The EngineeringResource ForAdvancing Mobility 400 COMMONWEALTH DRIVE WARRENDALE, PA 15f

ica

830567

Motor Vehicle Forward Lighting

Michael Perei, Paul L. Olson,

Michael Sivak and Jere W. Medlin, Jr.

International Congress & ExpositionDetroit, Michigan

February 28 - March 4, 1983

830567

Motor Vehicle Forward Lighting

Michael Perel, Paul L. Olson,Michael Sivak and Jere W. Medlin, Jr.

MOTOR VEHICLE FORWARD LIGHTING

ABSTRACT

This paper surveys the literature on motorvehicle headlighting and its influence on theability of drivers to avoid accidents. Thereview identifies the key relationships betweenheadlamp design characteristics and driver andenvironmental factors. The major safetyproblems associated with headlighting arediscussed, and issues needing the attention ofthe research community are identified.

THE NIGHTTIME ACCIDENT RATE IS OVER THREETIMES THE DAY RATE. Many factors can be blamedfor this difference including alcohol, youngreckless drivers, and fatigue. It is alsoreasonable to assume that the poor visualenvironment at night can affect safety since somuch of the information required by drivers isobtained visually.

There are a number of major differencesbetween the visual environment at night and daythat can make night driving a difficult task.At night, the broad fields of view of daylightnarrow to the region illuminated by artificiallighting—headlights or fixed road lighting.At night, objects are seen under lowercontrasts and non-uniform illumination, whichmakes detection difficult. As ambient lightdecreases, colors fade, shadows disappear,resolution is lowered, and object detailvanishes. Peripheral vision is reduced,causing the driver to modify search and scanbehavior. Increased glare causes eye strainand fatigue and further increases thedifficulty of detecting hazardous objects.

Various aspects of the highway environmentcan further degrade driver vision at night.The variation in pavement reflectivity fromasphalt to concrete can affect the contrastand, hence, detectability of potential

hazards. Adverse weather can reduce visibilitydistances in many circumstances. In fog,backscatter from headlamps increases glare andreduces the ability of headlights to illuminateroad delineation and other critical objects.In rain, the specular reflectance of light fromroadways makes it virtually impossible fordrivers to see lane markings.

Not all drivers are affected to the samedegree by the nighttime visual environment.For example, drivers with various visualdeficiencies can be more sensitive to nighttimedriving problems. Some of the visualcharacteristics which can reduce drivers'visual performance at night include glareresistance, contrast sensitivity, peripheralacuity, and night myopia. Drivers who arefatigued are more adversely affected by thenighttime visual environment because of changesin their visual search behavior and even lowersensitivity to peripheral cues. The use ofalcohol and other drugs can impair drivervisual information processing and make nightdriving more dangerous.

Vehicle characteristics can also play arole in influencing the nighttime visualenvironment. Headlamps are the primary sourceof illumination for drivers on unlightedroads. The design goal of headlamps is toilluminate the roadway and potential hazardswithout subjecting oncoming drivers toexcessive glare. There are a number ofheadlamp design characteristics which caninfluence driver vision and, hence, capabilityto avoid accidents, including beam pattern,intensity, and lamp construction. Theperformance that headlamps are designed toachieve can be degraded by misaim, dirt, lampoutage, and incorrect voltage.

In order to identify the major problems inheadlighting, the above driver, vehicle, andenvironmental factors were examined in a reviewof the literature to determine their frequencyof occurrence, their relation to crashes, andtheir relation to driver performance. The

0148-7191/83/0228-0567$02.50Copyright 1983 Society of Automotive Engineers, Inc.

review also describes the historicaldevelopment of the design of headlighting andtries to identify the safety-related rationalefor the various engineering developments andstandard requirements.

Most of the review focuses on data fromlaboratory and field tests and computermodels. Very little information existsdescribing the exposure characteristics ofdrivers, the relation of the headlampcharacteristics to accidents, and the rationalefor headlamp development over the years. Wherepossible, the relation of headlamp design tosafety was determined through the use of theproxy measures that have been used in lab,field; and computer modeling studies. Theserelationships show the sensitivity ofsafety-related performance to variations inheadlamp characteristics. The factors, such asdriver and environmental characteristics thatinteract with headlamp design to affectperformance were also examined. The goal ofexamining all the above types of data was todetermine the nature and extent of the hazardsassociated with headlight factors and theextent to which modifications to headlampcharacteristics or their interacting factorscould lead to improved driver performance.

One problem area that became clear at theoutset of the review was the need to improveperformance of the low beam rather than highbeam. This need arises from two basicreasons. First, most night driving is on lowbeam because of the high traffic densities thatare encountered. Second, high beam intensitieshave reached the point of diminishing returnsbeyond which no significant increases in seeingdistance could be gained without extremely highlevels of intensity. Thus, the majority of thereview will focus on low beam.

