to be superior to the dots-onlyset because (a) it contains many more usable signals; (b) thetime required to present complete cycles of its codes is gener-ally much shorter; and (c) it is likely to be more distinctiveazainst background lights.

I I I I I I I I I I I I I I 21 I , i ,

Intensity Coding

Signal lights can convey information through differ-*noos In Intensity. For example, arty light seems to growless intense with increasing distance, and a more intenselight may be seen further away than a loes intense one,providing some clue to distance. Anti-collision lights aregenerally more intense than position lights, so that in mostcases one may infer that If only the anti-collision lightcan be seen, the aircraft Is farther away than if both theanti-collision light and a position light are visible.

Unfortunately a number of factors limit the value ofthis kind of information. One of the most important Is atten-uation of light as it passes through the atmosphere. Figure 2shows the range in miles at various transmissivitles of theatmosphere for three signal light intensities. These curvesare based on an assumed "practical" threshold Illumination of0.5 mile-candle. The three signal intensities, 20, 100, and500 candles, represent approximately the order of magnitudeof low-, medium-, and high-intensity lights In present-daynavigation light systems. The 3-mile VFR visibility require-ment corresponds approximately to an atmospheric transmissivityof 0.25/mile. It is evident from the figure that range esti-mated from apparent Intensity is subject to considerable ambi-guity, if the atmospheric transmissivity is unknown. Forexample, when the tranamissivity is 0.85/mile ("very clear")a 20-candlepower signal observed about 4 miles away wouldappear equal in intensity to a 500-candlepower light observedat the same distance when the transmissivity is 0.4/mile("light haze"). Or, to look at it another way, a 100-candle-power light signal would appear at the "practical" thresholdwhen it is nearly 8 miles away, If the transmissivity is 0.85/mile; but if the transmissivity is 0.4/mile, it would appearat the "practical" threshold at less than half that distance.

The appearance of signal lights is very sensitive to at-mospheric tranrmissivity, even within the limited range of VFR.Pilots can estimate the condition of the atmosphere from visualcues, from weather reports, etc., although such estimates areoften subject to considerable uncertainty. Because of thevalue of range information, pilots should be trained to makesuch estimates, however uncertain they may be. Studies haveshown that ability to estimate range can be greatly improvedby training (Applied Psychology Corporation, 1962c).

Another factor tending to reduce reliability of rangeinformation derived from signal lights is the variability ofIntensity, both within the distribution of any given light asthe sighting direction changes, and between different lights.

If signals of different intensities appear simultaneously

-22-

30- ASSUMED VISUAL THRESHOLD:

0.5 MILE-CANDLE(•iPRACTICAL" THRESHOLD)

20

15

13II

98

a7_j Jj 6

iJ4

2-

EXCEP-VERY TIONALLY

LIGHT HAZE . CLEAR CLEAR _CLEAR

•2 .3 .4 .5 .6 .7 .8 .9 1.0TRANSMISSION/MILE

Fig. 2. Range vs. atmiospheric transmissivity torlight intensities of 20, 100, and 500 candles, based on"practical" Illumination threshold of 0.5 mile-candle.(A "mile-candle" Is the illumination at a distance of1 mile produced by a source with an intensity of 1 candle.)

23 -

or In close succession, the pilot's ability to estimaterange may be somewhat improved, although the results ofa laboratory study of such a system do not support this(Applied Psychology Corporation, 1962d). If Lwo intensi-ties are used, each has a separate threshold and corres-ponding range. Some anti-collision l2gh~s, for example,have alternating flashes of different intensities. Whenpicked up near the limiting range, only the higher-inten-sity flash is visible, and the flash frequency t. half thenominal frequency. Close in, the weaker flash is seen andthe frequency dcubles to nominal frequency. In addition tothe ohange of frequency, the difference in intensity betweensuccessive flashes is noticeable. The anti-collision lightswith alternating flashes of diff-rent intensities are goodexamples of the kind of unreliability of this distance esti-mation technique, mentioned above. The higher intensityflash is achieved oy optics which at the same time producemuch narrower vertical beam spread. Thus the beam which isPigher in intensity near horizontal is lower in intensityat elevation angles greater' than about 5 or 10 degrees. Ifthe sighted aircraft is in a bark, and only one of the twobeams is visible, it might be interpreted that the aircraftis at a great distance when in fact it is very close by.

Intensity coding may help to refine judgments of range,but does not overcome the irnherent difficulties describedpreviously. It may oe that Luch coding together with train-ing designed to make maximum use of its potential can resultin significant improvements. The training should include es-timating atmospheric transmissivlty, to minimize the uncer-tainties contributed by this factor. Possible improvementwith trainir•g In estimating range has been shown in bothground-to-air and air-to-air tests (Applied Psychology Corpo-ration, 1962c).

Intensity coding can be used to convey other information,particularly if two or more intensities are presented togetherin a spatial array or close together in sequence. Visual es-timates Cf intensity are very uncertain, but comparative Judg-ments can be fairly precise. However, spatial arrays o$ in-tensity-coded lights are not considered to be of value.fSequential arrays may be useful. It is felt that such codingshould be reserved for estimating range, since (a) this is animportant item of information, (b) no better techniques forproviding it have been found, and (c) the connection betweenapparent intensity and range, even if subject to uncertainties,is a close one.

1 This will be discussed in the section on array coding,

Simple intensity codes can be read quickly. Estimatesof apparent intensity are direct and immediate. If a sequen-tial code is used, there is an inherent delay in presentingthe time sequence, but the delay can be quite short sincethe time required for each element in the sequence need notbe long.

Array Coding

Spatial arrays of lights may convey information, andoffer attractive possibilities of conveying detailed informa-tion rapidly. The technique has been used as a major or in-cidental component in a number of navigation light systems.

Static spatial arrays, if not too complex or ambiguous,convey their information very rapidly. Those which build upthe array in a time interval may be somewhat slower but arestill quite rapid if the time interval is short. In the Mad-sen system, for example, the time to complete a flashing se-quence is about 1/6 of a second, and the repetition rate (fora single row) is 40 per minute.

However, several difficulties in the use of array codinglimit its usefulness.

Spatial arrays vary in apparent size with the distanceof observation. Arrays whose angular dimensions are of theorder of one minute of arc or less will be difficult to inter-pret, especially if they are complex. Thus if an array'scross-sectional diameter is 6 feet, it will subtend I minuteof arc at a range of about 4 miles. While it is sometimespossible to locate the elements of an array in such a way asto obtain dimensional cross-sections of 6 feet or more, oftenit is difficult or impossible.

For example, in the systems using rows of sequentiallyflashed lights along the fuselage, adequate dimensions areobtainable along the fuselage when the observer is abeam, butas the direction of observation is moved to the front or rearthe row of lights is foreshortened until it becomes unreadableas a row. The precise limit of readability will of course de-pend on the length of the row and on the distance and angle ofobservation.

Arrays are much more feasible when the information isintended for use in limited directions. If the direction ofinterest is abeam, the length of the fuselage is availablefor the array. If the direction is fore or aft, the wingtipsprovide adequate separation. If the view is from above orbelow, both the fuselage and wings provide potentially largecross section.

- 2 -

If the kinds of information coded into an array require

that it have the same appearance for all directions of sight-Ing, then array coding is likely to be altogether impracti-cable. If the array-coding system codes information by means

of the change of the appearance of the array with direction,it is more practicable, in principle, to provide necessarylocations about the aircraft, but usually at the coat of In-creased complexity. An example of such an array is the Lytlesystem, in which three flashing lights are mounted at theforward end of the fuselage, one at the nose,and two backalong the fuselage, one on top and one on bottom. Head-on,the three lights are in a vertical line. On either side ofhead-on, the three lights form a triangle, the apex of which(at the nose light) indicates how far the view is from head-on.A similar arrangement at the rear, with the three rear lightsflashing at lower frequency, completes the system. Sector in-formation is provided to the side by the array and the flashfrequency coding of the four fuselage lights. This systemseems to provide excellent azimuthal sector information whichis unambiguous and simple to learn. It also exemplifies manyof the problems of array coding, including those already men-tioned and others. These are listed, but similar problems arelikely to arise in the design of any complete array-coded sys-tem.

1. From the point of view of backscatter, the threelights at the front of the fuselage are in the worst possiblelocation. Of these three, the nose light and the one on topof the fuselage are in particularly poor locations for anyappreciably intense light.

2. From several directions, one or more of the lightsin any given array may be occulted by the wings, the empennage,or other structures. If one or more lights of an array ismissing, the infortuation may be uninterp!'etable or misleading.With any navigation light system, it is often difficult toavoid occultation in some directions, but when an array codeis used the difficulty is compounded.

3. Precision is often degraded when the observer is noton the same altitude, even when the elevation angle is rela-tively small. Near dead ahead, for example, if the observeris at a lower altitude, the nose light and the upper forwardfuselage light are closer together and may even merge, makingthe information difficult to read or ambiguous. If they areseparable, however, the information may be adequate and mayeven give some Indication of the elevation angle.

4. Locating the lights to provide adequate cross sectionat required distances poses a problem. Forward, for example,the maximum dimenslon is the height of the fuselage. On a verylarge aircraft this may be adequate. On small aircraft, it may

I I I I I I I I I I I 26

be quite inadequate. In general, the problem of obtainingadequate array dimensions on smaller aircraft is especiallydifficult to solve.

Under certain conditions some array-coded systems givemisleading information. Those systems using sequentiallyflashed rows of lights to indicate the direction of flightmay sometimes appear to flash in the opposite direction.This occurs if the observer's speed is greater than the pro-Jected component, parallel to the observer's direction offlight, of the sum of the intruder's speed and the speed ofprogression of the row of lights. In a typical installation(9-foot spacing, between lights of a row, .075 seconds be-tween successive flashes) the speed of the lights Is about80 miles per hour. If the intruder aircraft carrjing theselights has a speed of 200 miles per hour, the sum of the twospeeds is 280 miles per hour. If the observer aircraft isabeam, moving on a parallel course at a speed of 280 milesper hour, the lights will appear to be successive flashesfrom a single light in a fixed position; if the speed isgreater than 280 miles per hour then the direction of thesuccessive flashes will appear to be opposite to the actualdirection of the observer's aircraft. Although in such arelative-course situation no collision impends, the informa-tion is nevertheless misleading, and in other situations notso clearly noncollision, the false information could lead toserious confusion--if not to collision.

Some array codes use the standardized (CAR) location ofnavigation lights, for example on the wingtips. Occasionallythis can operate satisfactorily, but here too may presentproblems. If both lights can be seen the wingtip array canconvey banking information, but aspect information obtainedfrom the array of a wingtip light and the taillight is moreambiguous if the intruder aircraft structure is unknown. Thespatial relation between the wingtip and tail differs fromairplane to airplane. On some the wingtip is well forward ofthe tail, on others much closer to it or even rearward of it.

Because of the difficulties with array coding, it shouldnot generally be a primary part of a navigation light system.It may have incidental value in systems employing other typesof coding.

Combination Coding

Most navigation light systems combine coding techni.ues.The CAR "standard" system codes azimuthal sectors by color anddistinguishes the red anti-collision light from the red positionlight by flash coding and, 'Zo a lesser extent, Intensity coding.(As with many other systemvi, incidental array coding of the wing-tip lights gives bank indi.cation in limited directions.) In the

r -

Ummeywell-Atkins system, flash-frequency coding providesazimuthal sector Information while the wingtip array. e-cause of a sizable overlap (45 degreeb on either side offore and aft) sires bank information over a substantialfield of view. The Lytle system uses flash-frequencycoding and array coding.

Techniques can be combined readily, often without sig-nificant lose of distinctiveness for ~either type in thecombInation. Color coding combined with any type of flashcoding can be read, except for any long, oar'k interval, withlittle or no more difficulty than If used alone. A systemfor presenting large amounts of aifferent types of informa-tion would almost certainly combine techniques.

One way of adding information to a system is "timesharing," presenting different information in successiveintervals, or distinguishing signals by giving then an alter-nating characteristic. Onie navigation light system, for exam-ple, hau two tail lights, one red and one white, flashingalternately, thus distinguishing them from the left wingtiplight, which is a simple flashing red light. Dot-dash flashpattern codes are similar to the red-white taillight, In.that distinctiveness and information are gained by havingsuccessive fLashes differ In some respect. The more iLpor-tant type of time-sharing combinations would present dIffer-ent Information successively,. No system using this kind oftime-sharing has yet been formally proposed, and it is notlikely to be advantageous unless very large amounts of in-formation are considered essential.

Coding techniques can be combined in numerous ways, andthe optimum combination for a given requirement may not beeasy to choose. The choice requires careful evaluation interm of the four criteria for individual techniques, listedat the beginning of this section, and must in addition con-sider interaction effects.

One particular combination is of special interest: flashcoding and Intensity coding. The discussion of flash codingnoted that the fixity-of-bearing criterion to determine acollision course is likely to be m~coh le3s uccurate if thetarget light flashes than if it is steady. Such reduced pre-cision can be avoided to some extent if the light pulsatesrather than flashes, so that the signal may be considered toconsist of a steady light and a superposed flashing light.The flashes, more intense than the steady component, provideintensity coding as described earlier. When first picked up,the steady component is invisible, and the appearance Is thatof a simple flashing light. As the distance Is reduced, '-hesteady component becomes visible and there are no dark inter-vals. Within the range that the steady component can be seen,

4w- 28 oo

the fixity criterion should be usable with nearly as much,perhaps as much, precision as with a simple steady light.The difficulty in using fixity at longer ranges would con-tinue in such a system, but is not too much of a loss be-cause at longer ranges the usefulness of the criterion isnot very great anyway (Calvert, 1958; Applied ?sychologyCorporation, 1961a).

- 2 I

IV. INTENSITY, EFFICIENCY, AND VISIBILITY

The visibility of a light signal may be defined interms of the threshold illumination it produces at theobserver's eye under a given set of conditions. Thisthreshold is determined by a large number of physical andpsychological factors, among which are (a) the signal'sintensity, size and shape, color, movement, and distancefrom the observer; (b) luminance of the background; (c)other lights in the field of view; (d) the atmosphere;(e) backscatter; (f) the optical quality and location ofthe cockpit windows; (g) the portion of the observerlsretina on which the signal image impinges; (h) the observ-er's adaptive state, physiological environment, individualvisual capability, alertness, and search habits; and (i)distractions diverting the observer's attention.

This discussion will consider the more important aspectsas they relate to optimum design and operational use of air-craft navigation light systems. Because of their special im-portance, effects of the atmosphere, backscatter, backgroundlights, and cockpit visibility will be discussed separatelyin subsequent sections.

Navigation lights, because of their small size and therelatively large distances at which they are observed may beconsidered "point sources": only their photometric intensityneed be considered in determining visibility. It has occasion-ally been proposed that extended sources be used in navigationlight systems, either by using extensive light sources, suchas linear sources installed along appreciable lengths of wingedges, or by floodlighting the surfaces of the aircraft; thefloodlights themselves would be invisible generally but wouldprovide a visible signal oy illuminating the surface, Aswill be shown later, extending light signals into lines orareas drastically reduces the efficiency with which electricalenergy is converted into visible light signals. Because ofthe need for high efficiency, navigation light signals shouldbe limited to point sources.

Movement of a light signal in a field of view reducesits apparent intensity. If the movement is rapid, a signalmay become invisible, although significant effects of movementon visual performance occur at rates far in excess of thoseinvolved in visual collision avoidance (Miller, 1958). How-ever, movement may affect an observerls ability to detect asignal when the background contains other signals. 1

1 This is discussed in the chapter on Lights and Backgrounds.

- 1-

Individuals vary considerably In all kinds or visualcharacteristics. It Is not possible in this report to die-cuss this in detail; In general, only average performanceis discussed.

Visual Thresholds

In the absence of atmospheric attenuation, the Illumi-nation, E, produced at an observer's eye, located at a dis-tance, D, from a source of intensity, I, is given by the"inverse square law":

EmD2

Thus the apparent Intensity of a source of a given in-tensity seen at a given distance is the same as that of asource of four times the Intensity seen twice as faj away,if the two sources are alike in all other respects.

The threshold illumination, Eo, has been extensively in-vestigated, but the results vary over an extraordinarily widerange. Some variation Is due to the considerable variabilityamong Individuals, and in any individual from time to time.Another source of variation Is the uncertainty of the criteriaused to determine threshold. Threshold judgments by observersare largely subjective, and are affected significantly by theInstruction they receive as well as by the experimental design.

The reported results are also related to the scoringtechniques. One comwonly used method considers thresholdthat value of Illumination at which the observer's accuracyscore is 50%., a scoring criterion which theoretically has anumber of advantages; but others feel that a threshold repre-senting a higher likelihood of seeing Is more meaningful, andthus use a scoring criterion of 90 or 95% accuracy.

The order of magnitude of threshold illumination forwhite point-source signals observed against dark backgroundswith a high probability of seeing under favorable con itionsis 0.01 mile-candle (Knoll, Tousey, & Hulburt, 1946). Actualfield sightings with observers not knowing where or when asignal may appear occur at considerably higher threshold illum-Inations and are wdch more variable than those obtained or

1 A more detailed discussion of range is given in Chapter V.

2 The illumination produced by a source of intensity .01

candle at a distance of 1 mile in perfectly clear air, forexample.

-32 -

cited by Knoll, Tousey & Hulburt. For estimation and speci-fication it is convenient to agree on some field value ofthreshold illumination which may be considered reasonablyrepresentative of search situations such as prevail in air-craft collision avoidance. Many experts in this area agreeon a "practical" or "useful" threshold of about 0.5 mile-candle (Middleton, 1958, p. 99; Technical Committee 3.3.2.1,C.I.E., 1955; Stiles, Bennett, & Green, 1937, p.30; Gilbert& Pearson, 1951; Adrian & Jainski, 1960). While the "practi-cal" threshold may be used for general calculation, verydifferent thresholds may apply In particular cases (Koomen &Dunkelman, 1953).

The value of threshold illumination is not much affectedby variations in background luminance in the range from totaldarkness to starlit sky; but as background luminance increasesabove this range, the threshold signal illumination increasesby as much as 100 times greater than that against a very darkback round (Middleton, 1958, p. 97f; Knoll et al, 1946; de Boer,1951).

Adaptation, Fixation, and Signal Color

The adaptive state of the observer has a profound andcomplex effect on his visual capability. Measured in termsof ability to detect the presence of a faint signal, thiseffect may change the detection threshold by several ordersof magnitude. The reduction in sensitivity caused by an ex-cess of brightress in the visual environment is related to anumber of factors, and in any given cockpit situation is like-ly to be difficult to analyze. In one investigation, citedby Stiles et 41 (1937), background adapting luminances fromless than 10-J (moonless night sky) up to nearly 500 ft-L(overcast sky) were utilized. The ratio of the sensitivityI second after the adapting luminance was turned off to thesensitivity after an hour of dark adaptation was determined.Ratios as high as 1000:1 were obtained. While the necessityfor maintaining a minimal luminance level in the cockpit inorder to carry our pilot functions precludes the possibilityof attaining complete dark adaptation, carelessness in light-ing the cockpit can easily produce disabilities in ratiosranging from 5:1 to 100:1 and gross carelessness can resultin ratios considerably greater. It is thus of great impor-tance that pilot's vision be maintained In as nearly fullydark-adapted condition as is consonant with performing., otherfunctions, and thus ambient illumination in the cockpit shouldbe at a minimum.

Absolute dark adaptation, which requires a half hour ormore in total darkness, is probably only rarely achieved underactual flight conditions. Intermediate states of dark adapta-tion are common and can be considerably affected by factors

- -

subject to control within the aircraft. Ambient luminancesoutside the cockpit impose the basic limit or, the extent ofdark adaptation possible. (Backscatter from lights on theobserver's aircraft may contribute significantly to ambientluminance.) However, high luminances in the cockpit mayreduce the degree of dark adaptation or delay recovery ofthe fullest possible adaptation for appreciable lengths oftime (Wulfeck, Weisz, & Raben, 1958).

Two types of receptor elements are distributed throughthe retina, the rods and the cones. In the fovea, the centralpart of the retina, cones are densely distributed and there areno rods. Cones are distributed through the peripheral orparafoveal parts of the retina, outside the fovea, but theirdensity is very much less than in the fovea. When the eye isfully dark adapted only the rods are involved in sensing veryfaint lights. When the eye is light adapted, only the conesare involved in seeing. In intermediate adaptive states, bothrods and cones may be involved. The cone system is largelyresponsible for sensing color and appreciating detail. Colorcannot be identified when only rod vision is involved. Whenthe eye is dark adapted the se- sitivity of the rods is muchgreater than that of the cones for all colors except deep reds.The rod threshold for a faint green signal is of the order of100 times lower than the cone threshold and approaches a valuenearly 1000 times lower at the extreme blue end of the spec-trum (Wulfeck et al, 1958).

With complete dark adaptation, it is thus possible todetect in the periphery faint green or blue signals which areinvisible when the eye is fixated directly on the signal andits image falls on the fovea. It is, however, uncertain howimportant a role this plays in visual collision avoidance,since complete dark adaptation is seldom if ever achievedduring flight. Signal illumination levels at which sightingactually occurs are likely to be well above the levels atwhich pure rod response prevails. Middleton (1958, p. 98)says the pilot probably never uses parafoveal vision to dis-cover lights, an opinion he bases on the unlikelihood ofachieving any high degree of dark adaptation (mentioningprincipally the need to look at lighted instruments) and thefactors which tend to raise visual thresholds, such as in-ferior window quality and aircraft movement. These factors,plus the fact that detection is usually accomplished withoutthe pilot's knowing where or when to look, leads to the com-paratively high "practical" threshold, and supports Middle-ton's contention that sightings are almost always foveal.Without conclusive evidence, however, peripheral sightingsshould not be entirely rejected. Cockpit instrument panelscan be designed and operated so that instruments can be readwithout serious loss of dark adaptation, and other factorssuggest that at least occasionally the high sensitivity and

-4 -

other characteristics of a good measure of dark adaptationmay be effective.