REVIEW OF PERTINENT ACCIDENT EXPERIENCE

VISIBILITY - If the great bulk ofinformation necessary for the safe andeffective operation of a motor vehicle isacquired visually, and if darkness impairsvision, then lighting systems which improvevision ought to have an effect on safety.Researchers have been more successful incorrelating accident rates with fixedillumination systems characteristics thanheadlighting. The reason is that controlledstudies are easier to run with fixedillumination systems than headlights. Thereare thousands of miles of roads illuminated atdifferent levels. Lighting systems are beinginstalled where none existed, and old systemsare being upgraded somewhere at almost anytime. Sometimes it is even possible to alterthe lighting levels on a particular street andstudy the effects.

There are several good summaries of thisresearch. One of the better is provided byGallagher and Janoff (1)*. The results ofthese studies are quite variable. Among thereasons are that changes in lighting levels

were frequently confounded with changes intraffic volume, upgraded enforcementactivities, or other roadway modifications. Insome cases accident rates actually increasedwith illumination levels. But, the generaltrend is toward fewer accidents as illuminationlevels increase. Two examples will be offeredfor purposes of illustration.

One of the better recent studies is byScott (2). He looked at the relationship ofeight variables to accidents at 41 sites inGreat Britain. Average road surface luminancegave the best correlation and it could not beimproved by the addition of other variables.Scott's data are summarized in Figure 1, andshow that the proportion of accidents duringhours of darkness dropoed in a relativelylinear fashion as the level of illuminationincreased.

0.5 1.0 1.5 2.0Average road surface luminance (cd/m2)

Fig. 1 - Best fitted.relationship betweenDark/Day accidents and averageroad surface luminance (2)

Another recent and very interestinganalysis has been reported by Merritt et al.(3). Using data from the 1977 Fatal AccidentReporting System (FARS) developed by theNational Highway Traffic Safety Administration(NHTSA), they compared the ratio ofsingle-vehicle fatal accidents on straight andcurved road sections under three lightingconditions: daylight, night but lighted andnight not lighted. An update of this analysisusing 1980 FARS data is reproduced in Table 1.The percentages in the first two columns [i.e.,daylight and night (but lighted)] arestrikingly similar. However, the proportionoccurring on curves is higher on unlightedroads at night, as shown in the third column.These data suggest that improved lighting helpsto reduce accidents on curves.

*Numbers in parentheses designate References atend of paper

TABLE 1 - Summary Of 1980 Accident StatisticsRelating To Lighting Conditions, RoadGeometry, and Run-Off Road Accidents

SINGLE VEHICLE FATAL ACCIDENTSAS A FUNCTION OF LIGHTING AND ROAD GEOMETRY

DAYLIGHT NIGHT (BUT LIGHTED) NIGHT (DARK)

Straightand Level

Curvedand Leve-l

Totals

5454 (67*)

2693 (331)

8147 (100)

3009 (641)

1663 (361)

4671 (100X)

6758 (581!)

4876 (421)

11634 (IOCS)

Straightand Grade

Curvedand Grade

Totals

2137 (461)

2539 (5SX)

4676 (100X)

714 (451)

B73 (551)

1587 (lOOt)

2329 (391)

3575 (611)

5904 (1001)

All Straight(Level + Grade)

All Curved(Level + Grade)

Totals

7591 (591)

5232 (411)

12823 (1001)

3723 (591)

2535 (411)

6258 (1001)

9087 (521)

8451 (481)

17538 (1001)

It seems clear that better roadwaylighting reduces accidents at night. But, doesthis hold true for headlighting as well?Unfortunately, there is no direct evidence thatit does. The problem is that the variousconditions mentioned earlier that made itrelatively easy to run controlled studies offixed lighting have no parallels inheadlighting.

The best evidence for headlighting as afactor in accidents comes from in-depthstudies. For example, the Tri-Level studycarried out at Indiana University (4) foundthat improved forward lighting would have beenof at least possible benefit in about 3% of theaccidents studied. However, there are problemswith the in-depth approach that may tend towork against variables such as lighting. Forexample, a single-vehicle, run-off-the-roadcrash may be ascribed to driving too fast orpoor delineation. However, better limitingmight have provided the operator withsufficient information to properly adjust hisor her speed or see where the road was goingdespite poor delineation. In short, unless theoperator ran into something because of lack ofvisibility, inadequate headlighting could be acontributory factor that is easily overlooked.

Another assessment of the potential roleof headlighting in accidents was presented byHale and Blomberg (5). They estimated thepercent of pedestrian and bicycle accidentswhere poor visibility could have been a causalfactor. Since improved headlighting is oneapproach to increase visibility, theirestimates provide upper bounds on theheadlighting problem. Using data from in-depthaccident investigation studies, they calculatedthat from 3-23% of all pedestrian accidentswere attributable to poor visibility. The

range for bicycle accidents is 23-32%. Thelarge variability in these estimates is due inpart to the difficulty of judging the role ofvisibility in accidents.

GLARE - If glare reduces the ability tosee, then glare should cause accidents. Someevidence that it does is provided by Burger etal. (6), who used a critical incident approachto identify vehicle driver mismatch problemscontributing to accidents or near accidents.Based on 3,500 survey returns, citing nearly1,700 accidents or near accidents, the authorsconclude that glare from the headlamps ofoncoming vehicles is the single most importantdriver-vehicle mismatch problem to which NHTSAshould direct its attention.