Some experimental work suggests that above absolutethreshold the apparent intensity (foveal) of a red zignallight is greater than that of a white or green signal.Middleton-&Gttfried (19571, for example, working at levelsin the neighborhood of the 'practical" threshold, found thatred signals of equal photometric intensity appeared as muchas twice as intense as white signals. Other experimentaldata indicates that the relative peripheral sensitivity tored light is appreciably higher for signals above absolutethreshold than at the absolute threshold. A similar increasein red sensitivity is observed when the eye is adapted to abrighter background (Kinney, 1958). The results of theselaboratory experiments are very difficult to apply to fieldsituations, and serve mainly to underline the extraordinarycomplexity of the psychophysical phenomena involved. In theabsence of definitive experiments which relate directly tovisual collision avoidance, sweeping conclusions, such asavoiding the use of red signals because of the low sensitiv-ity of the dark-adapted peripheral retina for red light,should be avoided.

In any event, when the color of the signal light must beidentified, it is clear that higher thresholds apply, thatfoveal viewing is probable or required, and that there are novery large differences in threshold for the colors of ordinarythree- or four-color systems (Middleton, 1958, p. 102).

lashin Lights

The effective intensity of flashing lights has beenstudied extensively under a great variety of experimentalconditions (Projector, 1957). The generally accepted methodfor computing the effective intensity of a signal flash ob-served at or near threshold Is that proposed by Douglas (1957)(based on the classical investigation of Blondel and Rey, 1912):

-l I dtf t2Id

Ie .,.

.2 + (t 2 - tl)where

Ie = the effective intensity

I a the instantaneous intensityat any time, t, during the flash

tI and t 2 * times near the beginning and end ofthe flash, selected to maximize Ie

- 35 -

i

"Analysis of this equation shows that the flashing of alight signal at Intervals makes more efficient use of lumi-I

nous energy in producing intensity (and therefore visibility)than distributing the same energy in a steady-burning light.The increase in utilization of energy is of course inverselyproportional to the frequency of the flash, but such increaseumust be weighed against the increase of the duration of thedark interval between flashes, when no signal is visible atall. It is also evident that short flashes are more efficientin producing high effective intensity than long ones (forconstant total flAsh energy), but that the benefit of reducingflash duration reaches a maximum when the flash duration isappreciably less than the constant in the denominator, 0.2second. Thus durations less than about 1/20 second are almostequally effective (for equal flash energy), no matter howshort they may be.

While the variation of effective intensity of flashinglights with such other variables as color and background luml-nance is often more complex than the variation of the apparentintensity of steady lights with these variables, relationshipsestablished for steady signals, subject to the above equation,may be used for roughly approximating flashing signal charac-teristics.

Efficlency

The efficiency with which electrical energy is convertedto visible light signals is of great importance in designingand evaluating navigation light systems. On any aircraft,however large or high powered, the unnecessary consumptionof Dower, or the addition of excess weight or unnecessarybvlk, results in measurable reductions in over-all performance.O:1.1 smaller aircraft such waste may be critical, and decisivelan determining a design's feasibility. Since standardizationof light systems for all aircraft is essential, It is evidentthat efficiency, alwadr-an important consideration, may dic-tate the choice among alternatives.

In the final engineering design of a system to meet speci-fled performance requirements, one or another particular com-ponent may be called for to produce maximum efficiency. Ingeneral, however, there are no categorically superior compo-nents--for example, particular light sources--which can beselected in advance as components of choice for meeting allpossible requirements.

The efficie-.!:y with which electrical energy is convertedinto luminous signa1i3 consists of two parts: the conversion ofelectrical energy to luminous energy, or the "conversion effi-ciency,6 and the conversion of the luminous energy to light

--.. 6

signals with the required intensity distribution, color andflash characteristic, or "signal effi'iency."

Conversion efficiency. The effi lency of conversion ofelectricti -to luminous energy varies over wide limits. Whitefluoresc,;nt lamps may have efficiencies as high as 75 lumensper watt (LPW1), green fluorescent lamps may produce as muchas 125 LPW. Conventional incandescent lamps have efficien-cies in the neighborhood of 10-20 lumens per watt. But theerficlencies of all light sources are largely dependent oncompromises with other design criteria, most particular)-the design life of the source and the required maintens .characteristics (the manner in which light output charL -during the life of the source). For example, short-lifephotoflood incandescent lamps emit well over 30 LPW.

The efficiency of auxiliary equipment necessary tooperate the source may significantly reduce the over-allefficiency of the equipment. For example, the conversionefficiency of a conventional condenser-discharge light saof the order of 30 to 40 lumen-seconds per watt-second.'But the efficiency of the auxiliary equipment necessary tooperate condenser-discharge lights may be of the order ofless than 50%. Thus the over-all conversion efficiency ofcondenser-discharge lights is comparable to that of incan-descent lamps (Mullis & 4"rojector, 1958).

Signal efficiency--intensity distribution. Navigationlights are ail requirea to -ave a 3lstrloutlon of intensityin space. Minimum intensities are specified in a2l direc-tions. The design of equipment to produce distributionswith sharp cutoffs in any direction is generally easier ifthe light source is small. Whenever the light emitted by alight source is controlled or redirected by lenses or mirrorsor baffles, some degree of inefficiency, often substantial, Isintroduced. Light energy may be partially absorbed in theprocess of redirection; it may be scattered or redirected inundesired directions; and it may be impossible to redirectsome of it as desired.

It is essential to analyze efficiency in terms of inten-sity distribution rather than intensity in any one direction.If high intensity is desired only narrowly in one direction,it is possible, by redirecting in the desired direction thelij'ht flux enitted by the source in other directions, to

1 Energy units are used to evaluate flashing source effi-

ciency rather than the rate-cf-energy-emission units used forsteady lights.

- 37 -

obtain enorously high intensities. Thus, for example, a600-watt landing light consisting of a small incandescentfilament source and a parabolic reflector produces a narrowbeam with a maximum intensity of over half a million candles.The efficiency with which the light flux is directed into thedesired beam, measured in beau lumens per watt, is consider-ably less than the conversion efficiency of the filamentsource itself, although the intensity of the latter, in theabsence of a reflector, would be only of the order of 1000candles.

Signal efficiency--color. When the color of the fluxe*ittei by a 11hit source must be changed by the use of fil-ters, efficiency may be substantially reduced. Thme reductionsare measurable in terms of the transmissions of the filtersfor the light emitted by the source. With incandescent sources,the transmission of red and green filters (to produce Aviationcolors) is of the order of 20% to 25%. If, however, the lightsource is a condenser-discharge light, the transmission forgreen is about the same as for an incandescent source, butthe red transmission is only about half as much, or even less,depending on the particular lamp. The transmission for a bluefilter would be appreciably higher. Virtually no filteringat all would be required for bluish white, whe:*eas the bluish-white filter required for an incandescent source might have atransmission of 25-50%. It is evident therefore that differ-ent light source-filter combinations designed to produce thesame signal color may have radically different efficiencies.

In general, color coding implies reduction of efficiency,and this must be taken into account in over-all evaluation ofa system.

Sigal efficlenc--flashig lights. The signal efficiencyof a rlasnirg l~gHt can De evaluated by extending the computa-tion technique of Douglas for effective intensity (Projector,1958a). This nethod takes into account not only the effectof flash duration and instantaneous intensity, but also thedistribution of effective intensity, which, in determiningthe efficiency of a steady light, is analogous to intensitydistribution. As with effective intensity, the shorter theduration of the flash for a given flash energy, the higher theefficiency. Below about 1/25 or 1/50 of a second there islittle to be gained by further reduction in duration.

Thus condenser-Cischarge lights, with flash durations ofthe order of a millisecond or less, make maximum use of lumi-nous energy, but any light source, emitting a flash of veryshort duration, can operate with an efficiency not significant-ly less. For example, if the flash duration is of the order of

-38 -

1/25 second, the efficiency is about 80% of the efficiencyof a condenser-discharge light with the same energy content.If, however, the duration is of the order of 1/5 second, theefficiency may be only half that attainable with a condenser-discharge light.

Incandescent lamps can be flashed electrically, by in-terrupting the current supply, or can be designed into fix-tures which produce flashing signals by occultation, byrotation, or by oscillation of directed beams. Because ofthe thermal inertia of incandescent filaments, short dura-tion, high-intensity flashes are difficult to obtain byelectrical flashing. It takes time to heat the filamentsto incandescence and to cool them off between flashes. Oc-cultation with shutters also involves inertia, mechanical inthis case, and may also reduce efficiency if the flux emittedby the source during the dark interval is not utilized. Ro-tating beams, and, to some extent, oscillating beams canreadily be designed to emit flashes of very short durationand high signal efficiency approaching that of condenser-discharge lamps.

If flash-pattern coding consisting of dots and dashes isused, it is not possible to obtain the high signal efficiencyof short-duration flashes, since the dashes must be of appre-ciable duration to be distinguished from the dots.

The relative signal efficiency of flashing lights ascompared to steady lights can be calculated (Projector, 1959b),but to do this, the repetition rate of the flashing lightmust be taken into account. For repetition rates of about60-80 flashes per minute, relatively short-duration flashinglights may have signal efficiencies five times as great ascomparable steady lights.

Signal efficiency--extended sources. All light signalsoriginate in real or apparent sources tnat have extent. How-ever, if under given conditions of observation this extentsubtends less than a certain visual angle, the observer seesit as a point source. This means that only the intensity ofthe signal determines its visibility. If, however, the visualangle of the source exceeds the maximum angle for point-sou'ceobservation, then the visibility depends on the size and shapeof the source and on its luminance. The relationship betweensource size and shape and luminance has been studied by anumber of investigators (de Boer, 1951).

When the results of these investigations are expressedin terms of the illumination produced at an observer's eye,we have a direct measure of the relative signrl efficiencyof the sources of different sizes and shapes. Results showclearly that sources having extent are substantially lessefficient in producing signal visibility than point sources.

-39 -

When viewed against a moderately dark background, a circularsource subtending a visual angle of one degree must emit fivetimes as much flux as a source subtending one minute of anglefor equal visibility. The maximum angle for effective point-source viewing Is about a half minute to 5 minutes, dependingon viewing con'itions.

A special category of extended sour'ces is linear sources,such as tubular fluorescent lamps or closely spaced rows ofindividual point sources, where the critical visual angle isexceeded In only one dimension. Losses in efficiency withsuch sources are not as great as with extended-area sources,but are of the same character.

Another special case is that of designs in which lightsources illuminate aircraft surface*, which then provide thevisible signal. The efficiency of such arrangements Is es-pecially poor. They not only have the inherent inefficiencyof extended-area, sources but are subject to losses In reflec-tion and to losses arising from the difficulty of directingall the light from the source onto the surfaces, and all ofthe light reflected from the surfaces, in desired directionsonly.

Thus, maintaining high signal efficiency in a navigationlight system can be achieved only by using effective pointsources in all system components.

Other Engineering Considerations

Detailed engineering design of navigation light systemsrequires consideration of a number of factors which differamong aircraft. Por any aircraft, the usual aeronauticalengineering objectives of minimum weight, high efficiency,reliability, safety, and low cost are applicable. Since astandard navigation light system suitable for all aircraftis essential, some of these considerations will be of greaterimportance for some installations than others. For small air-craft, minimum weight, high efficiency, and low cost may bemuch more decisive. For high-performance aircraft, findingsuitable locations for .ighting fixtures may be more diffi-cult. Capability of withstanding high temperatures may bean important requirement for some Installations but of littleimportance for others. It is difficult to deal with suchproblem generally, but a standard navigation light systemwill require full consideration of the special difficultiesthat may be encountered on particular classes of aircraft,and may even require that different, but compatible. systemsbe standard for different aircraft categorfes.

V. THE ATMOSPHERE, BACKSCATTER, AND VFR

In a perfectly clear atmosphere, the illumination, E,at the observer's eye, is related to the intensity, I, ofa light source and the distance of observation, D, by theinverse square law:

E.

The atmosphere is never perfectly clear--even theoret-ically pure air attenuates light to some extent--and ingeneral, the inverse square law must be modified to takeinto account the effect of the atmosphere. The modifiedlaw is Allard's Law (Middleton, 1958, p. 137):

E sI tDD2

where t is the transmissivity of the atmosphere, orthe transmission per unit distance (same units thatD is measured in)

General flight in controlled airspace under visualflight rules, at altitudes up to 14,000 feet, is permittedonly under atmospheric conditions which would result In atleast 3-mile flight visibility in the daytime.

Table 2 gives the designation, daylight visual range, andtransmissivity for several atmospheric classifications frithe International Visibility Code (U. S. Coast Guard, 195;).

The two components of a light signal's attenuation--that due to the inverse square law and that due to atmos-pheric effects--contribute quite differently to the re-sulting illuminance. This is illustrated in Fig. 3, agraph of the relation between illumination and distance forvarious transmissivitles, as given by Allard's Law. Thesource intensity used as a basis for the curves in the figureis 100 candles, the minimum intensity required by the CAR foranti-collision lights in all directions within 5 degrees ofhorizontal. Anti-collision lights that meet the requirementsare likely to have peak intensities in excess of the requiredminimum. To convert the values for illumination given 4n thegraph to values appropriate to sources of any intensity, itis necessary only to multiply the given value of E by theratio of the new value of intensity to 100 candles.

Curves for four values of transmissivity are plotted.The upper curve, for a value of t a 1, represents the inversesquare law without any atmospheric attenuation. The next

40 41 -

Table 2

Visual Properties of the Atmosphere;

International Visibility Code

Atmospheric DaylightDesignation Visual Tranamissivity

Range, (transamission/mile)Miles

Exceptionally clear over 31 over .88

Very clear 12 to 32 .73 to .88

Clear 6.2 to 12 .53 to .73

Light haze 2.5 to 6.2 .21 to .53

Haze 1.2 to 2.5 .044 to .21

-

• i

1000 ILLUMINATION VS. DISTANCE

FOR SOURCE INTENSITY= 100 CANDLES

FROM ALLARD'S LAW:

100E= tDU)

ILL

4 -'

U0LJ

2

Z9

-i -URCTIo

(03S MI."C)

0.1 ,0 1 2 3 4 5 6 7 8

DISTANCE, D, (MILES)Fig. 3. Illumination from a lOO-candle signal at

various distances for selected values of atmospheretransmissivity.

- 43 -

curve, for t . 0.8/mile, represents a "very cleax" atmos-ere, The third curve, for t a 0.6/mile, represents a

clear" atmosphere. The fourth curve is very important:the value for t, 0.25/mile, represents an atmospheric con-dition when the daytime visibility is about 3 miles, themlniam flight visibility for VPR operation in control zones,control areas,, and transition areas.

The 'practical" threshold of Illumination, 0.5 mile-candle, In indicated on the graph.

It is evident from the figure that at closer ranges inclear atmospheres the principle source of reduction In il-uin-Ination with distance is the Inverse square law. At longerranges and at lower atmospheric transuissivities atmosphericattepuation Is increasingly Important. Atmospheres repre-senting the limits of VPR are decisive in limiting visibilityat longer ranges.

The figure shows that for marginal atmospheres, when

t a 0.25/mile, the illumination is below the practical"threshold at 3 miles (the threshold range is 2-1/2 miles).Since the "practical" threshold is, as noted, a very roughapproximation, and since actual anti-collision light Intensi-ties exceed requirements to some extent, it may be Judgedthat anti-collision lights currently in use do provide approxi-mately the visual ranges required.

If however an aircraft does not have an anti-collisionlight, and has only the presently required position lights,the visual ranges may be seriously submarginal. The CAR, forexample, specify a minimum intensity of only 5 candles forthe wingtip lights in directions from 20 to 110 degrees out-board--l/20 thi' intensity for which the curves in the figurewere computed. Applying this ratio to the illumination scaleIn the figure, it can be seen that when the intensity is 5candles the "practical" threshold illumination will be at-tained at 3 miles only if the atmosphere is "perfectly clear;"any atmosphere denser than "clear" will produce distinctlysub-marginal illuminations. On the other hand, IncreasingIntensity from 100 to 1000 candles will Increase the thresh-old range from 2-1/2 to about 4 miles for the mariZnal atmos-phere, or from 4-1/2 to about 7 miles for a "clear" atmospherewith a transuissivity of 0.6/mile.

By Increasing intensity, increased range can be obtainedfor any atmospheric condition; but in the marginal condition,atmospheric attenuation becomes so overwhelming that the gainin visual range becomes less and less as intensity is in-creased. Figure 4 illustrates this effect.

I I I I I I I I I I I I 4 II4-

Fj t 0 3 e1

c'e I ( m0 EQUIVALENT DAYTIME VISIBILITY, MILESQ) (4 0~ 'z f

elo11: - W Oj'At-' (a " DqN 0 A ) OD 0 0 0

Ca El 0 ý-r. "I's. 0 01 II

OCA D

'CD0 b

I-A. ci' a1 " AD CD a 0) ct r

0'O*CD 0 '-SOCD' :T Z

Ca CD (4

(D'0' #1 ____-_

v ~~1 C 0 '1

Ca 0 'S 0Cr1)r c. () " CI

0C :30(ZiC 91 ps0 0

I1) (D Oq m

Ea 8D < 0

~CD c'Ib

HOQ it ca C ct~0 :: m0ca -_

*t0~ CD D :3m__ _ ____0___- >r~

Ht 00C~ D 0- 0~ 'V

m Oj -- h 0D

I- W z F

14 0 ct Oq 40' C D~ ct-h -

"" C (D p IC I CD c

Another possible approach to increaping visual range isto restrict VWR to narrower ranges of atmospheric conditions.It may be seen in the figure, for example, that if VFR flightwere limited to "clear" or better atmospheres, with a daytimevisibility of at least 6 miles, then the marginal visual rangefor a 100-candle signal would increase from about 2-1/2 to 4miles, the same increase that was obtained by increasing theintensity from 100 to 1000 candles for an atmosphere with3-mile visibility. Or, for a signal of 1000 candles, therange would be increased from about 4 to over 6-1/2 miles.If light intensity is increased to 1000 candles and VFR re-stricted to "clear" or better weather, then the inhcrease ofrange under marginal conditionz would be from 2-1/2 miles toover 6-1/2 miles.

Signal Color and the Atmosphere

Under certain conditions the atmosphere may stronglyaffect che color of observed lights. The yellow or red ofthe sun when on the horizon or when observed through dust orsmoke is a familiar phenomenon. Within the range of condi-tions of interest in navigation light systems, however, selec-tive attenuation effects of the atmosphere are seldom of suf-ficient magnitude to produce such striking effects.

The selective characteristics of atmospheric attenuationare very complex and in many respects not well understood(Middleton, 1958). It is possible to describe some of thesecharacteristics in a general way to indicate their applica-bility to navigation lights. 1

As noted earlier (see Fig. 3) reduction in signal illumi-nation with distance consists of two parts, that due to theinverse square law, and that due to atmospheric attenuationThe inverse equare law is nonselective; all colors follow pre-cisely the same decay curve.

The contribution of atmospheric attenuation to reducedillumination is the difference between the illumination at agiven distance as computed by the inverse square law and theactual illumination at that distance with a particular at-mosphere. Thus, referring to Fig. 3, the total atmospherictransmission at 5 miles, when the transmissivity is 0.8/mile,is 1.3 + 4 = 0.325. When the transmissivity is 0.6/mile, the

1 It should be emphasized that this discussion is limited tothe appearance of observed signal lirhts at night. The effectof the atmosphere on the observed colors of objects in the day-time is substantial (Applied Psychology Corporation, 1961b).The effect of color on backscatter will be discussed later.

46

total transmission at 5 miles is 0.3 + 4 = .075. For shorterdistances (or clearer atmospheres) transmission 1s apprecia-bly higher. At longer distances it falls rapidly.

Selectivity in atmospheric transmissi~ity can affect theappea-ance of observed signals of different colors in two ways:(a) the apparent intensities of the different colors can decaydifferently with distance, so that two signals that appearequal in intensity when observed at short ranges can appearunequal at long ranges, and (b) the observed color of the sig-nals can change with distance.

Atmospheric selectivity has been stuled by a number ofinvestigators, often with surprisingly different results.Sometimes no selectivity has been found at all, at other timesthe shorter wave lengths (blues and greens) have been found tohave higher transmissivities than the longer wave lengths (redsand yellows). Most often, however, and particularly for thekinds of atmospheres of greatest interest in VFR flight, thelonger wave lengths have been found to have higher transmissiv-ities than shorter ones (Middleton, 1958).

Those experimental results considered by Middleton to bemost representative or the atmospheric conditions importantin flight are presented in Table 3 in a form indicating themagnitude of the effects as related to operational situations.The data for red and green were obtained with lamps and filtersof the general type that might be used to produce these colorsin navigation lights.

It is evident from the table that the effect of selectiv-ity on apparent intensity is not very large. The largest dif-ference in the table is in the ratio 2:1 approximately, oc-curring et 10 miles for green transmissivities .60 and .53/mile, and at 5 miles for green transmissivity .25/mile. Butcomparison with Fig. 4 shows that all three conditions wouldproduce signals well below the practical threshold even fcia l000-candle signal. The next largest difference is about1-1/2:1 at distances and transmissivities above threshold fora 1000-candle signal but below threshold for a 100-candle s ig-nal; differences in illumination of such magnitude are notconsidered very important. Most of the other differences areconsiderably smaller than 1-1/2:1. In sum, atmospheric selec-tivity may be considered to have only a minor effect on theapparent relative intensities of red and green signals. 1

1 White signals and yellow signals are intermediately af-;o fected, so that the difference between red and green is the

maximum to be found in navigation light systems. Blue signalsare attenuated differently, about as much more than green asred is less, but blue signals are not useful for long-rangez.Jgnaling and are not considered here (Middleton, 1958, p. 173).

- 47 -

0 %0 LA -400) A 00

0 .n4)c

0c U00 Ccm

0 *o 0 r- %

0) %0 AtmI"4)

ca 00 N-9 A c 0 0 '

".4) 00 '0 * H

09) 0%e~4-I 41 CA 4f0 8-

e-4 r.Nq 0 a *

'.440 cn

4) 09 03 >b

q4 -H 05 0

".4r 'v-44-) >~ %'

0 4 0 A N -0 4 C"0 a ) 4 0 r%

0- 94. -a

o00E-4 *

4) 02 S: > -$4 *- 0 WQg :c 0 w4 L00 . 0) 0% 4 0 LA N

H. H C i

4)o a

A -H

48.)

111iimi Mlk 1 jý t j ,. iit,4

The effect of atmospheric celectivity on the apparentcolor of signals depends not only on the magnitude of selec-tivity itself but also on the particular color. In general,longer wave lengths are less attenuated than shorter ones,and the apparent shift of all colors is toward the red endof the spectrum. Red itself is not noticeably affected,however, since all the wave lengths that compose red signals(of the kind used in navigation lights) are within a com-paratively narrow band. Green signals generally consist ofa broader spectrum of wave lengths from blue to yellow.Selective attenuation may make the color somewhat more yellow.However, with most broad-spectrum greens the magnitude ofthe strong components of the color in the green region ofthe spectrum is much larger than that of the blue or yellowcomponents, 3o that even in the more selective situationsthe total shift or color is relatively small. In the useof white signals, however,there are substantial componentsthroughout the spectrum. White signals therefore may benoticeably yellowed if observed when the selectivity is high.Even in such cases, the shift would rarely result in a con-fusion with red signals. However, if distinction from yellowsignals is required, it may be necessary to consider the

Kl possibility of confusion.

If the limits of VFR are changed to the boundary be-tween light haze and clear, the maximum color shift at anygiven distance will be reduced. The green-red ratio of trans-mission increases with atmospheric density for a given range,as may be noted in Table 3. Significant color shifts wouldthus be limited to very long ranges. If color identificationerrors occur at ýhe3e ranges, they are less likely to lead totrouble and would in any event be readily corrected as asighted aircraft closes in and the range is reduced.

At or near threshold, color- identification is more dif-ficult than at higher levels, the degree of difficulty vary-ing with the color (Stiles, Bennett & Green, 1937; Rautian &Speranskaia, 1955). The "practica~l" threshold, 0.5 mile-candle, is close to the color identification threshood, butmost observers suggest that somewhat higher thresholds, ofthe order of 2 mile-candles, are preferable if a high orderof certainty of identification ia reouired (Stiles, et al,1937; Rautian & Speranskaia, 1955; Hill, 1947).

Backscatter

The attenuation of light signals transmitted through theatmosphere results from absorption and scattering (due to re-flection, refraction, and diffraction). In "aerosols" (at-

I mospheres consisting essentially of air and water vapor)absorption of light is very small compared to scattering,and can usually be neglected (Middleton, 1958). It is only

4- 9 "

In atmoaphere5 containing large amounts of Industrialsmokes or dust that absorption may be significant.

Light Is scattered by the aerosol in all directions.That scattered at and near 180 degrees (the direction backto the light source) constitutes the "backscatter,t" and isof special importance. When backscatter Is very large, asin the case of fog or clouds, it can be so disturbing thata pilot must turn off his navigation lights to avoid seriousImpairment of his vision. However, such gross effects occurwhen the atmospheric condition is well outside VPR limits.However, the disturbing effects of backacatter may be signif-icantly large even witnin the limited range of "P"R atmos-pheres, and may impose important design requirements onnavigation-light systems.

The deleterious effects of backscatter may ivnlude (a)direct reduction of light-signal visibility by increase ofthe background luminance against which the signals are seen,(b) indirect reduction of signal visibility by loss of darkadaptation, ard (c) more or less 3evere distracting or dis-orienting effects on a pilot, especially if the light-sourceflashes. (In this connection it may be noted that backscatterfrom flashing signals originating in rotating beacons mayhave different effects from that originating in electricallyflashed lights.)

Visual ra,'ge (or transmissivity) and backscatter areclosely related (Curcio & Knestrick, 1958). There is verylittle backscatter in the clearest atmospheres, but back-scatter can reach serious levels near the limits of VFR.Curcio and Knestrick found a relation between meteorologicalrange, V, and the intensity, S, of backscatter:

V- Cs1.5'

where C is a constant.

To the extent tht. transmissivity is selective, so alsois backscatter. As noted earlier, selectivity varies withthe attenuating components that make up any given atmosphere.For aerosols In the range of interest in VPR, selective trans-missivities are shown in Table 3, indicating a higher trans-mission for red than for green. The corresponding backscatterwill be less for red than for green. While the effects of 3e-lectivit7 'are generally not large, they suggest that redsignals are preferable to shorter wave-length signals, bothbecause of higher transmissivity and because of lower back-scatter.

A more Important effect, favorinpg the use of red 31g-nals wherc backscatter- is a problem, arises from the spec-tral sensitivity distribution of the different parts of theretina. The central part of the eye is much more sensitiveto red light than is the periphery. Thus when the eye inmore or less dark adapted, it is much l.css likely that redsignal sightings will occur in the periphery. By the sametoken, however, red backscattered light will h0 far lessnoticeable in the periphery than will light of other colors.Since the fovea is relatively small, subtending approximately5 degrees of visual angle, and the periphery very much larger,the apparent difference between red backscatter and back-scatter from other colors is often strikingly large, eventhough the ability to detect signal-s in the forea iz notgreatly affected. Knoll et al (1946) have shown that thevisibility of a point source is affected only by its adjacentbackground, within about 0.4 degree.

The intensity of backscattered light is proportional tothe intensity of the sinal light that causes it. Thus asthe intensity of navigation lights Increases, backscatterbecomes more important and may on occasion decisively affectthe design of a light. The problems of backscatter ariseprimarily in connection with light projected forward intothe pilotts field of view; backncatter from rear projectinglights is -Lmportý.&it only when a p_'lot searches ýo the sideor rear.

A series of tests of the abllty of observers to detecttarget lights through backscatter and to Identify color isdescribed in Applied Psychology Corporation Technical ReportsNo3. 6, 9, and 14 (1962a). Theze tests were carried out ata field visibility te3t range, and all test conditions werecarefully controlled with the exceptior, of the properties ofthe atmosphere, whose transmissivity was measured. Red,green, and white target and ba'zkscatter lights were used,the backscatter light was operated both steady and flashing,and both foveal and peripheral :!ghting were tried.

Surprisingly, the results were largely negative. Thecondition of the backscatter light--its color, or whether itflashed or burned steadlly--had little or no effect on theability of pilots to detect the targets or to identify theircolor. These negative result- were obtained with a back-scatter light of relatively high intensity (about 5500 candles),and in much of the test the atmosphere was close to the limit-Ing VFR atmosphere c Jr'respondine to a daylight visibility of3 miles.

The significant differences that dld appear in the datawere generally attributable to atmosphere attenuation and toselectivity in the attenuation, and to the difference in eye

- 51 -

sensit•v!ty ftr different tar$1'-t c:1cor por1phereily anCfoveally.

It thus appears that backcncatter 1n the aourtht iikelyto occur with navigation light zyatems, even If they Includecomponents of appreciably higher intensities than are nowused, w1ll not significantly affect the detectability oridentifiabllity of intruder lights.

There are, however, other possible effects of back-scatter which may be important. These include, as notedpreviou-ly, distracting or disorienting effects. Pilotshave noticed and reported strong disturbing effects. Thesereports suggest that there are significant differences be-tween flashing and steady backscatter, and between back-scatter of different colors. Investigation of the psych-logical effects of backscatter will be increasingly impor-tant as high-intensity navigation lights come into use.

It is therefore u!!eful at the present time to designnavigation light systems that minimize backscatter as muchas possible, even though some of its expected deleteriouseffects have not been confirmed by tests. Backscatter effectson a pilot c-an bc minimized in a number of ways;

1. By lateral separation of the light source from thepilot. Thus, wingtips constitute the most favorable locationfor forward-projecting lights.

2. By longitudinal separation of the light source fromthe pilot. This does not result in as great a reduction inbackscatter as does lateral separation, but the divergence oflight beams results in a marked reduction in luminous fluxdensity (and with it the brightness of backscatter) with din-tance from the source, particularly very near the source.It is thus advantageous to locate a fuselage-mounted light(an anti-collision light for example) as far back from thecockpit as practicable.

3. By intensity distributions with sharp cutoffs in-board (this should be coupled with lateral separation). Thi3results in very low backscatter brightness in the areas di-rectly forward of the cockvlit, but permits high signal in-tensity forward for the benefit of the pilot of an Intruderaircraft.

4. By limiting high-intensity forward-projected lightsignals to red, for reasons described above.

-52-

VI. LIGHITS IN: ThIi DAYTIFE

In Chapter iV it was indicated that the thresholdillumination of a light signal is affected by the lumi-nance of the background against which it is seen. As theluminance of the background increases,the threshold illumi-natlon increases. For dark backgrounds up to the range ozmoonless starlit night skies, the 4ncrease in thresholdsignal illumination is not large and does not significantlyaffect the "practical" threshold. As sky luminance in-creasee above that of the moonleas sky. the threshold il-lumination rises rapidly. In a bright moonlit sky, thethreshold is about 100 times greater than in a dark sky.In the daytime thresholds are 1000 to 10,000 times higher(Blackwell, 1966; Knoll et al., 1946; Jainski, 1960).

Thus, extremely high inten•itie3 would be required toprovide adequate signal visibilities throughout daylightconditions. The intensity required to provide a visualrange of 3 miles when the meteo.'ological range is 3 milesand the background luminance is 1000 footlamberts is about250,000 candles. Even this intenzIty Is not enough, however.While 1000 footlamberts is fairly representative of clearblue sky luminance, the luminancez o fparticular portions ofthe skvy often attain much higher values. For example, Hop-kinson(1954) measured values az high as 8000 footlambertsnear the sun. The sur, itself has a maximum luminance abouthalf-a-billion footlamberts. The extreme difficulty oflooki.ng into the sun, or even near it, makes it generallyimpracticable to attempt sightinZ in the3e directions Byusing dark glasse3, especially wher. the z'*n's disc itselfcan bc obstructed (for example, by aligning it with a partof the aircraft structure) it fz zometimes possible to searchin the neighborhood of the zun. Even so, It is far beyondpracticability with the aid of lights.

InteriL'c• 4£ 25v,000 candies can be produced in light-ing equipment, but only in limited directions and with con-siderably higher expenditures of power (and increased weightand bulk) than are likely to be considered feasible, even forvery large aircraft. Six-hundred-watt landing lights, forexample, have peak intennities of about 600,000 candles, buttheir beam widths, measured at the points of 10% of peak in-tensity, are only about l1x12 degrees. Complete coveragew.th a reasonable distribution of such high intensities inall directions would require many thousands of watts andconsiderable bulk and weight and would present many otherengineering problems such as high temperatures, reduced aero-dynamic performance resulting from necessary projections, etc.

-53 -

It is ciear then that ýIghts cannot be expected toprovide a complete synrtem fer daytime visual effectiveness.Nevertheless, daylight incl..des visual conditions notsharply separated from n4ght. It is therefore appropriateto inquire how far into daylight conditions and under whatparticular conditions lights may be useful.

According to the CAR, navigation lights are to beturned on between the hours of sunset and sunri3e. Theluminance of the sky near the horizon at sunset or sunriseon clear days is as low as 100 footlamberts. (Koomen, Lock,Parker, Scolnick, Tousey, & Hulburt, 1952) If clouds are

present, much lower luminances may occur. Furthermore, thebackgrounds against which aircraft may be sighted may bedark clouds or dark terrain rather than clear sky. Thedata of Iholl et al. (1946) suggest that at a backgroundluminance of 100 footlamberts threshold signal illuminationwill be 100 to 1000 times the threshold on a starlit night.If this is compared with Fig. 3 of the last chapter, it isevident that with intensities of 100 candles, raising thethreshold by a factor of 100 will result in sighting dis-tances against 100 footliabert backgrounds of well under2 miles.

It is evident that navigation lights conforming to

present requirementL !inimum of 100 candles in the horizon-tal plane for anti-collision lights) give substantially less

than 3-mile visibility under many marginal conditivns intwilight. Aircraft without antl-collision lights, havingonly conventional position lights, are required to presRentonly 5 candles of intensity from 20 to 110 degrees outboardon each side of the aircraft. Against a dark background, asignal with an intensity of 5 candles would have a "practi-cal" visuai range of about 3 miles in a "perfectly clear"atmosphere, but in an atmosphere with 3-mile visibility, therange would be only about 1-1/3 miles. Against a backgroundwith a luminance of 100 footlamberts, such a signal wouldhave a negligible visual range.

With lights that barely meet the present minimum re-quirements, there is some question whether their intensitiesare adequate even to cover the hours from sunset to sunrise.On the other hand, backgrounds at sunset or sunrise may haveluminances appreciably less than 100 footlamberts, the valueon which the above discussion is based. It may therefore bereasonable to turn navigation lights on before 3unset (orkeep them on after sunr.!se) to take advantage of the possi-bility, however small, that some conspicuity may be gained.

It is of value to compare the daytime visibility of theaircraft with the visibility of the light signal itself. Be-cause of the many variables in each situation, this comparison

.54-

is complex. A3 indicated earlier, light signals can bevisible in the daytime if their intensity can be made veryhigh. But only large aircraft could conceivably, providepower for sufficiently high intensities, and the large air-craft are already more visible because of their size.

Howell (1958) investigated the visual range of high in-ten.zity lights in the daytime and of the aircraft carryingthe lights. Three sizes of lights were used, each mountedin an oscillating fixture so that the signal was a flashingwhite light. The lamp power consumptions were 250, 600, and .900 watts. The effective intensities of these signals areestimated to be (very approximftely) 15,000, 50,000, and200,000 candles, respectively.

The 250-watt unit with an estimated effective intensityof 15,000 candles was seen appreciably less far than the air-craft itself (a DC-3); the aircraft threshold averaged 17.1miles and the signal light threshold 8.3 miles.

The 600-watt unit with an estimated effective intsnsityof 50,000 cindles was seen on the average about as far asthe aircraft itself. The 900-watt unit with an effectiveintensity of 200,000 candles w s seen farther away than theaircraft in each of four runs.

Data such as the above suggest that little is to begained from daytime use of lights of practicable intensities.If the data are examined in detail, however, the suggestionthat lights can be of occasional value in the daytime appearsto have more merit. T'he extreme variability of daytime visi-bility of the aircraft is poor. If this should occur whenthe background is dark, lights may be helpful.

1 Static data "were given in the report for the intensitydistributions of the lights used, but insufficient dat2 aboutthe oscillatory mechanism to determine the time-intensity dis-tribution. Furthermore, there is considerable uncertainty inthe value of the constant, a, in the Blondel-Rey equation, foreffective intensity computations appropriate to daylight view-ing. The estimates of the effective intensities are thereforevery crude and indicate only the order of magnitude.

2 The sighting distances in this series of tests (from 10to 20 milea) were well in excess of those obtained under normaloperational conditions. Howell (1957) found that sighting cis-tances in operational situations are 3 times shorter than thosefor observers who knew where and when to look and were not dis-tracted by other duties. Marshall and Fisher (1959) found azimilar ratio of sighting ranges for a small aircraft in a ter-minal area but at very much shorter distances and with a numberof completely missed detections.

- 5, -

A contrast of an aircraft target against a baci:groundmay be positive, negative, or a mixture of'both. The con-trast of a light against a background is always positive.

A high negative contrast, as occurs, for example, witha backlighted aircraft seen against a background of brightsky, is probably the oommonest mode of seeing aircraft -ar-gets. A light signal, unless it is overwhelmingly intense,cannot improve the conspicuity of such an aircraft. (On rareoccasions, it may even noticeably decrease if by effectingan apparent increase in the luminance of the target aircraft.)If, however, the background is dark and the contrast is posi-tive or small, lights of moderately high intensity may im-prove conspicuity.

Under a heavy overcast, sky luminances as low as 13 foot-lamberts have been found just before sunset (Hopkinson, 1954).Middleton (1958) has summarizedthe available data on theluminance of the sky near the horizon at sunset on an over-cast day and found it to average about 3 footlamberts. Darkclouds and some terrestrial backgrounds may often have lumi-nances in midday of 100 footlamberts or less, and terrestrialbackgrounds near sunrise and sunset may be very dark.

Thus, while lights with the intensities available todayare of little value in the daytime, their occasional useful-ness under marginal conditions suggests that they be turnedon well before sunset, kept on until well after sunrise, andturned on on all other occasions when there is a possibility

that they may serve a useful function. As higher-intensitylights become available, usefulness will extend further intomarginal daylight conditions, and rules governing their useshould be correspondingly changed.

The above discussion is based on the probable sightingF range of lights in daytime or in marginal lighting conditions.

In addition to providing detectability, navigation light sys-tems serve the function of giving sector-coded aspect infor-mation, which is of value within sighting ranges. Por example,in the vicinity of airports, traffic mu~st be kept in sight andtracked at separation distances cften less than one mile. Insuch cases navigation lights may be helpful and should be usedwell into marginal daylight conditions.

It is conceivable that the most effective lights for uay-time use may be different from those used at night. Back-scatter, for example, which dictates the use of red as th?color of anti-collision lights, is not a daytime problem.It may be desirable therefore to usa a white anti-collisionlight in the daytime with five times the intensity of presently

-56-

used red anti-collision lights. On the other hand, sincepresent lights extend sa- little into marginal daylight andare therefore almost ex•'-usIvely night lights, it may bepreferable to avoid nonstandard lights and use the same systemin the daytime as at night.

I I57 -

Vii. LIGHTS A1.D BACKGROUNDS

It was shown in Chapter IV that the threshold illumi-nation of a signal light increases as the luminance ofthe background against which it Is seen is increased. Whenthe background itself consists o:. numbers of small lights,such as the lights of a city or a starlit sky, an elementof confusion enters, and an otherwise visible signal maybe undetected. Against a starlit sky, where all the back-ground lights are of a single character, most aircraftsignals are fairly distinctive, especially if they appearmore intense than the stars, and can be distinguished bytheir color or flashing characteristic as aircraft lights.City lights and ground or tower obstruction marker lightsmay be extremely diverse in color, intensity, flashingcharacteristic, etc., and may be easily confused with air-crfst lights. If ground lights emit signals similar incharacter to aircraft lights (as is the case for example,with red flashing ground obstruction markers and red flash-ing anti-collision lights), identification is possible onlyby using secondary visual cues such as relative movement.

Under cruising conditions, at intermediate and highaltitudes, it seldom happens that the confusion of aircraftlights with ground lights is serious, if the intruder air-craft is near enough to one's own altitude to constitute apossible collision threat, ground llghts in the backgroundare likely to be far away and faint. At low altitudes, how-ever, especially in densely populated areas and terminalareas, the difficulty of distinguishing aircraft lights maybe very great.

In terminal areas, a pilot may be confronted not onlywith the general difficulty of distinguishing aircraft lightsfrom ground lights, but he may have to do this for severalaircraft at the same time. Once a light has been detected,any motion against the background of ground lights will dis-tinguish it as an aircraft light. But under certain relativespeed conditions there may be no motion against the back-ground, or the motion may be very small.

The general subject of the distinctiveness of lightsagainst backgrounds of other lights has been investigated,

1 This difficulty is unnecessarily compounded by lack ofstandardization. If the number of different signals thatare found on aircraft can be reduced, it will be easierfor a pilot to identify many lights as ground lights.

but not in the context of aircraft operations. The resultsare of interest here, but definitive experiments that includethe Important operational factors are yet to be undertaken.

Langmuir and Westendorp (1931) investigated the con-spiculty of a flashing light against backgrounds of numbersI of steady lights. They found that if the target light waswell separated from them .(3 degrees), the background lightshad little effect on target visibility, even when the back-ground lights were hundreds of times more intense than thetarget. However, when the target light was close to any ofthe background lights (0,75 degree) then intense backgroundlights multiplied detection time by as much as five times.

In two experiments by Crawford (1959, 1960), the effectof steady and flashing background lights on the detectabil-ity of a steady or a flashing target light was studied.The target light was always yellow and the background lightsred and green. It was found that the most conspicuous tar-get was a flashing light against a background of steadylights, and the least conspicuous a flashing light againsta background of flashing lights. When various mixtures ofsteady and flashing lights were used as background the ad-vantage of a flashing light over a steady light was lostwhen only a very few of the backaground lights flashed.

The experiments described above were done in staticconditions. An experiment carried out under more realisticconditions is reported in Applied Psychology CorporationTechnical Report No. 13 (1962e). The backgrounds were ac-tual city lights as seen from a nearby mountain, Whilethere was no relative motion (the observers and the targetlights were stationary during the observations), the dynamiccharacter of the lights in the background (moving vehicles,flashing signs, etc.) was a factor in the test design.Twelve distinctive parts of the background were used, rangingfrom rural unlighted areas to intensely lighted commercialsections. Red, green, and white signals were used, and theywere flashed either regularly or in a dot-dash code. Asmeasured by the time it took to discover them the red sig-nals were found to be more conspicuous than white. Greensignals were intermediate between white and red. No differ-ence was found between the two modes of flashing the signals.

While experiments such as the above are suggestive theycannot be considered conclusive. The work of Crawford suggeststhat there are not likely to be general conclusions valid forall signals against all backgrounds.

It is possible, however, to present some tentative gen-eralizations.

I60I

1. Increased intensity will help distinguish air-craft lights from backgrounds.

2. Plashing aircraft lights will provide good con-spiculty except against backgrounds that have a number offlashing lights, especially when they have characteristicssimilar to the signal.

3. The regularity of the flash of aircraft lights,Swhether pattern coded or not, probably helps distingL•ish

them from background lights that flash erratically or thatflash in regular but markedly different ways.

14. Pattern coding of flashing lights probably doesnot add to conspicuity, but should help resolve ambiguity.For example, an anti-collision light coded with a dot-dashI altitude code can be distinguished readily from a regularlyflashing uncoded tower obstruction light.

5. The most reliable factor that will help identifyan aircraft light is relative motion against the background.This motion should also help make the light more conspicuous.

IIIf

- 61 -

VIII. SEARCH TECHNIQUES AND COCKPIT VISIBILITY

Most collisions occur at times when atmospheric eon-ditions are favorable for visual sighting of aircraft, but,for one reason or another, the colliding aircraft Is notdetected at all or is detected much too late. Methods forincreasing the conspiculty of aircraft lights have been din-cussed elsewhere in this report, but many factors withinthe cserving pilot's cockpit pr-ofoundly affect the likeli-hood of detection of a target. These include the pilot'ssearch habits and technique, and the physical arrangementsof the cockpit itself: the visibility afforded by thewindows,, the quality of the glazing, and the visual environ-ment.

Marny of these f actors are much the same at night andduring the daytime.. The quality of the glazing, for example,similarly affects both day and night visibility. Inadequatewindow area or structural obstructions in the field of viewlimit visibility in the same way both day and night. Insome respects, however, visual processes differ materially,and this difference has a bearing on search habits andeffectiveness.

It has long been recognized that visibility from thecockpits of modern aircraft is severely restricted, so muchso that for some collision approaches the pilots of bothaircraft are unable to see the other, regardless of con-spicuity (Edwards and Howell, 1956; Fisher and Howell, 1957*Calvert, 1958; Zeller and Burke, 1958; Wulfeck et al., 1958).Conflicting demands make it impossible to provide sufficientcockpit visibility to meet the requirements of visual coll-sion avoidance. It should nevertheless be possible to pro-vide better visibility than now exists and particularly toeliminate "double blind" approaches.

In general, there are no "safe" directions of approaches,although under present conditions it is too much to expect apilot being overtaken to detect and be responsible for de-tecting the overtaking aircraft. Nevertheless, responsi-bility should be placed on both pilots, requiring that presentvisibility limits be expanded not only in the horizontal planebut vertically as well.

1 A discussion of some aspects of daytime search is con-

tained in Applied Psychology Corporation Final Report No. 1,The Role of Paint in Mid-Air Collision Prevention (1961b).

I6I

Pilots are expected, when flying VFR, to see and avoidother aricraft. But they are required to spend much of theirtime on other visual tasks. On long flights they are subjectto fatige, which lowers their visual alertness. Furtherore,a piloers attention is limited to a single direction at atime, and complete, effective coverage of the field of viewmust take tpoe. In thdudngthe maxymm visual effectivenesstIs lmited to a small region in the center of the visualfield--so much so that most detections are believed to occurhen the eyes are directed narrowly toward the intruder.Lamvr, 1960; Middleton, 1958).

The proportian of the eye at excht can differ radicallyfrom the properties during the daytime, affecting both the

effectiveness of search and the kind of search techniquethat should be used. Visual capability in the periphery ofthe llght-adapted eye In very poor compared to that of the

fopea vhence the likelihood that daytime detection will takeplace n the center of the eyel, When the eye s a dark adapted,however, renuitevety of the peolpherb to a light slinal ismuch hgher than thet of the foreas, except for red l95ht.It is thus possible, if dark adaptation can be achieved,that search can be effective through the much lakier peri-pheral visual field and detection can be made without targetfixation. In this case., much less scarzdmg would be required

for adequate coverkgp, and it would be much loens iely thattargets would be missed (Mertens, 1956).

Unf~ortunately, complete dark sdaptatlon is seldom pos-sible, Light, both inside and outside the cockpit, pre-cludes ;his in general. Middleton has stated the ambientlight insude the cockpit Is likely to be so unconductve to

dark adaptation that it is likely that sightings are actuallymade foveally at night as well as in the daytime. The so-called "practical" threshold of signal illumination (seeChapter IV) is well above the actual limits of visual sensi-tivity, suggesting that the high sensitivity of the fullydark-adapted eye is not exploited in operational situations.

Dark adaptation Is a matter of degree. Between completedark adaptation and full light adeptation there is a con-tinuum of intermediate states. In this range, near the darkend, there are sensitivity advantages to be gained eventhough they are not of the large magnitude characterizingcomplete dark adaptation. By Judicious control of the am-bient light in the cockpit (the instrument lights, lightfrom the cabin, etc.) pilots can often achieve and maintaina fair degree-of dark adaptation, which can maximize sensi-tivity generally, and may occasionally result in peripheralsightings that might otherwise be missed.

Scanning usually consists of a series of eye movements,preferably covering the field of view systematically. Ifthe target light is flashing, and no considerable dark adap-tation has been achieved, trin detection is likely to occurif the flash takes place while the eyes are fixated in theneighborhood of the target. If the flash occurs when theeyes are moving, It is less likely to be seen than if theeyes are stationary. Measures of the effective intensityof flashing lights are based on steadily fixated eyes; itis presumed that the energy from the flash is used to forman image at a single point on the retina. If because of eyemovement the image is spread along a line on the retina, theenergy density at any point is very much reduced and the sig-nal correspondingly less visible. If, however, the durationof the flash is extremely short, the image may be completelyformed in a single retinal location even though the eye ismoving, and the target's visibility might not be reduced.Mackworth and Kaplan (1962) found just such an effect with atarget illuminated by a condenser discharge light having aOlash duration of about one microsecond.

Flash durations as short as one microsecond cannot beobtained in navigation light systems, even with condenser-discharge light sources, but it is worthwhile to investigatethe possibility that during operational search very fastflashes of practicable durations may provide greater conspi-cuity than slower flashes.

The serious liritations on cockpit visibility suggestthat it may be worthwhile to consider the use of rear-viewmirrors or optical scanning devices in order to providesearch capability in otherwise blind directions (Howell,1958; Fisher, 1957). Optical aids of this type have notbeen found fully satisfactory, but even limited capabilitywould be of value, particularly when a pilot desires toassure himself that the air is clear in the direction towardwhich he wishes to maneuver (Applied Psychology Corporation,1962g).

- 65 -

IX. WULEE3-OF-THE-ROAD

Although regulations covering IFR and VFR flight alti-tudes offer a measure of protection against air collision,there are many situations in which two aircraft encountereach other at the same altitude, usually when one or bothare climbing, descending, or turning, or when two VFR air-craft are on headings within the same east or west half ofthe compass. In the latter case their paths may be inter-secting at any angle from nearly head-on to parallel.

If a pilot is confident that there is no danger ofcollision with a sighted aircraft, he need do nothing otherthan reaffirm occasionally that his judgment is still cor-rect. The occasions when a pilot must consider that acollision possibility cannot be ruled out may be groupedin two classes:

1. Those in which the fixity of bearing is or mightbe a major source of information;

2. Those in which the fixity-of-bearing criterion isnot applicable because one's own aircraft is maneuveringor the intruder is known or presumed to te maneuvering.

The second class of situations characterizes terminalareas, and at busy airports may involve several intruderaircraft at once. In such cases the pilot must anticipateaircraft movements and attempt well in advance to avoidany location or course which might lead to collision. Whencollisions or near-collisions occur under such c'rcumstances,there is usually a surprise element due either to failureto detect an intruder or failure to analyze visual cuescorrectly ora in time.

Collision situations in which the fixity-of-bearingcritericn is applicable occur generally under cruise con-ditions. If the intruder is sighted reasonably early, thereis time, if closing speeds are moderate, to evaluate thethreat and determine what action, if any, is needed. Inprinciple, it should be possible to devise rules-of-the-roadwhich, when fixity of bearing is observed, will insure thata reasonable avoidance action is taken, and that if bothaircraft maneuver the maneuvers will be complementary.

Such rules-of-the-road are of limited value when air-craft are maneuvering, and the complex nature of the situa-tion near airports requires a much greater degree of flexi-bility and improvization than any set of rules could provide.The question is what rules could best cover general cruisingconditions.

67 -

One such set of rules is incorporated in the CAR,parc 60, as follows:

* 1. When two aircraft are approaching head-on, orappeoximately so, each will alter its course to the right.

2. An overtaken aircraft has the right of way, andthe overtakirg aircraft will alter its course to the right.

3. When two aircraft are on crossing courses at ap-proximately the same altitude, tne aircraft which has theother on its right will give way so as to keep clear.

An aircraft which has the right of way may ordinarilymaintain its course and speed, but the pilot is not relievedof his final responsibility for taking action necessary toavert a collision. An aircraft obliged by the crossingrulesto "keep out of the way" is expected to avoid the otherwithout passing above, below, or ahead of him, unless pass-ing well clear.

These right-of-way rules, identical for day and nightflight, raise a number of problems. There is nothing toinsure that the "responsible" pilot (the one not having theright of way) will see the aircraft having the right of way,whose pilot expects to continue his heading and airspeed,at least until a more critical moment. At what angle does"approximately head-on" change to "crossing," where the ruleis different? And does "give way" mean to turn, climb, des-cend, or change speed? What happens when an overtaken air-craft decides to turn right, and so loses its right of way?

The rules-of-the-road in the CAR were taken from rules-of-the-road devised for marine navigation. There is somequestion whether these rules are adequate even for marineuse. There can be little doubt that the important differ-qices between air and sea navigation require different rulesfor the air (Calvert, 1961).

The time available after sighting for decision and man-euver in the air is very much less than on the sea. Wheretime-to-collision at sea may be measured in minutes, it ismeasured in seconds in the air. With so little time, andwith sighting failures not uncommon, every pilot shouldhave avoidance responsibility at all times, regardless of"right of way."

Ships at sea can control speed over a wide range, whileaircraft (except helicopters) have only a limited range ofcontrol and cannot drop below a minimum speed which itselfis many times higher than marine cruising speeds.

1168-

Ships at sea can maintain alert watches in all direc-tions, without difficulty, and generally can ass g-n to watchduty personnel who have this as their sole functlion. Pilotsmay have copilots to share observation duties, but limita-tions on cock-it visibility and the many other functionsrequired of pilots (and copilots) limit the amount and theeffectiveness of the attention that can be given to colli-sion avoidance.

On the other hand, aircraft operate in three dimensionsand so have the important possibility of altitude avoidance.Altitude maneuvers are often used by pilots to avoid colli-sion, but the present rules-of-the-road do not take thismaneuver into account.

To design rules-of-the-road unambiguous about pilotresponsibility, insuring complementary maneuvers, and pro-viding the most effective maneuver for all possible colli-sion situations, is by no means a simple task.

A complicated set of rules, based on a detailed analysisof sight bearing, closure speeds, and required magnitude aswell as direction of maneuver, was devised by de Vienna(1956). When a maneuver is indicated for either or bothaircraft, it is always a right-hand turn, followed by anequivalent left.-hand turn to get back on course.

Calvert ('1960) devised a set of avoidance maneuverswhich called for combined turn and altitude change maneuvers,to be used whenever fixity of bearing indicated a possibil-ity of collision.

One interesting set of rules for avoidance in a hori-zontal plane, is based on the principle that a pilot maneu-vers so that the sight line between himself and the intruder,If the intruder course Is not changed, rotates counterclock-wise. (The intruder drifts to the left in his field of view.)

This proposal has been analyzed by Hollingdale (1961)and Calvert (19061). Figure 5 is a course diagram illustratingthe manner in which this system operates. In general, bothaircraft are required to maneuver, and maneuvers consist oftur:ns, in specified amounts, and speed changes. As designed.,the system is principally applicable to marine navigation,and is particularly well suited to ships with navigationradar equipment. Although based on the simple requirement ofcounterclockwise rotation of the sight line, its applicationto air navigation is complex and difficult for some approachdirections. The system does not conflict with present rules-of-the-road in that the aircraft on the right passes ahead ofthe other aircraft; the head-on and overtaking rules would beessentially unchanged.

-69 -

B2 oI

\% /4% I

st p AN So

~sI\ I /

" ~/\

* I/A 2i and/

i I\

IL I

i,3 movement betwen.. ,- to and tI

1 2

Position of .t roft

/

/

A0

Fig. 5. Illustration of rule requiring counterclockwiserotation of sightline S, between aircraft. Subscripts referto given times during problem. Problems begin at time zero(t I when maneuvers are atarted. Time 2 (t 2 ) is time whencollision would have occurred if neither aircraft had man-",vered.

-TO

To sum up, while no set of rules-of-the-road has yetbeen devised that may be considered completely satisfactoryfor collision avoidance, it is evident that the presentrules do not take advantage of the possibility of altitudeavoidance, and in most cases they put the burden of res-ponsibility, at least initially, on only one of two air-craft involved.

-b71-

X. PROCEDURES FOR EVALUATING AIRCRAFT

NAVIGATION LIGHT SYSTEMS

From the time when position lights were first installedon aircraft until about 1940, little interest was displayedin critical examination of the adequacy of the system or inthe development of improved systems. Beginning about thattime, efforts were made to improve the basic system by vari-ous flashing arrangements, by adding lights to the basicposition lights, and by moderate increases in intensity.

During the 19508, interest in improving navigationlight systems increased greatly and was accompanied by theintroduction of manay proposals for new systems, some con-taining novel features differing radically from ex.stingsystems. There were no guidelines by which these proposalsmight be evaluated and no systematic procedures or facili-ties for dealing with them.

To facilitate experimentation by inventors and operatorswho were anxious to test some of the proposals, Special Regu-lations 361 and, later, 392, were issued, authorizing in-stallation of nonstandard systems, in limited numbers, forthe express purpose of collecting information from what mightbe called "user tests." However, the "experimentation" car-ried out under this program was haphazard. The reportedresults were fragmentary and generally consisted of littlemore than opinion polls and testimonials obtained under un-known and generally doubtful circumstances. Accordingly,at its last termination date, Jure 25, 1962, SR 392 was notextended.

Generally, the approach to improving navigation lightsystems has suffered from a preoccupation with "hardware."This has had two unfortunate results. Those proposals accom-panied by developed equipment received a great deal of atten-tion, but only in terms of the "package," without seriouseffort to separate engineering detail from the eosence of aproposal. On the other hand, proposals of considerable merit,not embodied in demonstrable hardware, were largely ignored.

By the middle of the 1950s, more systematic approacheswere undertaken, and there was some criticism of the inade-quacy of customary experimentation. Wright Air DevelopmentCenter contracted for a broad study of exterior lighting(Laufer, 1955), which resulted in an excellent account ofthe relation between collision geometry, required warningtime, light intensity, atmospheric condition, and visibility.

In 1957, the Douglas Aircraft Company issued a report in

.-m . , , ,

which factors involved in the evaluation of navigationlight systems were analyzed. In a subsequent report (Zied-man, 1957) these factors were considered further and therequirements for evaluation procedures were elaborated.

Also in 1957 the Bureau of Aerona'itics contracted fora systematic evaluation of several navigation light systems,which resulted in the development of objective test pro-cedures (Robinson, 1959; Fisher Noffsinger, & Robinson,1958). Projector (1958b; 1959a) analyzed various factorscharacterizing navigation light systems and stressed theneed for dealing with them separately.

In 1958 the Airways Modernization Board contracted fora study of the status of navigation light systems (Projector& Robinson, 1958). Sharply critical of the experimentalprocedures that had been in general use, this study suggestedvery strongly that evaluation procedures, to have value, mustbe systematic and objective. The present contract was anoutgrowth of this preliminary study. One of the tasks underthe contract called for the design of a systems evaluationprogram having as its goals (a) maximum reliability, (b)objectivity,,(c) minimum cost, and (d) maximum generality.

To achieve these goals, a large part of the effort hasbeen directed toward obtaining general information which canbe applied to the evaluation of any system. Another parthas been to develop experimental facilities at the NationalAviation Facilities Experimental Center (NAFEC), in coopera-tion with NAFEC personnel. These experimental facilitiesplus others available at NAFEC could beused in an evaluationprogram.

A considerable amount of basic information has been ac-cumulated, and has already served to provide criteria bywhich lighting systems have been evaluated. Such informationis by no means complete, and can never be complete because ofadvances in the state-of-the-art, new formulations of re-search questions requiring new information and so forth.Ongoing programs should continue to fill major gaps.

Rigidly detailed procedures for evaluating proposalscannot be established. Each proposal will require indivi-dUal handling. It is nevertheless possible to set forthgeneral guidelines which will suggest a rational, step-by-step program. In a given case, steps may be omitted, ormodified. The objective always should be to use as muchexisting information as is applicable, to avoid expensiveor difficult procedures when inexpensive, simple ones willserve, and to consider all aspects of the problem, includingthose (such as regulatory problems) that often escape theattention of inventors.

I7I

The first step in evaluation is a detailed analysisof the proposal in terms of the information it conveys,the light-coding technique used to convey it, and engineer-ing and economic considerations. These separate elementsare then evaluated in terms of information available. Itis often possible to reject a proposal on the basis ofthis analysis, if, for example, it attempts to present in-formation known to be without value, or if it employs alight-coding technique known to be subject to excessiveinterpretation error, or if its cost is inordinately high.If the system cannot survive such analysis, no furtherevaluation should be done on it.

However, it may be determined that some defects of thesystem as proposed are not basic. For example, engineeringdefects might be corrected, or a system conveying useful in-formation with a poor light-coding technique might be accept-able if a better method of coding were used. In such casesit may be desirable to proceed with evaluation of the ac-ceptable parts of the proposal, or to redesign it to make itmore acceptable.

If corroborative information or test data has been sub-mitted with the proposal, this should be critically examinedas part of the analysis. Test data, in particular, shouldbe evaluated for objectivity, adequacy of test design, andreliability of results.

If a proposal survives the analytic screening, it isthen subjected to experimental investigation. This may in-clude, as needed, photometric and engineering measurements,visibility tests, or psychological tests. As much as possible,essential aspects of the system should be examined separately.It is not possible to specify in advance precisely what testswill be needed, but generally laboratory, simulator, or fieldtests are to be preferred to flight tests. It is not alwayspossible to avoid flight tests, even for preliminary screen-ing. In any event, if a proposal survives preliminary screen-ing, it must ultimately be subjected to extensive flight testsbefore it can be considered as a new standard system underthe regulations.

At all stages in the investigation a prime considerationmust be that no system offering only minor or doubtful im-provement over the standard system passes the screening pro-cess. The results may be interesting and may add to the fundof information to be used in subsequent evaluations, butchanges in the regulations which would modify a standard sys-tem should not be undertaken except for substantial gain.

75 -

XI. NAVIGATION LIGHT SYSTEMS AND THE

CIVIL AIR REGULATIONS

Two apparently contradictory generalizations describethe present situation:

1. In several respects navigation light systems aremuch less effective than has been generally assumed.

2. Navigation light systems can be made much moreeffective than they now are.

The technical background on which these generaliza-tions are based has been summarized above. Technical as-pects of navigation light systems are inextricably linkedwith the CAR, and the effectiveness of light systems oftendepends critically on the regulations and how they are ad-ministered.

The Need for Standardization

The effectiveness with which navigation light systemsprovide communication between pilots depends largely on thespeed and accuracy with which their "language" can be read.

Before a pilot takes a new model of an aircraft offthe ground, he becomes thoroughly conversant with Its in-struments, controls, indicators, warning devices--its in-ternal communication system. He continues his trainingin flight and is tested before he is assumed fully compe-tent. Pilots require similar training in using navigationlight systems. They must learn to "read" the informationaccurately and quickly, and to make correct decisions basedthereon. A pilot sighting a signal constituting part of anavigation light system on an aircraft need not have beentrained in operating that aircraft, but he ought certainlyto have been trained in using the signals he sees. A pilotflies his own aircraft with the aid of its internal commun-ication system, but his navigation lights are part of acommunication system used by other pilots.

In these circumstances standardization is essential.With high-speed aircraft the time available for making de-cisions is often insufficient under the most favorable con-ditions. Under many marginal conditions any loss of timecaused by lack of standardization may decisively determinethe outcome.

Essentially, present navigation light systems, in addi-tion to indicating the aircraft's presence, are intended toIdentify azimuthal sectors around the aircraft. To provide

ss77

this relatively simple information, a pilot njay be con-fronted with a great variety of signals and combinationsof signals, including steady red, green, and white lights;flashing red, green, white, yellow, and bluish-white lights;alternately flashing red and white, yellow and white, greenand white, red and green lights; bluish-white lights flash-ing at any of three different frequencies.; lights flashedsequentially in rows; and many other arrays and combinations

of lights too numerous and complex to list here. The firsttask of a pilot confrontod with a light signal should be to"Wread" the Information. In the present situation he must

first, ".natead, identify the system to which it belongs.This is often difficult; it is sometimes impossible, becausethe saw signals appear in different systems with differentmeanings.

Lack of standardization has superimposed on the pilot'salready difficult task a superfluous task, characterized bythree possible degrees of difficulty:

1. At the very least, he must identify the system towhich the light signal belongs--sometimes easy, but in anyevent a task that should not be necessary.

2. The pilot is confused by the need to determine thesystem to which a signal belongs. He must know all possiblesystems in detail and be able to place the observed signalunerringly within the correct system. With some signalsthis is theoretically possible, but often the diversity ofsystems leads to delay, uncertainty, or error.

3. If there is a definite ambiguity, the pilot cannotresolve It unless and until he gets additional informationfrom the light system.

Lack of standardization may also make it more difficultto distinguish aircraft signals from lights on the ground,many of which are similar to those found on aircraft. Thisconfusion can be reduced by reducing the number of differentkinds of aircraft light signals.

The first step is to achieve standardization at theearliest practicable date--meaning that every aircraft shallhave all of the standard system and nothing else. This doesnot mean, necessarily, that every aircraft would have anidentical system; different detailed configurations may bepermitted (even called for) on different aircraft. It doesmean that each authorized configuration would be part of anover-all unambiguous standard, whereby a pilot sighting anaircraft light signal could interpret it wi'hout hesitationor uncertainty.

To -

Heretofore, navigation light systems have been judgedoutside the context of standardization. It has not beenappreciated that the virtues of any system, however greatwhen considered in isolation, are lost or seriously dilutedif such a system is one among several being flown simul-taneously. The one criterion by which any system oughtto be judged is its suitability as the standard system.

Before taking up the details of a program intended toachieve and maintain genuine standardization, it is usefulto consider some general aspects, particularly two technicalareas of special importance: the role of intensity as re-lated to the CAR, and the limitations of visual collisionavoidance with navigation light systems.

Problems and Procedures in Standardization

Once genuine standardization has been achieved, anysubsequent major change in regulations should require con-verting all aircraft to the new standard. Thus, it wouldbe ill-a•VTsed to change regulations frequently or for minorgains incommensurate with the cost of refitting. Promisingnew systems should be tested and proved with sufficientthoroughness that when one does become the standard, thereis reasonable assurance it can remain standard for a numberof years thereafter. It is impossible to guarantee thatgreatly improved systems may not be developed or discoveredvery soon after a standard is made effective, but it is en-tirely possible, through painstaking evaluation, to avoidhasty or inadequately supported changes.

There are special difficulties in requiring standardi-zation of small, low-powered aircraft, which often havepower plants too small for a complete standard system.While It is highly undesirable to grant exceptions for thisreason, it may nevertheless be unavoidable. In any event,the authorized sxception should be a version of the standard,with similar signals, free of ambiguity.

High-performance aircraft also present difficulties.Conventional locations for lights are often unavailable, orelse are subject to severe restrictions of weight, size, andenvironmental factors. At the high speeds of such aircraft,the limitations of visual collision avoidance are so greatthat is Is unreasonable to impose a difficult lightin. re-quirement on them. However, they land and take off at mod-erate speeds, often mixed in with conventional traffic. InVFR mixed traffic at moderate speeds, the lighting require-ments should be the same as for other aircraft. Recent engi-neering advances, including retractable anti-collision lightsand very small high-powered position lights, suggest thatstandard navigation light systems for high-performance air-

as 79

craft will be feasible (Godfrey, 1959; Grimes drawingo-8455o).

Aircraft with unusual flight characteristics, such ashelicopters, VTOL's and dirigibles, might require distinc-tive lighting systems. Any such systems should be standard-ized and should dovetail with the standard system on con-ventlonal aircraft.

Once standardization has been achieved, any contemplatechange must be examined for confusions or ambiguities thatmight arise during the transition from the existing to thenew standard. It has been the practice to allow relr ,ivelylong periods of time between promulgating a regulati.n andputting it into effect. This transition time probably cannobe avoided, especially with a major change. Ambiguities suclas exist today should be allowed only for overwhelmingly co-gent reasons. In any event, transition periods should bekept to a minimum. It may be possible to provide relativelylong delays before new regulatlons become effective, but toconfi•e the actual changeover time to a short period justbefore the effective date.

Standardizing air traffic internationally would requirecooperation of all national regulatory agencies through theInternational Civil Aviation Organization. This aspect ofthe problem has not been stressed in this report, becausethe diversity of systems within the borders of the UnitedStates is almost entirely domestic in origin. While thereis some diversity in foreign aircraft, virtually every sys-tem appearing on foreign aircraft also appears on Americanaircraft. Ultimately, when the United States is preparedto attempt genuine standardization, the effort should be co-ordinated with other countries through ICAO.

Intensity arid the CAR

It was shown In Chapter IV that the navigation lightintensities called for In the CAR are marginal. The mini-mum requirement of 100 candles for the anti-collision lightin the horizontal plane provides a visual .'ange of about2-1/2 miles when atmospheric conditions are minimum VFR(3-mile visibility). The miniomm Intensity of 5 candlesfor the wing position lights from 20 to 110 degrees outboardgives a visual range of only 1-1l miles, based on the "prac-tical" threshold, and must te considered submarginal. Sub-stantial increases in intensity for all navigation lightswould be of considerable value.

Since the CAR prescribe only minimm intensities, anaircraft operator may install equimnZ of much higher in-tensity without deviating in anr way from the requirements.

II l II I l I Iso

Nevertheless much of the effort toward developing improvedsystems has been directed toward higher-intensity systemsnot in accord with the regulations, In the mistaken beliefJ Thait this was the only way to obtain higher intensity. Asindicated earlier, any system can be designed for higherintensity if the deis--ner is willing to pay the price interms of increased power consumption. In achieving andmaintaining genuine standardization, it is unnecessaryand undesirable to depart from the standard in order toincrease intensity.

I CAR requirements for the intensity distribution ofposition lights at various angles from dead ahead are showni in Table 4. The higher requirement forward is based on thefact that collision courses in these directions, for anygiven aircraft, involve higher closing speeds than thosefrom other directions. Therefore to provide a given warningtime, sighting ranges must be greater and signal Intensitymust be correspondingly greater forward. Laufer (1955),analyzing the intensity requirements as related to collision-I course convergence angle and aircraft speeds, obtained re-sults which suggest that the abrupt drop in required inten-sity 20 degrees outboard is not Justified.

But the problem is far more complicated. Slow aircraftmay collide with fast aircraft. The required distributionof intensity around the slower aircraft, for equal warningI time, is much more uniform than for the fast aircraft. Inthe limiting case, a hovering helicopter for example, theintensity requirement should be uniform in all directions.I In the case of a fast aircraft in a collision situationwith a slow aircraft, no rearward intensity is requiredsince the slow aircraft cannot overtake the fast one. De-pending on the speed ratio, the zero-intensity zone mayS extend well into the forward hemisphere; the required in-tensity ratio under these circumstances Is infinite. Fi.nally, at very high aircraft speeds, visual collision avoid-I ance may be difficult or impossible, or the signal intensi-ties required may be far beyond practicability.

The Implications of the above for the CAR will betreated In detail later. Two general conclusions may bestated:

1 1. Intensities should be generally increased.

2. Position light Intensities, especially in the zonefrom 200 to 1100, should be increased substantially.

Limitations of Visual Collision Avoidance

A number of factors limit the usefulness of efforts to

! i i l l l l l l lall

Table 4

CAR Minimum Intensity Requirements for

Position Lights

Angle, Light Minimum intensity,left or riht candles

00 to 100 wing position,

left red, right green 40

100 to 200 30

200 to 1100 " 5

ii0 to 1800 tail, white 20

I I I i !

avoid collision by visual techniques. In desigring lightsystems and irn drafting regulations prescribing standardsystems, it is essential to understand precisely what asystem can and cannot do. In some cases, the limitationsare inherent and not subject to control. In other casesimprovements are possible, through more effective use ofexisting or readily available equipment and procedures.

It has been suggested that the limitations of visualavoidance are so great that it must be considered seriouslyinadequate or, often, not possible at all (Emerson et al.,1956; Calvert, 1958; Fiore, 1959). It is felt that suchpositions are In some respects extreme. The limitationsof visual avoidance are indeed serious, but to a considerableextent, improvement is feasible. In any event, visualmethods are still required and will be used for a long time;the problem is to maximize their efficacy and recognize anddefine the limitations.

Cockpit visibility. Most cockpits lack sufficient visl-bility. Pilots are "'lind" in many directions and can seein some other directions only with a degree of strain (Ed-wards & Howell, 1956; Fisher & Howell, 1957). Visibilityis particularly poor rearward and is often severely re-stricted vertically. In some cases climbing or descendingovertaking courses involve visibility angles blind to thepilots of both aircraft. It is not surprising thereforethat the p-eponderance of accidents involve overtaking orrear-approach situations in spite of the fact that closingspeeds are much slower for such courses (Fisher & Howell,1957). It is not suggested that these are the only reasonsfor the high incidence of overtaking collisions. Rules-of-the-road and traffic management result in relatively uni-directional traffic in high density locations, thus pro-viding many more opportunities for overtaking collisions.The combination of high opportunity rates with serious im-pediments to adequate vision probably account for the pre-ponderance of this class of collisions.

It is possible that rear-view mirrors or other opticaldevices may help to provide visibility in the otherwiseblind directions (Fisher, 1957; Howell, 1958). These areof limited usefulness and require extra attention on thepart of the crew member who uses them. Nevertheless sincemore time is generally available and required ranges aretherefore shorter for such collision approaches, even limitedadvantage gained from such devices may be helpful.

The quality of the glazing itself may have an importanteffect on visibility. Windshields of extremely low pitch,as on high performance aircraft, have low transmissions andconsequently seriously reduce visibility. At a pitch angle

20 degrees from horizontal, for example, the transmissionof double glazing is about 45% (Fiore, 1959).

Pilct attentiveness. Collision threats may occur inVFR hlight at any time and from any direction. Therefore,a pilot can use visual collision-avoidance techniques onlyto the extent that he is able and willing to pay constantattention. But the workload of pilots is so severe thatthey are often too distracted to be able to search regularlyand effectively. In the neighborhood of airports, wheretraffic density is heaviest and collision likelihood greatest,workloads are particularly heavy. Marshall and Fisher (1959),In an experiment on the daytime conspiculty of small aircraftin a terminal area, found that pilots often failed entirelyto notice small aircraft on a collision course. When theydid detect the intruder aircraft, they often failed to notethat it was on a collision course and thus took no evasiveaction. The pilot subjects in this test had been instructedto watch for and avoid local traffic, but were not aware thattheir ability to do so was the subject of interest in the test.While the small target aircraft were by no means very easy tosee--more often than not they were detected at short but rela-tively safe ranges and were avoided by effective maneuvers--the number of occasions when they were not detected or notrecognized as threats suggested very strongly that pilot dis-traction interferes seriousli with collision-avoidance dutiesor that many pilots need to allot more time and attention topossible collision threats.

The obverse of the problem presented by the heavy workload of pilots is that resulting from inattentiveness duringtimes when conditions are "too" favorable. It has been sug-gested that good weather, comfortable physical environment,and the monotony of long, uneventful cruising can seriouslydiminish a pilot's attentiveness to his tasks and his abilityto detect and respond correctly to unexpected and infrequentstimuli (Trumbull, 1962).

Measured in terms of the probability that any aircraftwill be involved in a collision at any time, the riak is ex-tremely low, even in high-density areas. But the consequencesof collisions are often disastrous. A pilot flying VFR ispresumed to be capable of avoiding collisions by visual tech-niques, and he is responsible for doing so. But to the extentthat a pilot does not watch alertly for other aircraft., hecannot be considered effectively under VFR, no matter what histechnical status.

The difficulties of collision avoidance, especially inthe presence of high-speed aircraft, require that the pilotsof both aircraft be responsible for avoidance. It can be oflittle aolace to a pilot involved in a collision that he had

the right of way. The entire concept cf right of way,derived from marine traffic rules, is antiquated even forthe environment for which it was formulated. In the com-plex, high-speed environment of air traffic, it can bedangerous. It is the obligation of ever pilot, whetherhe himself is flying VFR or under tra--T I control, toutilize his visual capability at all times that atmosphericconditions perwUtt.

Sector information. Azimuthal sector, or aspect, In-formation is coded nt-to all systems currently in use or pro-posed. Although it has not been specified in what way, ithas generally been taken for granted that pilots could usesuch information to determine whether a collision courseexists. Analysis has shown, however, that pilots can makesuch determinations only crudely, even if extremely precisesector information were coded into the system (Calvert, 1958;Applied Psychology Corporation, 1961a). Marshall and Fisher(1959) found that pilots were unable to use aspect informationto determine the existence of a collision course. Althoughthey speculated that this inability may have been due to thesmall size of their target aircraft, it seems much more likelythat aspect information is inadequate for the purpose.

Although sector information cannot be used for preciselydetermining collision hazard, it can serve to segregate air-craft into those with which it is impossible to collide andthose which will require further attention. And especiallyin terminal areas where pilots must keep track of a numberof aircraft, it can be helpful In anticipating their probablefuture locations.

Sector information also permitts segregation of collisionpossibilities into categories of urgency, but again verycrudely.

To facilitate these rough screenings, sector-coded aye-tems should be quadrantal, distinguishing left from rightand forward from rear aspects. The screening rules to beused with quadrantal sector coding are:

1. If to your right you see the intruder's right, orto your left, his left, no collision impends.

2. If to your rear you see his rear, no collisionimpends.

If the possibility of collision cannot be ruled out bythe above rules, then the urgency of the situation may beclassified as follows:

1. If in your forward hemisphere you see his forward

aspect, the situation Is likely to be urgent,

I 2. If to ycur rear you iee nis forward aspect, or Ifforward you see his rearward aspect, the situation Is likelyto be less urgent.

Obviously these Judgments can be refined by feeding intothe problem the precise bearing, relative to one's own air-craft, on which the intruder is sighted, and the knowledgeof one's own aircraft speed.

It may appear that categories could be further refinedby coding more finely divided sector information into thesystem. However, the results of the analysis show (a) it isdifficult to use precise sector Inflormuation, (b) codes givingmore information grow increasingly complex, (c) other informa-tion better suited to the purpose Is available$ and (d) otherkinds of coded information might be more useful. These con-siderations suggest that quadrantal information is optimalfor sector coding.

Use of the fixity-of-bearing criterion. Limitationson the-use or the rixity-or-Dearing criterion were discussedin Chapter II. They are:

1. The criterion is valid only %hen both aircraft areon straight-line, constant-speed courses. A pilot knowswhether his own course meets this requirement but often isuncertain about the intruder's course.

2. There is no experimental data on the ability ofpilots to detect movement of the sight bearing. Laboratoryexperiments show that observers can detect relatively smallsignal movement, of the order of I minute of arc per second(Leib:witz, 1955). In an operational situation, detectionthrestiolds are undoubtedly very much higher, principally dueto tbe Instability of the piloL observer's frame of reference.Calvect (1958) has estimated thresholds to be about 0.4 de-gree of arc per second, but this must be verified experi-mentally. The operative threshold profoundly affects theability of a pilot to use the fixity criterion successfully(Applied !',jc nology Corporation, 1961a).

3. Because of uncertainties due to the difficulty ofprecise determination of fixity, the fixity-of-bearing cri-terion Is probably of limited value at longer ranges or athigh closing speeds.

4. Flashing signals adversely affect the precise de-termination of the existence of fixity.

In spite of the above limitations, the fixity-of-bearing

criterion is one of the principal avoidance techniquesavailable with navigation light systems, and pilots shouldmake full use of it, while at the same time recognizing itslimitations.

High-speed aircraft. High-speed aircraft impose seriouslimitations on visual collision avoidance. At high speeds,aircraft are slow to respond to maneuver control, and themarmvezm themselves are slow relative to the speeds. Formany collision convergence angles, closing speeds are sohigh, even when one aircraft is much'slower than the other,that the possibility of successful visual avoidance islimited. If both aircraft are high speed, closing speedsmay be so high that the small possibility disappears alto-gether, and the outcome of a potential collision situationmay be beyond the pilot's control. The closing speed of twoaircraft on a collision course depends so critically on theconvergence angle that visual collision avoidance cannot besaid categorically to be impossible with high-speed aircraft,although it may be impossible under certain conditions.

A number of factors operate in favor of visual collisionavoidance, even for high-speed aircraft. If certain alti-tudes are reserved for traffic in one general direction, therange of possible closing speeds is held down to more manage-able levels. (On the other hand, overtaking approaches areparticularly difficult from the point of view of cockpitvisibility.) At lower altitudes, especially in terminalareas, high-speed aircraft are generally operated at rela-tively low speeds. A pilot usually knows little or nothingabout other aircraft in his vicinity, but his knowledge ofhis own aircraft may serve to lessen his problem. If hisis a very high-speed aircraft he may sometimes be assuredthat no aircraft will overtake him, but that he could easilyovertake others. If his own aircraft is very slow, he mustexpect trouble from any direction.

The limits of VFR. Part 60 of the CAR, Section 60.30,gives vM, minimum weather conditions as shown in Table 5.Flight visibility is defined in Part 60 as "the averagehorizontal distance that prominent objects may be seen fromthe cockpit." In the recently issued Federal Aviation Regu-lations, Part 1 (1962), this definition was revised to "theaverage forward horizontal distance, from the cockpit of anaircraft in flight, at which nrominent unlighted objects maybe seen and identified by day and prominent lighted objectsmay be seen and identified by night."

The revised definition provides for nighttime observa-tions of lights, but the original operational difficultiesof the daytime definition are substantially unaffected. Also,the addition intended to cover night operations is exceedingly

i67.

Table 5

Minimum Flight Visibilities

Area Flight Visibility

Control zone 3 miles

Control area and transitionarea 3 miles

Continental control area 5 miles

Outside controlled airspace 1 mile

vague. It has been shown that the actual visual ranges ofaircraft in the daytime are less than the "flight visibility,"and often very much less (Applied Psychology Corporation,1961b). The designation of 'prominent lighted objects" astargets for observation of flight visibility at night isinexact, since the principal determinants of visual rangeare the intensity of the lights and the background againstwhich they are seen; prominence of the object is irrelevant.

Even if the definition were precise, it is difficultto assess atmospheric conditions to determine their suit-ability for VFR operation. Atmospheric conditions varyconsiderably from time to time and from place to place.Ground observations are of particularly limited use in de-termining flight visibility (Douglas, 1953). Direct esti-mate of flight visibility by pilots in flight is difficultbecause of the genee'al lack of "prominent objects" at knowndistances. At night, it is even more difficult.

It is possible to train pilots to improve their esti-mates of flight visibility from the meager cues sometimesavailable, including the amount of backscatter from theirown lights, but at best such estimates must be consideredcrude.

A note in the CAR Manual, part 60, sect. 60.30, sug-gests that the prescribed minimums be treated ý.onserva ivelyby pilots: "Good operating practice requires that regular orcontinued flight in near-minimum weather conditions be a-voided." It may be desirable to call for even more conserva-tism. Figures 3 and 4 in Chapter V suggest that when theequivalent daytime visibility is 3 miles, a 100-candlelight signal will be visible Just under 3 miles at night.But for an equivalent daytime visibility of 5 miles (theminimum for the continental control area) the visual rangeof the light will be less than 4 miles. Furthermore, ifthe background is relatively bright (moonlit sky or dusk)the visual range will be appreciably reduced. However, ifsignal intensity is raised to 1000 candles, then the visualrange of the lights will correspond well with daylight visi-bility, even if the observing conditions are not ideal.

The limitations of visual collision avoidance suggestthe advisability of raising VFR minimums above their presentlevels, in order to insure longer sighting ranges and there-fore more decision time. It may also be advisable to specifymore particularly the limiting conditions of VFR, not onlyas related to atmospheric conditions but also to the speed ofthe aircraft, the range of possible closing speeds as deter-mined by air traffic rules, and the intensity of the naviga-tion lights carried. While this may seem excessively complex,the limitations of collision avoidance are correspondingly

Slill II-8g..

complex; the alternative is extreme conservatism, elimi.nating a large range of relatively safe conditions in orderto eliminate rare risks, or else exposure of aircraft tooccasional situations in which visual collision avoidanceis not possible.

A convenient and more conservative demarcation betweenVPR and non-VPR atmospheres might be the boundary between"light haze" and "clear" (see Pig. 4, Chapter V). At thisboundary the transmissivity is 0.53/mile and the daytimevisual range is about 6 miles. The visual range of a 1000.candle signal light is slightly greater than 6 miles. ifthe assumptions about the fixity-of-bearing criterion madeIn Applied Psychology Corporation Technical ieport No. 1(1961a) (that is, threshold detection angle a 20 and escapeacceleration w 1/4 g) this sighting range is probably ade-quate for lower ranges of probable closing speeds, not inexcess of about 400 or 500 miles an hour, but even this mustbe considered marginal. It Is likely to be practicable underfavorable conditions, If head-on convergences can be elimi-nated by traffic rules.

0I -

XII. rOSSIBLE CHANGES IN THE CIVIL AIR REGULATIONS

The following discussion assumes that effective andgenuine standardization will be the first objective of anychange in the CAR. The formidable problems involved oftenmake it difficult, even impossible to effect changes of un-questioned value. These problems will be considered andchanges in the rules which can achieve fundamental objectiveswith minimum Impact on the aviation community will be out-lined.

Three phases are considered:

Phase I---Regulatory action to achieve genuine standard-ization under preseiit conditions.

Phase II--Improvements on the standard system achievedin Phase I, calculated to obtain optimum performance withinthe basic intent of the design.

Phase III--More or less radical major or long-range im-provements in the standard system, having substantial impact.This long-range program should lead to regulatory changesonly if the improvements are proved in careful, extendedevaluation, and are commensurate with the uost and otherconsiderations.

Phase I should be instituted with minimum delay. Thetimetable for Phases II and III should be flexible. PhaseII, depending on considerations discussed below and un thetime required to effect Phase I, may be combined with PhaseI in order to avoid a change in the regulations too soonafter the completion of Phase I.

Phase I: Regulatory Action to AchieveGenuine Standardization

All navigation light systems being flown today providesector-coded information. To achieve genuine standardizationwithout scrapping all present systems, it is necessary toexamine each for adequacy and to consider which (or whichelements) would constitute the basis for a regulated stand-ard system meeting the following criteria: (a) providesadequate sector information, (b) requires little or no de-velopment involving extended evaluation of doubtful outcome,(c) has minimum impact on the aviation community, (d) interms of available equipment, sacrifices no important attri-butes such as intensity or reliability, (e) permits smoothestand fastest transition.

-91

The system that comes closest zo satisfying these re-quirements is apparently the one prescribed by the presentCAR: steady-bu'ning red, green, and white position lightsand one or more flashing red anti-collision lights.

Anti-collision lights are ovnidirectional: equallyvisible In all azimuthal directions. The angular coverageof the position lights Is as follows:

red wingtip: 00 to 1100 leftgreen wingtip: 00 to 1100 rightwhite tail: 1100 left around the rear to 1100 right

This 3-sector system does not provide optimal quadrantalsector coding, since the white light in the rear does not dis-tingulsh left from right. Also, the sector boundaries abeamare at 1100 instead of 900.

However, no system now in use provides better sectorinformation. Some use additional lights, as on the fuselage,providing an array that may enable pilots to interpret sec-toring more precisely or even to distinguish left from rightto the rear, but array codes, as noted earlier, we confusingand unreliable.

Since the system has been in use for several years,many, perhaps most, aircraft already have it; thus, no de-velopment is needed. For similar reasons, the i.act onthe aviation community would be less than for any acceptablealternative. Large numbers of aircraft already fly the sys-tem; many others would require only minor changes, such asdisconnecting excess lights or bypassing flashers.

Two major groups of aircraft might require changes ofsome consequence: the small number carrying nonconformingequipment which would have to be replaced entirely; andthe aircraft having position lights but no anti-collisionlights, this latter group including many small, privateaircraft. The cost of one or In some cases two anti-colli-sion lights might seem large to these owners. More impor-tant, the limited power capacity of these small aircraftwould prohibit adding more electrical equipment. For genu-ine hardship cases solutions might have to be worked out,but no acceptable alternative would involve less hardship.

Table 6 compares approximate costs of some of the sys-teus and components of interest (only the equipment cost Isgiven; installation costs van• considerably).

As for the requirement that no important attributes ofcurrent systems be sacrificed (in other words, that thestandard not represent a step backward), the prese:st system

_92-

Table 6

Approximate Costs of Navigation

Light System Equipment

Proposed Standard System

j Position lights (three required) $ 15 each

Anti-collision light (one or two $ 100 eachj required)

Cost of complete system $ 150-250

I United Air Lines*

Oscillating position lights (threerequired) $ 375 each

Cost of c~mplete system $ 1100

I Honeywell-AtkinsOver

Two wingtip units $ 2000(Exact costs are difficult to esti-mate. Designs and prices havevaried considerably.This estimate is based on a May 1961quotation of ?2350).

I Since this report was prepared, a version of the UALsystem has been developed for small aircraft. It Is under-I stood that this lighting system will be in a cost categorycomparable to that of the proposed standard system.

- 93 -

Is an acceptable minimum. An alternative that appears tosatisfy most of the requirements consists of position lightswithout anti-collision lights; this is simpler and cheaperthan the proposed system, and is already Installed on aneven larger number of aircraft, but is unacceptable forthe following reasons:

1. It would substantially reduce intensity when pres-ent minimums appear to be too low. Anti-collision lightsare generally more intense than position lights in all direc-tions, and considerably more intense in most.

2. If the position lights were burned steadily, thenthe conspicuity advantage of flashing lights would be lost.On the other hand, if they were flashed, there would be nosteady lights, reducing the effectiveness of the fixity-of-bearing criterion. In the proposed system, the flashinganti-collision light provides the most conspicuous warningof aircraft presence at long range, and the steady positionlights are useful as targets for observing bearing fixityat shorter range.

3. The simple position-light system could conceivablybe modified to meet the above objectives by using higher-intensity fixtures and pulsating the lights, rather thanflashing them with a completely dark ofd period. By thisarrangement, they would appear to flash at long range butwould pulsate at shorter range without going out entirely.But such modification would require replacing equipment onalmost all aircraft, thereby imposing a heavy burden onvirtually the entire aviation community.

The present system meets very well the fifth require-ment, enabling a fast, smooth transition. Since burdensomechanges are required In relatively few cases, the transitionperiod could be very short. The time required to draft andcirculate regulatory proposals, to prepare, submit, andanalyze comments, and to hold conferences need not be pro-longed by considerations of excessive, burdensome change-overs by the entire aviation community.

Because all pilots are already familiar with the system,no retraining is involved.

It has been suggested that anti-collision lights shouldbe white rather than red. Simple and inexpensive replacementof the red cover glass with a clear cover would yield a whitesignal with 4 to 5 times the intensity of the red signal. Onthe face of it, the suggestion appears attractive. It is notrecommended at present, however, for the following reasons:

1. Ant.ý-colllslon lights are mounted behind the cockpit

and project their beams forward directly in front of thepilot. Backscatter, plus occasional direct Illumination ofaircraft structures in the pilot's view, can be a seriousproblem. While the effect of backscatter on pilots is notfully understood, it has been found that red backscatter issubstantially less bothersome than white, especially in mar-ginal atmospheric conditions. This is explained partly bythe much lower sensitivity to red light In the periphery ofthe eye. While this means that an observing pilot will seethe white signal better than the red in the periphery ofhis eyes, it has been found that the visibility of red sig-nals in the center of the eye is significantly better thanindicated by photometric measurements (on which the 4 to 5times increase in intensity of white over red is based)(Middleton & Gottfried, 1957; Mullis & Projector, 1958).Furthermore, it Is seriously questioned that signals aresighted In the periphery of the eye under operational condi-tions (see Chapter IV). These effects are far too complexand insufficiently well understood to warrant a change towhite at this time.

2. The higher intensities being recomended wouldmagnify the deleterious side effects mentioned above.

3. Because It is simple and inexpensive, the change towhite could be made at any future time; thus there is noreason to consider it until its desirability is well esta-blished.

To require general increases in intensity as part ofPhase I would mean replacing the navigation lights on vir-tually every aircraft. Also, serious problems of installa-tion, wiring, and power might Jeopardize the standardizationprogram, and would certainly delay it. Standardization isso greatly needed, and its accomplishment so relativelyeasy, that it is felt desirable to proceed immediately.However, the need for higher intensity need not be ignored;it can be strongly recommended without being required.

While the CAR prescribe minimum intensities, higherIntensities with the standard configuration would In no wayviolate the regulations, nor depart rrom genuine standardi-zation. Equipment providing substantially higher intensi-ties Is already available or in an advanced stage of develop-ment. Some high intensity equipment not In accord with theproposed standard is already in use. In changing to the newstandard, there is every reason why operators of aircraftwith nonstandard high intensity lights should use equip.r-twhich will result in no loss of intensity.

In order to facilitate immediate standardization, itmay be inexpedient to require increased intensity at this

- 95 -

time. However, it may be feasible to require all new equip-

went to meet higher Intensity requirements, especiaT~y fortransport and other large aircraft. The option of ope.iatorsto equip their aircraft with higher-intensity lights in theinterest of increased safety already exists and would re-main available after standardization.

The cost figures quoted above are for units conformingto present intensity requirements. Obviously, higher-inten-sity versions of this equipment will cost more, but thereis no reason to believe their cost would be as great asthat of other presently available high-intensity equipment.

Phase II: Regulatory Action To Improve ThePhase I Standard System"

The objective of Phase I was narrowly limited to theessential requirement of achieving genuine standardizationat the es-liest practicable date. Because it is felt thatthis could be done without the necessity of developing newequipment, of extended evaluation, or of serious impact ordelay, it has been felt desirable to consider imyrovementsin the Phase I system as part of a second phase. Threepossible regulatory actions are considered: (a) dividingthe symmetrical white tail sector into two arts, in orderto distinguish left and right to the rear; fb) increasingminimum-intensity requirements; and (c) changing trafficrules and minimum conditions for VFR in order to increaseavailable warning time in possible collision situations.

Convertin the Standard 3-Sector to a 4-Sector QuadrantalSystem "-

The major defect of the present 3-sector system is thatit does not distinguish left from right in the rear. Also,while the two forward sectors are 1100, a quadrantal system

1 It is not meant thereby to rule out the possibility thatpart or all of Phase II might be combined with Phase I in asingle regulatory action. Only the regulatory agency itselfcan decide questions like these, and many factors other thanthose considered here may underlie decisions. Furthermore.the basis for decisions changes in time, depending on thestate-of-the-art, the availability of information, the re-ceptiveness of the aviation community, etc. Thus Phases Iand II are separated here because they do seem to involvedifferent elements or elements of differing importance. Atthe time of making decisions about regulatory action, thesefactors may be weighed differently than they are now.

with sharply distinguished 900 sectors is considered pre-ferable. It is important, therefore, to change the presentsystem to make it more nearly quadrantal. If has been shown,however, that a sector-coded system is not and cannot be aprecision device In visual collision avoidance. Its majorfunction is to enable a pilot to avoid potentially dangeroussituations, Once he is in such a situation, sector codingIs of little or no use in determining the existence of acollision course or deciding what avoidance maneuvers willbe am t effective. In short, while the usefulness of quad-rantal sector coding makes it worthwhile, the limitationson its usefulness make It desirable to pay special attentionto the cost and impact of methods of obtaining it.

The essential change proposed is replacing the 140-degree white taillight with a light or lights covering two' 90-degree quadrants to the rear, the left yellow and theright bluish white. To minimize the impact, it is proposednot to require a change in wingtip position lights untilthey are being changed for other reasons (for example, toincrease intensity), at whicb time their coverage could simul-taneously be changed to 90 degrees quadrantal.

Substituting bluish white and yellow for the white sig-nal raises problems, the most important of which is the speci-fication limits for these colors.

j In all signal color systems, including the 3-color sys-tem of the CAR, an effort is made to permit as much tolerancefor the colors as is consonant with easy and accurate identi-fication. There are two major reasons for this: it is dif-ficult to manufacture colored glassware to narrow tolerances,and it is desirable to obtain glassware of as high a trans-mission as possible. For many colors, notably red and green,high color purity is obtained at the expense of very lowtransmission. In view of the need for high-intensity signals,the requirement for high transmission makes it especiallydesirable to determine tolerances carefully.

For the conventional 3-color system acceptable toler-T ances, worked out over the years, are described in FederalStandard No. 3 as the "Aviation" series of colors. Trans-missions for good red and green filters in this series areof the order of 20-25%. White, of course, is obtained withclear glassware. This system is considered acceptable as a3-color system; it Is not intended that Aviation Red andAviation Green be used in conjunction with more than oneother color intervening In the range of yellow to bluTlMwhite. Another system is described which Is suitable as afour-color system. This system, the "Identification" series,includes red, green, yellow, and "lunar white" (a bluishwhite), but calls for much purer reds and greens to make

97--

room fir yellows and bluish whites of fairly wide toler-ances. Identification Red and Green glasses have trans-missions approximately half the transmissions of AviationRed and Green.

It would be possible to change the glass covers ofAviation Red and Green wingtip position lights to Identi-fication Red and Green. This would requtre changing everypiece of equipment now in use, but the cost would be rela-tively small per unit. More Important, however, would bethe reduced intensity, a penalty that would be paid withexisting equipment as well as new higher-intenslty equipment.

If the red and green in the proposed system are in theAviation series, then there is no existing specification foryellow and bluish white suitable for making up a reliable

r-color system. It is felt that satisfactory tolerancescould be specified for these two colors. These toleranceswould be appreciably narrower than existing ones, but ifcarefully determined would be broad enough for reasonablemanufacturing. Fortunately, both the yellow and bluishwhite color limits could be met with glassware of relativelyhigh transmission, of the order of 30-50%.

To sum up, it is felt practicable to add a yellow anda bluish white to the existing red and green Aviation colorsto make a 4-color sector-coded system, but color tolerancesfor these new colors would be different from existing colortolerances for any similar colors. Careful experimentationwould be required to determine specification limits forglassware in the new colors to insure a proper balance be-tween adequate color distinctiveness and manufacturing capa-bility.

Another problem with the four-color system is ambiguityin the overlap regions between sectors. (This problem existsat present in the 3-sector system, but would be more seriousto the rear in the proposed quadrantal system.)

The overlap dead ahead is not serious. The location ofthe wingtip position lights at the extremities of the wingsmeans that at almost any distance of observation the twosignals, red and green, are seen as separate signals. Notonly is there no merging, but the visibility of the twoseparate signals at and near dead ahead identifies the aspectof the sighted aircraft quite precisely.

Blue and purple are other available signal colors, butare not considered suitable for long-range signaling systemslike those under consideration here.

II II II I II "I"• I II

Abeam, there is :reater possibility of confusion, de-pending on the relative location of the wingtip and tallon any given aircraft and on the distance from the observer.If we take one minute of angle as the minimum separationat which two light signals can be identified as separate(Moon, 1936, p. 423), then if, in a given sighting direc-tion, the lateral separation of a wing and tail light is10 feet, they will appear as a single light source beyondabout 7 miles. In an operational situation, It is doubtfulif the eye can resolve two sources as close as one minute ofangle, and it is quite unlikely that their separate colorscan be Identified. But lateral separation Is often appre-ciably greater than 10 feet, and 7 miles, in terms of the

j usual visual range with present equipment under marginal. conditions, is a considerable distance. It is true that

in the clearest weather or with high-intensity lights longsighting ranges can be obtained, but it is likely that under

* most circumstances wing and tall lights will be separable.

When merged, two position light signals are seen as onej because the color of the signal is the additive combination

of the separate colors. In the case of the present 3-colorsystem, the overlap zone at 1100 on the left would appear toshade from red to white through orange and yellow or pink.On the right side, the overlap would shade from green towhite through greenish white. In the proposed four-colorsystem the overlap zone on the left would shade from yellowthrough orange to red. There would be no doubt of its beingthe left side. If the present wingtip coverage Is retaieed,the overlap zone would extend 20 degrees from 90' to 110(plus the cutoff regions). If the two signals are separable,then the observer can tell that he is in the 20-degree over-lap zone. If the wingtip units are replaced with 90-degree"quadrantal units, then the overlap zones are much smaller,as they are at other zone boundaries, their size dependingon the sharpness and precision of the cutoffs of the lightunits involved. The boundary relationships on the rightside would be similar except that the color shading wouldrange from bluish white through greenish white and pale greento green.

1 The merging of light signals Is a problem more fre-

quently between an anti-collision light and any positionlight. The clarity of separate signals is always subjectto degradation when merging occurs, but the flashing charac-teristic of the anti-collision light offers some help inproviding a distinction against a steady-burning positionlight.

The most troublesome boundary would be that at therear. Here the light signals originate from a single unitor from two units very close together. Merging would occurat all meaningful observation distances and the colors wouldrange from yellow through white to bluish white. To mini-mize ambiguity, the overlap zones would have to be as smallas possible. Ingenious but somewhat complicated techniquesfor resolving the overlap zone ambiguity in the rear havebeen proposed, but would have to be evaluated carefully be-fore being considered as ready for application. In anyevent an ambiguous overlap zone of as little as one or twodegrees to the rear is a considerable improvement over thepresent ambiguous coverage of 140 degrees.

The ability of pilots with second- or third-class medi-cal certificates to distinguish the four colors of the pro-posed system will require Investigation. Holders of thesecertificates are now required to distinguish Aviation Red,Oreen, and White. It may be that a few pilots with colorvision defects sufficiently minor to enable them to pass thepresent requirement would have difficulty with the 4-colorsystem.

Transition to a 4-sector system would not present seri-ous problems. The forward sectors would remain essentiallyunchanged. There would be ambiguity between the white ofthe present system and the bluish white and yellow of thenew system, and a pilot would have to consider his abilityto distinguish among these three colors as unreliable. Oc-casionally, especially If he-could observe two or three ofthe colors simultaneously or in rapid succession, his capac-ity to identify an observed signal color might be substantiallybetter than normal, but conservative practice would requirethat he consider the distinction unreliable until he can besure that the transition is complete and that his choice liesonly between yellow and bluish white. During the transition,therefore, quadrantal rear sectors would be of uncertainvalue--but the pilot would be no worse off than he is now.

Increasinr the Minimum-Intensity Requirements

Since most presently available equipment meets but doesnot substantially excaed present requirements, any significantincrease In requirements would mean replacing virtually allpresent equipment.

If the Phase I recommendation of higher-intensity equip-ment has been accepted, the impact of a mandatory requirementwould be lessened. It Is unlikely, however, that many air-craft would be equipped with higher-intensity lights as theresult of a recommendation. In view of the hardship involved,regulations might recognize four categories for transition:

- -

1. On new aircraft, the higher requirements could beimposed early and completely.

2. If equipment is being replaced for other reasons(for example, replacing the taillight with a two-color light),the intensity requirement could be raised at the same time.

3. Aircraft might be categorized by class of operation,for example, transport and commercial.

4&. For small private aircraft the expense of replace-ment might be compounded by inadequate power. For such air-craft it might be necessary to impose lses stringent require-

tments.

The present CAR call for minimum intensities for positionlights in the horizontal plane (azimuth), as shown in Table 7.Above and below horizontal, the minimums shown in Table 8 arerequired. To provide reasonable cutoffs in overlap zones,maximum allowable intensities in zones cutaide the intendedcoverage zone for a given position are specified, as shownin Table 9.

Anti-collision lights are required to nave uniform in-tensity at all azimuthal angles, with minimums for variousangles of elevation as shown in Table 10.

Because of the complexity of collision situations andthe extraordinary gamut of conditions that may be encountered,it Is difficult to determine precise intensity requirementsf for an ideal navigation light system. Also, there are prac-tical limitations on the intensities that can be requiredby regulations on all aircraft.

In drafting regulations it is necessary to considermany factors and to reconcile conflicting requirements.Therefore the higher intensities suggested below are In-tended as guidelines, setting forth objectives reasonablein terms of the present state-of-the-art and engineeringand economic considerations. It may prove necessary orappropriate to reduce the size of the suggested increasesor to set up different requirements for different classesof aircraft. It is also worth repeating that the require-ments are minimums. Where possible, these requirements mayand should -exceeded at the option of the aircraft operator,especially if the recommended increases cenmot be imposed byregulations. A further discussion on thic point, with speci-

S fic recommendations, is contained in the next parts of thischapter.

Position light intensitl requirements. The present re-quiremients ror rorward posit;ion lighns, as shown in Table 7,

ij Table 7

CAR Mininum Position Light Intensity

in the Horizontal Plane

Direc tion Light Intensity, candles

00 (dead ahead)to 100 Left red wingtip ndright green wingtip 40

100 to 200 30

200 to 1100 5

110° to 1800 White tall 20

- 102 -

Table 8

CAR Minimum Position Light Intensities

Above and Below Horizontal

Angle above orbelow horizontal Intensitya

0o 1.00 I

00 to 50 .90 150 to 100 .0 1

100 to 150 .70 I15' to 200 .50 1200 to 30° .30 I300 to 400 .10 I400 to 900 .05 I

a "I" is the minimum intensity in the horizontalplane specified in Table 7.

s05 -

Table 9

CAR Maximum Position Light Intensities

in Overlap Zones

Maximum Intensity, CandlesBoundary 100 to 200 beyond beyond

boundary 200

0°, red-green 10 1

llO°, red-white on left orgreen-white on right 5 1

00 , red-greena 10% of peak 2.5% of peak

a Paragraph 4b.634.1, of CAR 4b gives maximum percentages

of peak inten3ity for forward position lights in overlap zoneswhen peak intensities are higher than 100 candles.

-lot,

Table 10

CAR Minimum Intensity for

Anti-Collision Lightsa

"Angle above or below Effective Intensity,horizontal plane Candles

0 0 to 50 100

50 to 100 60

* 100 to 200 20

3200 to 300 10

a Because of the difficulty of obtaining completeazimuthal coverame in aome installations, due to ob-struction by empennage, obstruction is allowed up to amaximum of .03 steradians of solid angle within a solidangle of 0.15 steradians centeted about the longitudinalaxis in the rearward direction. However, for aircraftcovered by parts 3, 6, and 7 of the regulations, obstruc-"tion up to 0.5 steradian in any direction is permitted.

aili ~~UI iil ili]f

range from 40 candles near 0 degrees (dead ahead) to 5 candlesfrom 20 to 110 deerees. The large difference is based on thefact that closing speeds at or near 180-degree (head-on)approaches are mroch higher than those at smaller convergenceangles. If it Is desired that warning times be equal for allprobable collision courses, then the visual range of thelights must be proportional to closing speed. To achievethis it Is generally necessary to provide much higher in-tensitles forward than at angles more abeam. The extraordi-nary range of possible aircraft speeds and the variability ofatmospheric transmissivity within VFR limits make it impossibleto specify a distribution of intensity that precisely accom-plishes the objective. Rough calculations on some selectedsample cases are Instructive.

Table 11 gives relevant data for two aircraft, eachtraveling at 300 miles per hour, for various angles of con-vergence, based on a required warning time of 30 seconds andan atmospheric transmissivity of 0.53/mile (the boundary be-tween "clear" and "light haze," equivalent to a reported day-light visibility ol' about 6 miles). The intensity data aretaken from Fig. 4 of Chapter V.

If this table is compared with Table 7, it is evidentthat:

1. The CAR minimums are much lower than necessar7 tomeet the requirements on which the table is based through allangles of convergence from 180 degrees (head-on) to 60 de-grees (aspect angles from 0 to 60 degrees).

2. The abruptness of the reduction in the intensityrequired by the CAR at 20 degrees aspect angle does notcorrespond with the much less abrupt changes in intensitynear 20 degrees, shown In the table.

3. The range of intensities in the table covers amuch wider gamut than the range in the CAR.

But the assumptions on which the table are based arevery narrow. Similar tables covering all possible valuesof the parameters would be encyclopedic. It is useful hereto consider qualitatively how changes in the assumptionswould change the results.

1. If the atmospheric transmissivity were less than0.53/mile, for example 0,25/mile, equivalent to a daytimevisibility of about 3 mileu, all the required intensitieswould be increase. At the lower convergence angles, theincreases would be substantial, for example, from 13 tonearly 100 candles at 40 degrees convergence angle. Nearhead-on, the increases would be enormous. At 180 degrees

| 1 | S 06|

0

0 o 0 0 0 0 1 0 M090 (D

Do

-4~~ aw o

0 0 0) 0' 0~ 0 owaQC0 0 0. 0 000 0~'

r cCt

A) Is0 0 0 0 0 0

R Ph

P& m c 0

141 0 'At 01e L&I 0 0

0 L

'S 0

H1 0

v0-1

ceI

Ithe required Intensity would Jump from 300 to nearly10,000 candles. I

2. If the atmospheric transmisslvity were greaterthan 0.53/mile, the required Intensities would be lower,significantly so near head-on. At 0.73/mile, the boundarybetween *clear* and "very clear," the requirement of 300candles for the head-on ease would be reduced by nearly jhalf.

3. If the assumed speed of 300 miles per hour wereiwreased, substantial increases in intensity would be re- Iquixed, again especially near head-on. If for example thespeeds were 50% greater (450 miles per hour) the intensityrequired for the head-on case would be about ten times great-er, or about 3000 candles.

4. If the assumed speeds were reduced, the requiredintensities would be substantially lower. At speeds of 200 Imiles per hour, the intensity required dead ahead would be40 candles. j

5. If the speeds of the two aircraft were unequal, thenfor a given convergence angle the aspect angle for the sloweraircraft would be greater, and the aspect angle for the fast-er aircraft would be less than in the case of equal speeds.Therefore, the variation of required intensity with aspectangle would be greater for the fast aircraft than for treslow one. However, the maximum value, at 0 degrees aspect [angle for the head-on collision, would be the same for bothaircraft. This does not imply that intensity requireme'n--are similar for fast and slow aircraft; If both aircraft arehigh speed, then the possible closing speedBE-nd therefore,the required intensities) include higher values than If onlyone of the aircraft is high speed, and the corresponding re-quired Intensities are much higher.

6. If the assumed warning time of 30 seconds is changedby some ratio, this has the same effect as a change in air-craft speeds in that ratio. Thus increasing the requiredwarning time to 45 seconds has the same effect as Increasingspeeds from 300 to 450 miles poer hour, discussed above.However, there Is no single warning time applicable to allcollision situations (Applied Psychology Corporation, 1961a).In general, required warning time is greater for higher-speedaircraft, and for larger and less easily maneuverable air-craft, but it is very difficult to specify precisely whatwarning time is required in any given situation.

7. If the assumed threshold illumination of 0.5 mile-candle underlying the data in the figure from which intensi-ty values were obtained is inapplicable to a given situation,

-Ie -

I

Tthen the required Intensity must be changed proportionatelyto the applicable threshold. Factors such as backgroundluminance, the presence of other light sources, and pilotalertness can profoundly affect thresholds.

It is evident from all the above that no simple seti of intensity requirements, even if substantially increasedover present CAR requirements, can provide visual rangessufficient to insure effective visual collision-avoidancecapability in all situations possible under today's trafficconditions. The choices open to the regulatory agency there-fore must reduce to one of three general approaches.

1. It must accept a degree of unavoidable risk, under-standing that there will be occasions when VFR pilots cannoteffectively control the outcome of a possible collisionsituation.

2. It must so restrict VFR operations as to reducethe incidence of these occasions to a small fraction ofJ the current number.

3. It must accept and operate under VFR rules con-I sidera.y mrore complex than present ones, including generalincreases in intensity of navigation lights, differentspecified intensities for different classes of aircraft,and new definitions of limiting VFR conditions tailoredmore to traffic conditions.

These three approaches are not sharply distinguished.t Compromises might have elements of all three approaches.Intensity requirements, for example, must necessarily belimited by important considerations, including the possi-r bility of deleterious effects due to backscatter.

In specifying intensity minima it is helpful to con-sider the difference in requirements for aircraft in dif-ferent -peed classes. Even though they may sometimes beinvolved in collision courses with faster aircraft, thosewith low maximum speed or whose operations are restrictedt to low-speed airlanes do not require intensities as highas aircraft with high cruising speed. On the other hand,the intensities they require must be nearly uniform in alldirections, since the slow aircraft may be approached fromany direction by a fast aircraft, and its speed and the ap.proach direction do not affect the net closing speed as pro.foundly as when both aircraft are fast.

The minimum intensity requirements for position lights,shown in Table 12, has been formulated to provide guidelines,"not to prescribe a detailed regulation. In view of the

109 -

Table 12

Sugested Minimum intensity Requirements (Candles)

for Aircraft Position Lights

Angle from right or leftof longitudinal axis, Low-speed High-speedmeasured from dead ahead aircraft aircraft

00 to 200 100 500

200 to 400 75 300

400 to 1800 50 100

UO -

uncertainty of much contributory data, the values mustbe based or, somewhat arbitrary determinations. There islittle doubt, however, that the minimum intensities listedwould be much more appropriate to actual needs than pres-ent requirements. Different requirements are given forhigh-speed and low-speed aircraft. The class cutoff mustbe somewhat arbitrary. It Is suggested that a maximumcruising speed of the order of 200 miles per hour be used,but detailed investigation may suggest a different or evenflexible classification.

4The present CAR prescribe lower minimum Intensities

at elevation angles above and below horizontal than In theI horizontal plane. This conforms generally with reasonablerequirements, since the visual ranges required in the pres-ence of altitude differences (and therefore elevation angledifferences) decrease with the difference. There is no evi-dence that the present CAR requirements for position lightsare inadequate in this respect. Since they are defined asa proportion of the intensity in the horizontal plane, anincrease In the requirement for the latter is accompaniedby a proportional increase at all elevation angles. It issuggested therefore that this part of the present CAR beT left intact.

Maximum-intensity requirements in the overlaps shouldI be rewritten, in order to insure reasonable cutoffs. Es-sentially, what is, or ought to be required, is that theintensity of the unwanted signal from the adjacent sectorbe sufficiently lower than the desired signal in a givendirection that the desired signal dominates and thereforeadequately identifies the sector. The present CAR recog-nizes the engineering difficulty of obtaining an exactboundary and provides no maximum intensity In the first 10degrees adjacent to a boundary. In the next two adjacentten-degree zones, specific maximum Intensities are given.This makes no allowance for lights of appreciably higherintensities than those prescribed as minimum. It is un-reasonably difficult to meet the same cutoff requirementsin both high- and low-intensity lights when the cutoff max-f imums are given in definite values of intensity. This dif-ficulty, recognized In one instance, resulted in Paragraph4b. 634-1 of CAR 4b, which specifies the cutoff maxi•tmsfor the forward overlap zones with high-intensity lightsas percentages of the forward peak intensities. While thisis an improvement, the intent of the cutoff requirementwould be served best if (a) all cutoff requirements weregiven as percentages and (b) the requirements were givenas percentages of the doginant signal intensity in any givendirection.

-l111 -

This would insure adequate cutoffs in terms of whatan observer would see in a given direction, and would pro-vide equally effective cutoffs with lights of any intensity.

Accordingly, it is suggested that the maximum overlapIntensity requirements be changed as Indicated in Table 13.

It is generally desirable to provide sharp cutoffs atsector boundaries. The somewhat loose requirements givenin Table 13, and in the present CAR, reflect the need tocompromise this requirement with engineering feasibility.If davelopments in the state-of-the-art permit, manufac-turers should sharpen the boundaries as much an is practi-cable, and revision of the regulations should be consideredto call for more precise boundaries.

A particular problem arises with the quadrantal sector-coded system discussed earlier. Here the rear overlap intro.duces a more pervasive ambiguity than do other overlaps.Since there has been no engineering experience with thisproblem, it is inappropriate at this time to require sharperoverlaps at this boundary. Hcwever, as equipment is devel-oped to provide quadrantal coverage, the objective of re-quiring an especially sharp boundary to the rear should bekept in mind.

Anti-collision light intensity requirements. Anti-collision llght~s are omniolrectlonalj, and their intensityat any angle of elevation Is uniform in all azimuthal direc-tions. Therefore the CAR requirement is stated in terms ofthe minimum intensity required at various elevation angles.The possibility (and desirability) of using lights of appre-ciably higher intensities than the required minimums suggestthat the requirement be stated in two ways, a minimum in-tensity in and near the horizontal plane, and, for otherelevation angles, proportions of actual intensity in thenear-horizontal zone.

It might l.e argued that there ls no harm in designinga light with a very intense but very narrow horizontal beam,provided it is above some minimum at other elevation angles.However, this would not constitute good design; by the verynarrowness of the intense part of the beam, its usefulnesswould be more limited than the numerically high value ofpeak Intensity would suggest. The beam shape contemplatedby the over-all requirements should be the goal of designers.In designing higher-intensity cquipment there is no parti-cular difficulty involved in preserving beam shape, andevery reason to insure good design by doing so.

As in the case of position lights, the proposed higherintensities shown in Table 24 are given in classifications

g | e| e || | | | | | i - 112-

Table 13

Suggested Maximum Intensity Requirements

for Position Light Overlaps

Overlap zone Maximum intensitya

100 to 200 from sectorboundary .10 1

More than 200 from sectorboundary .025 I

a "I" is the actual Intensity of the required signalin any specified direction.

- li3 -

Table 14

Suggested Minimum Effective Intensities for

Anti-Colllaion Lights

Angle above or beiow Minimum intensitynorizontal plane Low-spee' High-spe .

aircraft aircraft

00 to 5o 500 candles 2000 candles

50 to 100 .60 Ib

100 to 200 .20 I

200 to 400 .10 1

a If complete coverage Is obtained with two anti-collistnnlights, each intended to cover one hemisphere, the upper orlower, then the indicated directions shall refer only to thecovered hemisphere.

b I is the lowest measured intensity in the 00 to 50 zoneof the anti-collision light under examination.

- 114

!. ." ý, , -r-in' for' jow-:peed aircraft. The general-11,curli~n -j the rean)nlinf behind t-e classification forncnitlon i ahts applies here. The intensity requirements,since these are fl'ahinv lights, are in terms of "effec-tive Intensity," computed from photometric data, as calledfor in the present CAR. In addition to increased Intensity,It will be noted by comparison with Table p0, thai an in-crease in elevation angle coverage from 30 to 40 is calledifcr. Thid is to insure that the anti-collision lights ofbankinT" aircraft Viil remain reasonably visible.

Because of the difficulty of avoiding obscuration byaircraf't sti0uctures, some incompleteness of coverage is per-mitted under the present CAR. It Is almost always possibleto solve this problem, but it is often at the cost of addi-tional light fixtures or troublesome installations. Air-craft covered by Part 4b of the CAR are permitted no morethan 0.03 steradian of obstruction, which must be confinedto 0.15 steradian centered about the longitudinal axisrearwards (a cone with a half-angle of about 12P°). Sincethis obstruction is confined to the rear where possibleclosing speeds are at a minimum, it may be argued that theloss of the high-intensity flashing anti-collision lightand the consequent dependence on the much lower intensitysteady-burninp rear position light does not seriously reducesafety. This argument cannot be used for aircraft coveredby other parts of the CAR, for which a much larger obstruc-tion is allowed (0.5 steradian) without restriction of di-rection. To allow any obstructions at all seems question-able, even if they are restricted to rearward. Statisticsindicate that this dpproach direction accounts for a highproportion of collisions (Fisher & Howell, 1957), eventhough analysis suggests it ought to be the safest direction--an anomaly perhaps explained by the fact that many collisionsoccur in more or less unidirectional traffic near airports.It may also be connected with the fact that pilots are gener-ally 'blind" to the rear so that in an overtaking collisionthe entire responsibility for avoidance rests on one pilot.Finally, if an overtaking collision also involves climbingor descending courses, both pilots may be "blind" (Pisher &Howcl2, 1947; Calvert, T759).

Under these circumstances, greater intensity than nor-mally necessary for detection may be needed to attract thepilot's attention. It is difficult to assess the effect ofseemingly small exceptions to requirements, but deviationsconstitute some increase in risk and should be allowed onlyafter a determineJ effort to obviate their apparent necessity.

Power required for increesed Intensity. Increased in-tensi y can be obtained generally oUny at the expense of in-creased power consumption. since the suggested minimum

- 11) -

intensities for navii-ation ,htir ar, - ,stantjq ij h1ihe:-than present ý.A: ininimum., ther pr.wer reqj'reientn will btccorrespondingly higher. On larqe or h~ih-powered aircraft,these power demands may not seem excess4 'e. r,n a'. ler or

4 even medium-powered aircraft, conrideratlonis of power avail-ability may be decisive in determiningthe feasibility ofthe recommendations, or the extent to which they may haveto be compromised.

Since the proposed requirements differ from the presentCAR requirements not only in retard to generally higher in-tensity, but also in distribution of intensity, and poociblyin color, so for as the taillight is concerned, estimatesof power cannot be based on a simple ratio. In estimatingthe new power demands, the characteristics of a variety ofconventional and newly developed navigation lighto havebeen examined, and the present state-of-the-art in illumi-nating engineering has been considered. The estimates arenevertheless only rough approximations. When actual equip-ment to meet requirements is designed, it may be found pos-sible to improve on these estimates, or If necessary, indifficult cases, to accept lower efficiency and therefore,higher power consumption.

Table 15 lists the power consumption of present typicalnavigation light equipment and estimates of the power con-sumption required to produce the suggested higher intensitieslisted in Tables 12 and 1i.

Traffic Rules, Rules-of-the-Road, and Minimum VPR Conditions

It has been shown that avoiding collisions visually Isdifficult and sometimes impossible when (a) closing speedsare high, (b) flight visibility is marginal, and (c) ap-proaches occur from directions in which cockpit visibilityin restricted.

As the number of aircraft and their cruising speedeincrease, these difficulties increase, and pilots are lessabl6 to detect intruder aircraft In time to analyze collisionprobabilities and to take effective action when indicated.

Suggested improvements in light systems would helpreduce the number of uncertain or impossible collisionsituations. Other steps can be taken by regulatory actionor by operational procedures to improve the situation fur-ther.

From the viewpoint of visual collision avoidance alone,a number of seemingly straightforward steps are suggested.When these are viewed in the context of the entire scope offlight operations, they become more complex.

- 116 -

Tal,,e 159

Power Consumption of Navipation Light Systems MeetingPresent CAV Requiremnents Compared with )Fstimated Power

Consumption of 'yatema Meeting Proposed HigherIntensity Pequirements

Estimated PowerConsumption to

Light Power Consumption Meet Proposed Re-of Present Equip- quirement, Wattsment, Watts 'Iw-Speed Hign-spee

Aircraft Aircraft

Wing(2 required) 50-80 200 750

Tail(I required) 15-30 200 750

Anti-collision(2 normallyrequired) 175-220 300 1200

Complete system 235-320 700 2700

- 117 -

For ex3mplP, there are manry raaF•: 17 th"! *Ir' -e!IJc

col'lsions. There would be little porit to reducirn ' unedanger if in the process the danger from other. hazards wa.increased even more. -conomic conniderations carinnt beignored, competing demande for pilot attention ari for airtraffic facilities complicate the problem of reconcilingdifferent requirements, especially when regulatory actionis involved. The proposals herein have been geared to prac-tical considerations and to reasonable (if heavy) demandson enrIneerinr capability. The suggestions in this sectionmay not lead to regulatory action at all or may be only par-tially incorporated into regulatiomns. Furthermore, the pro-posals are prebented in general form, since the detailedcontent of regalations or procedures is in some cases sub-ject to more investiat•on or analysis.

The proposals may be grouped in the following cate-gories: (a) restricting possible collision courses, (b)raising VFR minimums, (c) increasing cockpit visibility,and (d) changing rules-of-the-road.

Restricting possible collision courses. Paragraph60.32 orh ;e CAR, Part OU, Air TrarrIc Rules, restrictsdirections in VFR level cruising flight 3000 feet or moreabove the surface as shown in Table 16.

These restrictions reduce the possible incidence ofnear head-on collision situations, especially in the densereast-west directions. To the extent that they are effective,they reduce the number of high-closing-speed situationsshown tv *, Jiifficult or impossible to cope with.

According to the rule, opposing traffic along an east-west airway should be at different altitudes. But othertraffic crossing the airway, or airway traffic changingaltitude or for some reason failing to hold to specifiedaltitude, and traffic outside airways, may be involved inconverging courses at any angle. The incidence of nearhead-on approaches is very much reduced by th3 rules, butnot eliminated.

Every effort should be made to set up rules to insureagainst head-on approaches. Quadrantal altitude separationwould be much more effective than the two-sector separationof the present rules, but would be complicated, requiringeither a large altitude change to get to the next altitudewith the same direction, or, alternately, too close an alti-tude segmentation for operational safety.

The use of one-way climbing and descending corridorsis also helpful, aa are the general corridor restrictionsnear airports. In general, pilots of high-speed aircraft

-118 -

Table 16

VFR Crusing Altitudes

AtitutdeMagnetic Course 2e53ow i•,'o bDoVCe i,9PU•

feet feet

00 to 1790 odd thousands 4000-foot inter-plus 500 feet vals begiLnning

at 30,000 feet

1800 to 3590 evnn thousands 4000-foot inter-plus 500 feet vals beginning

at 32,000 feet

- 119 -

wil. find it dlfficult c!r ifmpossi.e tc u av:)Id c;AI.j irwith other air'~raft in head-on approacte; even withmoderate-speed aircraft thie time fPr jeclkian i 3ncrtand the correct decision difficult to determine.

Raising VFf minimums. In genceal, v'Fn minimums shouldbe ra-sec, regar•-ng MoIn fnight visibility and separationfrom clouds. The present regulations (Part 60s paragraph60.30) set up three different minimum flight visibilities,1, 3, and - miles, for four categories of air space. Inthe continental control area, the 5-mile requirement recoe-nizes that speed ranges are high at high altitudes, thusrequiring higher visibility for effective VFR. Even so,with the intensities required by 'he CAR, signal visibilityIn marginal conditions is inadequate at the speeds commonlyencountered.

If intensities can be increased along the lines recom-mended, and head-on approaches eliminated, the presentlyrequired minimum visibility may be satisfactory. Neverthe-less, it is suggested that the minimum be raised to 6 miles,partly to provide a safer limit, and partly because thiswould restrict VFR to atmospheres designated by the Inter-national Visibility Code as "clear" or better. If head-onapproaches cannot be ruled out, then visual avoidance forvery high-speed aircraft must be considered unlikely undermarginal conditions and very difficult at best.

A note in the CAR (paragraph 60.30) indicates that VFRminimums are to be treated conservatively. This advicecannot be overemphasized. As indicated repeatpdly, when=-" f- r ioa marginal--flight visibility, cockpit visi-5fity, pilot attentiveness, convergence angle, speed, etc.--the pilot's task is difficult. When several factors aremarginal, the difficulty may be enormous. In all areaswhere the pilot has a choice he should be very cautious.

For this reason the 3-mile minimum flight visibilityrequirement for control zones, control areas, a.A1 transi-tion areas is too low. These areas may contain both higb-speed and low-speed aircraft. Furthermore, many aircraftare not (and may not be, for a long time) equipped withlights of high intensity by current standards. Trafficdensities are higher, and demands on pilot attention meresevere. It is also likely to be more difficult to insureagainst head-on approaches.

For all these reasons, therefore, it Is felt thet the3-mile visibility minimum should, as suggested for the conti-nental control area, be raised to 6 miles, "clear" or bette-.This change is especially important with present light in-tensities, but is recommended even if intensities can besubstantially increased.

W 1W0

The req-ilreux-nt of %-Mile vqbJ1Jty o0t'aide on-trolled airspace is Var too lo% for avoiding colI2.sicnunless the possible conditions under which collision canoccur are severely restricted. One-mile visibility is a"thin rog" condition. A signal of 100 candles has a rangeof about 1-1/4 miles, and a 1000-candle signal about 1-3/4miles--adequate for relatively controlled low-speed flightnear airports, but not suitable for general flying. Thisrequirement should be raised or should be narrowly restrict-ed to situations where such low visibility can be consideredadequate.

The requirements for cloud clearances are not suffi-ciently conservative. Cloud distances are often difficultto estimate. They should be sufficient for an aircraftnear them to evade an aircraft emerging from them. Therequirement should be reexamined in terms of the accuracywith which separation from clouds can be maintained andthe collision problems presented by emerging aircraft.

The CAR's definition of flight visibility is vague,particularly the part covering nighttime visibility: "theaverage forward horizontal distance, from the cockpit of anaircraft in flight, at which prominent lighted objects maybe seen and identified at night." The definition of nightvisibility should not refer to "prominent objects," whichare irrelevant, but to lights of specified intensity asseen against specified backgrounds. Although It is evidentthat the definition is inadequate, satisfactory definitionsare difficult to establish. The question of a suitabledefinition related to actual determinations of visibilityshould be investigated.

Increasing cockpit visibility. Increased cockpit visi-bilit7 is a-n evIcent and Iong-recognized requirement forachieving effective visual collision avoidance. Generally,a pilot must be concerned with intruder threats from alldirections, yet windows on most contemporary aircraft pro-vide very limited visibility.

Mirrors, periscopes, and similar scanning devices canpartially overcome visibility restrictions, but they areinferior to direct line-of-sight viewing and are thereforepoor substitutes for increased window space. Pilots canalso increase directional coverage by maneuvering, butthis is hardly a useful general technique.

Although no pilot should assume that any approachdirection is safe, the degree of danger in different di-rections varies with different aircraft. Slow aircraftare subjected to nearly uniform danger in all azimuthaldirections. They should therefore be required to have

-121 ,

much la'.---c- indcw areas than taut aircraf't. Fortunate-lythis confornnn with current -. ctlice.

A teriou viciiility deficiency of' modern aircraft isirA sight lines above and below. Most double-blind approaunesinvolve such sight lines. Every effort should be made torequire and obtain larze incrcaseu in upward and downwardvisibility.

No conreivab. ý increases in cockpit window areas canbe expected to provide full visibility. It Is importanttherefore to develop optical devices to supplement directvisual coverage. Since they provide coverage in otherwiseblind directions, their limitations may be tolerable. Theyare likely to be partl.ularly valuable to a pilot in deter-mining before he maneuvers that airspace is clear. Whensuch devices are fully developed and their capability proved,regulationa requiring them and governing their use 8ho:ld beconsidered.

Changin7 rules-of-the-road. Paragraph 60.14 of the CARcontains rules governing maneuvers for avoiding collIsionand specifying rights-of-way in potential collision situa-tions. Many of these rules are concerned with giving prece-dence to aircraft in distress or to less maneuverable air-craft. Othe:'s, however, modeled after marine rules, givethe "right-of-way" to one of two involved aircraft dependingonly on relative sight-line "earing. A note 3uggests thatnD pilot is ever relieved of responsibility for avoidingcollision. However, it is recommended that all potentialcollision courses require complementary avoidance by bothpilots. It may ofte-nTappen that one pilot will not seethe other, so that the burden of avoidance will fall bydefault on only one pilot. But if both see each other,the short time available in many situations requires thatboth maneuver. Furthernore, the pilot who does not havethe right-of-way may fail to detect the other, or he mayevaluate the dang-er of collision incorrectly. In any event,prudence reauires that any pilot who determines that -.collision risk exists take measures to avoid collidir;g.

Designing rules-of-the-road to satisfy this requirementis not easy; any set of rules is likely to involve incorrectmaneuvers when an inevitable degree of uncertainty exists,especially when the closing speed is high. Pilots shouldtry to evaluate collision risk carefully before engaging inan avoidance maneuver but should not delay such a maneuveronce the risk has been Judged to exist. In general, pre-mature maneuvers will not reduce the risk of collision, tothe pilot's knowledge; late maneuvers will not provide su--T-ricient time to escape the zone of possible collision.

-122 -

The rCe3ent :,uiez dci not :e>uire or 3urf&;e3t the Dossi-ulilty or avoidiri;, collision uy climbtinr or de3cendinr,2uch mareuvers are occasionally used and syatims of com-bined turninv and altitude-ch-nging maneuvers have beendevised. Such maneuvers are more troublesome for pilots;in the absence of better altitude-difference informationthan lz zonerally available, ttese maneuvers may be ofdoubtful value for systematic use. However, they areuseful for last-moment acts of desperation. But if alti-tude-coded light systems come into use and provide adequatealtitude information in collision situations, then altitudeevasion might be the preferred maneuver.

Under present conditions, altitude avoidance maneuversmight oe considered for rule-m kint if it can be shown thatthey are r'er.crally effective and wouid oe more so if rulescould be designed that would assure comnlementary maneuvers.

Double-blind overtaking, with one or both a4rcratchanging altitude, is one of the most dangerous collisionapproaches; it has been the cause of a high proportion ofcollisions. Improved cockpit visibility would greatly re-duce this danger.

Rear-view mirrors and optical aids are only a partialanswer, and in any event, are not now available or adequate-ly developed.

A change in the rules may be of some help in reducingthe dan'aer of double-blind approaches. Pilots are requiredto watch for collision possibilities in all directions.Neverthele4z it would be helpful to spell out that thepilot of any aircraft about to engage in a climbing or des-cending maneuver is responsible for making sure that theair is clear of collision threats from normally blind di-reCtions, even though he believes the present overtakingrules qIve him the r.iht of way.

In effect such a rule says that a pilot flying levelhas the right of way over a pilot changing altitude. Thismay sound superfl]jous or self-evident, but the special dan-ger of double-blind vertical collisions suggest that theemphasis of a specific rule might be helpful.

Phase I1I. Long-Range Program to DevelopNew and Improved Nayvgation Liht Systems

The sugrestions in this phase constitute a comprehen-sive prog-ram tc Investigate methods of improving navigationlif[ht systems and of increasing their usefulness. Some of

- 123-

the suggeationa call for radically new concepts, such asaltitude coding. Others are directed toward establizhingdetails of design to insure maximum effectiveneac for agiven concept. Finally some partr of the program are in-tended to Improve visual avoidance techniques, and to as-certain their precise limits of usefulness.

One of the most important techniques available to apilot is the use of the fixity-of-bearing criterion. Itwas pointed out that a key factor in the success of thetechnique is the precision with which a pilot can deter-mine bearing fixity. In order to be able to assess theusefulness of the criterion, it is essential to carry outexperimentE which will provide valid estimates of the thresh-old of bearing fixity detection in operational situations.The thresholds must be determined for steady and flashinglights, and for other possible parameters such as intensityor color. Leibowitz (1955) found that movement discrimi-nation is aided by reference lines in the field of vimw.This suggests the possibility of providing a fine grid inwindow areas, and this should be investigated.

The effect of backscatter on target visibility hasbeen found to be unrelated to differences in the backscatter-Ing light (Applied Psychology Corporation, 1962a). Subjec-tive evaluation of backscatter suggests striking differencesnevertheless. It is thus strongly suggested that the possiblepsychological effects such as distraction, disturbance, ordisorientation, be investigated. The results will be impor-tant in deciding such questions as the choice between red andwhite for the color of anti-collision lights and between dif-ferent flashing modes.

The conspiculty or aircraft lights against backgroundsof city li'hts has been investigated in a static situation(Applied Psychology Corporation, 1962e). Further investiga-tion, with relative movement characterizing most operationalsituations, should b, undertaken.

Various light coding techniques were discussed in Chap-ter III. A broad study should be undertaken to determine,urder comparable operational conditions, the relative effi.-cacy of these techniques, in terms of speed and accuracy ofInterpretations, and suitability for conveying differentamounts of information.

It has been found that pilots can be trained to improvetheir ability to estimate distance and relative altitude ofintruder aircraft (Applied Psychology Corporation, 1962c).In view of the importance of this information in visual col-lision avoidance, a broader investigation should be under-taken to determine what the limits of estimating capabilityare in general operation.

-I -

The use of lights to improve daytime conspiculty hasbeen sugested frequently, It was shown in Chapter VIthat there is little likelihood of providing effectivedaytime lighting for the full range of daytime require-ments. The gravity of the daytime problem suggests thata systematic investigation should be undertaken to deter-mine precisely what can be done with lights.

The most promising development in the informationcontent of light systems is the possibility of codingaltitude in'ormation into navigation lights. This hasbeen discussed Ir detail In Technical Report No. 1 (1961a)and summarized In Chapter II of this report. Two testshave been made. One, in NAPEC's F-100 simulator, suggestedan advantage with altitude coding In pilots' ability todetermine altitude, and a substantial advantage in deter-mining courses involving change of altitude (Applied Psy-chology Corporation, 1962b). The results of the secondtest, a flight test, showed a marked improvement on bothcounts with altitude coding (Applied Psychology Corporation,1962f). These preliminary results support the desirabilityof a systematic investigation. A number of questions re-quire answering:

1. What altitude segmentation would be optimum?

2. What cycle lungth would be optimnum?

3. How shall coding be achieved?

4. What rules-of-the-road would most effectively usethe information?

5. Under what conditions is altitude coding useful?(high altituees? near airports? etc.)

6. What are the engineering problems of deriving analtimeter signal and converting it into light signals?

7. What inaccuracies, confusions, errors, etc., willbe encountered and what will their effect on operations be?

8. What special problems will arise If altitude codingIs to be standardized under the CAR?

Other kinds of Information that may be useful If codedInto navigation light systems are (a) maneuver, (b) speed,and (c) angle of elevation.

Of these, maneuver appears of greatest interest, parti-cularly change of altitude (which may be of particular In.terost if altitude coding should be rejected as Infeasible)

- 125 -

and turn indication, of interest in the use of the fixity-of-bearing criterion. In using the criterion, both air-craft must be on straight, constant-speed courses. Apilot knows whether hib own course meets the requirementsbut general3y does not know about the intruder. Turin-maneuver indication would help him determine this.

The indication may be rudimentary, showing only thatsome mmneuver is under way without specifying what it is;or It may show which of the four basic maneuvers (or two,in a simplified version) is in process. Detailed maneuverinformation would enable a pilot to know that correct comr-lemntary avoidance maneuvers have been undertaken byoth aircraft. (It should be noted that one of the 19

navigation light systems on diaplay at NAPEC includes sucha 4-element maneuver indication.) A research programshould investigate questions of feasibility, usefulness,detailed requirements, etc., of the various possible typesof maneuver coding.

Speed coding appears to have some usefulness, but itis felt to be less promising than either altitude or maneu-ver coding. At the present time it is considered unlikelythat any additional complexity in navigation light systemis justified for this purpose, particularly if altitude ormaneuver information (or both) is added to sector coding.

Elevation angle coding, similarly, appears much lessuseful than altitude coding and does not justify the addedcomplexity.

I I I I II II I I I

REFERENCES

Adrian, W., & Jainaki, P. Experimental studies on the re-liability of the visual threshold internationally agreedupon for the calculation of luminous intensities of lightsignals. Re . No. 5-4-4. Sixth International TechnicalConference on houses anm Other Ais to .aVi Ot-Ion,V-73. coast quardl, 1.96o,

Applied Psychology Corporation. Analysis of the usefulnessof coded information in visual collision avoidance.Tech. Rep. No. 1, under Federal Aviation Agency contractFAA/ZHP-i27T, I9W1a.

Applied Psychology Corporation. The role of paint In mid-aircollision prevention. Final Rep. No. 1, under FederalAviation Agency contract FAA/BR-127, 1961b.

Applied Psychology Corporation. Effects of backcaatteredlight on target light detectability in a ground testenvironment. Tech. Revs. Non. 6. 9,& 14 (combined).under Federal Aviation Agency contract FAA/BKO-127, 1962a.

Applied Psychology Corporation. Flight simulator tests ofaltitude-coded lights. Teob, Rep, No. 8, under FederalAviation Agency contract'NAt/BlI7T127,1902b.

Applied Psychology Corporation. Pilot Judgments of rangeand relative altitude: ground-to-air and air-to-airobservations. .Tech Rees, Nos, 10 & 11, under FederalAviation Agency contract FAA/,5W-1 1962c.

Applied Psychology Corporation. Distance estimation of fre-quency-coded and uniformly flashing lights. Teb, ReNo. 12, under Federal Aviation Agency contractFAA/B-7T27T962d.

Applied Psychology Corporation. Conspiculty of selectedsignal lights against city-light backgrounds. Tech ReNo. 13, under Federal Aviation Agency contract 7AA7BRD•T2--Tp62e.

Applied Psychology Corporation. Flight tests of altitude-coded aircraft lights. Tech Rep. No. 16, under FederalAviation Agency contract FAA/BRD-I27, 962f.

Applied Psychology Corporation. The role of optical devicesIn mid-air collision prevention. F l 2Reg. No. 5. underFederal Aviation Agency contract FfaIWTTTog.

Blackwell H. R. Contrast thresholds of the human eye. J.opt. so,. Amer., 1946, *, p. 24.

- 127 -

Blondel, A., & Rey, J. The perception of lights of shortduration at their range limits. Transactions. Illuml-natin Engineerit Society, London." VoJ. V1 No. a,

Breckenridge, F. C. A correlation of signal color recogni-tion tests a pe.o -- 2 it nentoa

Technical Con

Breckenridge, P. C. U. S. standard for the colors of signallights (tentative). Prepared In collaboration with theU. S. National Committee on Colors of Signal Lights,under sponsorship of the U. S. National Committee of theC. I*.So# 1962.

Calvert, E. S. Some operational aspects of the use of aircrew-Interpreted devices for preventing collision in the air.Royal Aircraft Establishment,, ICAO Working Paper AIR C'-

Calvert, E. S. Collision avoidance by sea and air, somerecent developments In theory and practice. Royal Aircraft Establishment Tech. Memo. No. EL 1840, JI"nury 96o.

Calvert, E. S. A comparison of two systems for avoidingcollision. Journal of the Institute of Navigation (London),1961, a4, P. 379g.

C. I. E. Technical Committee 1.3.3. On the colors of signallights. C. I. E. (International Commission on Illumina-tion) Proceecings, Voi. 1, Annex to jecommenaatzons 55.

C. I. E. Technical Committee 3.3.2.1. Aviation ground light-ing, C. I. E. Proceedings, Vol. II, Committee Report, 1955.

C. I, E. Colours of signal lights. Publication No. 2(-33) 1959.

Civil Air Regulations, Part 60. Air traffic rules. CivilAeronautics Manual 60, May 15, 1961.

Cohen, J., & Dinnerstein, A. J. Flash rate as a visual codingdimension for information. USAF WADD Tech. Rep. 57-6J, May1958.

Cook, K. 0., & Beazley, R. H. Determination of identifiablelight flash patterns. J. Engng Psychol., 1962, ., pp. 19-25.

Crawford, A. The perception of light signals. (2) The effectof the number of irrelevant lights. Road Research Labora-to!Z, De-t. of Scientific and Industrilal 9emeawon, RFeschNW NO NjJfj0IVZAU, turea; Brit2 in) uclover A~YMP

-12S -

Crawford, A. The perception of light signals. (3) Theeffect of mixing flashing and steady irrelevant lights.Road Research Laboratory, Dept* of Scientific and Indus-

Curcio, J. A., & Knestrick, a. L. Correlation of atmospherictransmission with backscattering. J. Opt. Soc. Amer., 1958,48, p. 686.

de Boer, J. B. Visibility of approach and runway lights.Philips Research Reports, 6, June 1951.

de Vienne, J. Avoidance of aerial collisions using pictorialtraffic display in the cockpit. Soc. Automotive Engineers,National Aeronautic Meeting, April 1956.

Douglas Aircraft Co. Anti-collision light study. Report No.sM 227, Warch 1957.

Douglas, C.A. Some factors affecting the relation betweenreported visibility and visibility from aircraft. Nation.al Bureau of Standards PFo. No. 2715, August 12, 153.-

Douglas, C.A. Computation of the effective intensity offlashing lights. Illuminating Engineering, 195','. 52, p.64l.

Edwards, T. M.$ & Howell, W. D. Recommendations on cockpit.visibility standards for transport-type aircraft. CivilAero Admin., Tech. Dev. Center, Tech. Dev, Rep. No. JFeDruary 199o.

Emerson, 0. 0., Metcalf, R. D., & Olover, H. C. The inade-quacy of visual search in avoiding mid-air collisions.USAF WADC Technical Note 2§-145 , March 1956.

Federal Aviation Regulation, Part 1. Definitions and abbre-Vlarionse may 90d.loe

Federa2 Standard No. 3, Aeronautical Lighting Colors, 1951.

Fiore, M. V. Visibility, Is it possible at today's highspeeds? Northrup Aircraft Co. Rep. No. 29-250. presentedat teohnicai conrerenoe ox Ill. zngng Zo0., AircraftLighting Committee, April 1959.

Fisher, R. B. Preliminary flight tests of the Farrand Opti-cal Scanner. Civil Aero. Admin. Tech. Dev. Center, Tech.Dev. Rep. NO*_ _j25, October' l.y'ff

Pisher, R. B., & Howell, W.D. An investigation of someparameters related to mid-air collisions of aircraft.Civil Aero. Admin,, Tech. Dev. Center, Tech. Dev. Rep, No.

r Ly:)-t *- 129 -

Fisher9 W. C., Noft'singer, L. R., & Robinson9 J. E. Flighttest procedures for the obaectlve evaluation or aircraftexterior light systems. U. S. Navy, Bureau of Aeronautics,1958.

Oebhard, J. W. Target coding by means of visual flicker.Institute for Cooperative Research, Johns Hopkins Univer.lie,. 100..I-OOp'WUODWF .LY�O 8

Oebhard, J. W. Preliminary reconuiendations on the types ofvisual codes prepared for integrated electronics displaysyitems. Institut, for cboperative Research, Johns Hop-Icizis Uraiver., plO�I-2O5, November 1949.

Gilbert, N. S., & Pearson, N. J. C. A study of the visibilityand glare ranges of slope-line approach lights. CivilAerQ Adml.ak. Tech * Dev. Cen�er, Tech. �v. Rep. .�77T�November i4i.

Godfrey, 0. we Aircraft position light system. Research anddevelopment Interim Report, North American Aviation Rep.No. NA 59H-�l5, November 195!i.

Grimes Mfg. Co. Retractable rotating anti-collision beacondrawing No. o-8450.

Hill, N. E. 0. The recognition of coloured si�nal lightswhich are near the limit or visibility. Proc. PhysicalSociety (London), 1947, 59, part if, p. 56u.

Jfollingdale, S. H. The mathematics of collision avoidancein two dimensions. Journal of the Institute of Navi abion(London), 1961, li#,

Honeywell Aeronautical Division. Honeywell-Atkins maximumsafety light. N! Aero Document R-ED 7822, 1961.

Hopicinson, R.0. Measurements of sky luminance distributionat Stockholm. J. Opt. Soc. Amer., 1954, 111f, p. 455.

Ho��ell, W.D. Determination of daytime conspicuity of trans-port aircraft. Civil Aero. Admin. Tech. Dev. Zenter

Howell, W. P. An investigation of some means to improve thedaytime conspi@US.ty of aircraft. Civil Aero Aduin. Tech.

Jainski, P. Calculation of luminous intensities of lightsignals at great 5QrZ'OWhdiflg bk'ightne.s. Re rt No. 1-il,Sixth International Technical Conference o

. , r

-l�i.

Kinney, J. A. S. Comparison of scotopic, mesopic, and pho-topic spectral sensitivity curves. J. Opt. Soc. Amer.,1958, 48 p. 185.

Knoll, X. A., Tousey, R., & Hulburt, E. 0. Visual thresholdsof steady point sources of light in fields of brightnessfrom dark to daylight. J. Opt. Soc. Amer., 1946, 36, p. 480.

Koomen, N. J., & Dunkelman, L. Visibility of light sourcesthrough fog. Naval Research Lab. Report 245, December 1953.

&

Koomen, M. J., Look, C., Packer, D. M., Scolniok, R., Tousey,R., & Hulburt, E. 0. Measurements of the brightness ofthe twilight sky. J. Opt. Soc. Amer., 1952, 42, p. 353.

- Lamar, R. S. Operational uaokgruund and physical considera-tions relative to visual search problems. In Ailene Morris,& E. P. Hore, (Eds.) Visual Search Techniques. Washington,D. C.: National Research councia, Fub'lication 712, 1960,pp. 1-9.

Langmuir, I., & Westendorp, W. F. Study of light slgnals inaviation and navigation. Physics, 1931, 1, p. 273.

Laufer, M. K. Study to determine optimum exterior lightingfor aircraft, USAP WADC Tech. Rep. 55-309, 1955.

Leibowitz, H. Effect of reference lines on the discrimina-tion of movement. J. Opt. Soc. Amer., 1955, !2, p. 829.

Mackworth, N. H., & Kaplan, I. T. Visual acuity when eyes arepursuing moving targets. Science, 136. May 4, 1962, p. 387.

McNicholas, H. J. Selection of colors for signal lights.Bureau of Standards Journal of Research, 1936, Ila p. 955.

Marshall, P. R., & Fisher, R. B. Determination of the day-time conspicuity of a small airplane in a terminal area.Federal Aviation Agency, Tech. Dev. Center, Tech. Dev..Ep. =o. joe, january Ly:y.

Mertens, J. J. Influence of knowledge of target location uponthe probability of observation of peripherally observabletest flashes. J. Opt. Soc. Amer., 1956, 16, p. 1069.

Middleton, W. E. K. Vision through the atmosphere. Toronto:U. of Toronto Press, ±9".

Middleton, W. E. K., & Oottfried, H. S. T. Apparent intensi-ties of colored signal lights. Illuminating Enineering,1957, .2, p. 192.

1 131 -

Miller, .7. W. Study of visual acuity during the ocularpursuit of moving test objects. II Effects of directionof movemefit, relative movement, and illumination. J. Opt.Soc. Amer., 1958, 48, p. 803.

Moon, P. Scientific basis of illuminating engineering. NewYork: -NlI-L, 1930.

Mullis, W. P., & Projector, T. H. Comparative tests of theeffective intensity and luminous efficiency of severaltypes of anti-collision lighte. National Bureau ofStandards Report 2.1/21N-2/7, JuTy j190.

Projector, T. H. Effective intensity of flashing lights.Illuminating Engineering, 1957, 52, p. 630.

Projector, T. H. Efficiency of flashing lights. Illuml-nating Enilneering, 1958a, 22, p. 60.

Projector, T. H. Intensity as a factor in the evaluationof aircraft navigation light systems. Tech. Conf.,Aircraft Lighting Comm. or Illuminating Engineering Soc.,May 1958b.

Projeoctor, T. H. Factors in the design of navigation lightsystems. Technical Conf., Aircraft Lighting Comm. ofIlluminating Engineering Soc., April 1959a.

Projector, T. K. Efficiency of flashing lights: comparisonwith steady-burning lights. Illuminating Engineering,1959b, 1, p. 521.

Projector, T. H., & Robinson, J. E., Jr. Mid-air collisionavoidance with navigation light systems. Airway Moderni-zation Bd., September 1958.

Rautian, 0. N., & Speranskaia, N. I. Colours of marinesignal lights. Trudy Goo. Ordena Lenina OpticheskogoInstituta. Im. S. I. Vavilova, 24(144): 1-42, 1955,National ReseaorH Council or qarnaa, Technical Trmnsla-

Robinson, J. E., Jr. Human e ineeri tests of selectedaircraft anti-col11lo .2atss _ ITICn"7. :

u. 3. Wavy Sureau orAeronau mos, 19.

Sperling, H. 0. Visual aspects of air anti-collision lighting.

Amendix B to 14H Aero DocumentR-ED 7822,, Honerwell Aore-

R1111011 Divsio-. 132 .

Stiles, W.S., Bennett, M. G., & Green, H. N. Visibilityof light signals with special reference to aviation lights.Aeronautical Research Committee Reports and Memoranda No.17•93, Brilt•S1 Air Prinistr•, aY- 97 . ...

Stone, R.A. Untitled paper presented before Aviation Light-ing Committee of the Illuminating Engineering society,1954.

Trumbull, R. Accidents and weather Aerospace MedicalAssociation conference, April 1992.

U. S. Coast Guard. Aids to navigation manual. Chapter 23,June 1956.

Wulfeck, J. W., Weisz, A., & Raben, M. W. Vision in militaryaviation. USAF WAX Tech. Report 58-399. November 1958.

Zeller, A. P., & Burke, J. M. USAF mid-air collisions.USAF Aero Medical 3afety Division, orte of Flightsarety Research, Publication NO. i•-5$, November 1y5.

Ziedman, K. Evaluation of anti-collision lights. Doui.UAircraft Co. Report No. SM-22834, July 1957.

-