What does not become clear from these datais whether the glare problem is caused byfailure to dim, misaim, or inadequate low beamdesign. Clearly, the solution recommendeddepends very much on the source of the problem.

In sum, the two attributes most importantin headlamp design, illumination and glare,both seem to relate to the probability ofaccidents. What we do not have is therelationship of various levels of headlightillumination and glare to accidents. Further,the difficulties of acquiring such informationare such that it is unlikely to be available inthe foreseeable future. In lieu of accidentdata, researchers must rely on surrogatecriteria such as visibility distance toevaluate the problem of nighttime visibility.

WEATHER - Adverse weather can make therelatively poor visual conditions at night evenworse. However, there is only limitedinformation on the relation betweenheadlighting, adverse weather, and accidents.

Koth et al. (7) reviewed data from anumber of studies of fog and accidents. Theconclusions of this review were that nighttimeincreases the incidence of fog accidents. Thescope of fog accidents is minor, representingabout 2 or 3% of all accidents, but theseverity of fog accidents can be great.

Rain can increase glare and affect thecontrast of critical objects illuminated byheadlamps. Accidents in rain were investigatedby Haghighi-Talab (8) and reviewed in an OECDreport (9). The results of these studiesindicate that the degrading effects of rainfallon vision contribute to accident causation.

None of the studies of adverse weatherwere able to associate headlamp design withaccident rates. However, the influence ofheadlamp design on visibility in adverseweather is an important factor that should beconsidered in headlamp evaluations.

REVIEW OF EXPOSURE AND OTHER EXPERIENCES OF THEDRIVING PUBLIC

The seriousness of the night drivingsafety problems is demonstrated in part by thepoor performance of drivers in being able tostop safely, read signs, and follow the roadpath. It is also a function of the exposure of

drivers to the more hazardous lamp designcharacteristics and environmental conditions.Another component of the night driving hazardis the extent to which driver limitations canreduce the seeing distance provided byheadlamps. Unfortunately, there is onlylimited data on the exposure characteristics ofthe night driving environment and the range ofcapabilities of the driving population.

VEHICLE AND LAMP CHARACTERISTICS - Aconsiderable number of vehicle and lampcharacteristics affecting night drivingperformance vary from car to car. Theseinclude beam intensity, beam pattern, beamcolor, aim, lamp mounting height, driver eyeheight, lens and reflector deterioration,windshield tint, dirt, inoperative lamps, andlamp voltage.

Beam intensity and pattern, althoughregulated by FMVSS 108, can vary significantlyfrom manufacturer to manufacturer. FMVSS 108references SAE photometric requirements, whichset maximum and/or minimum intensity values fora limited number of points* in the beampattern. The permissible range of values forsome of the test points can be relativelylarge. Thus, a variety of beam patterns andintensities can be designed to meet thestandard. The quality control employed in themanufacturing process can also affect variationfrom lamp to lamp.

Most U.S. vehicles use sealed beam lampswhich are basically similar in terms ofintensity pattern. Intensity levels do varyfrom lamp to lamp. However, no systematicsurvey of this variability has been published.A small selection of tungsten sealed beam lampswas examined by Bhise el al. (10), who foundvariations in maximum output as great as two toone.

Motorcycles can use European non-sealedheadlamps with a "sharp cut-off" beam pattern.It is not known how many "sharp cut-off" lampsare currently in use in the U.S.

In recent years the frequency of use ofhalogen sealed beams has increased. One reasonfor the advent of halogen sealed beams is totake advantage of the raised maximum intensitynow permitted on high beam by Federalstandards. No published data has been foundwhich reports the actual intensities of halogenhigh or low beams. Also no information on thepercent of cars equipped with halogen lamps wasfound.

The halogen lamps operate at higherfilament temperatures and hence emit a whiterlight than tungsten units. The effect of suchcolor differences on safety are nil althoughsome drivers have the impression that they cansee better with the whiter light. In France,headlamps are yellow, a policy based on theassumption that it reduces glare. Yellowheadlamps (other than fog lights) are illegalin the U.S. In Europe the lamp color issue hasbeen extensively researched (11, 12, 13). Thefindings indicate that if there is any benefitto yellow headlamps, it is slight. These

benefits must be balanced against certaindisadvantages, such as a slight loss in seeingdistance, and change in apparent color ofroadside objects.

Lamp mounting height and driver eye heightin cars have been decreasing since the 1950'sas car manufacturers have sought to change thelooks of their cars. In the early 1950's lampheight was about 32 inches. Currently theminimum mounting height is 24 inches measuredto the center of the lamp. For reasons ofinternational harmonization and improvedaerodynamics some proposals have suggested thatthe height be slightly lowered even further.Driver eye heights in cars have also droppedover the last 30 years, from about 3.95 feet to3.5 feet (for the 15th percentile) (14).Although there are many drivers in vans,pickups, and trucks with higher eye heights,the exposure of drivers with low eye heights isthe critical factor determining the degree ofglare hazard.

Misaim and dirt on lamps are encounteredrather frequently by drivers. They arediscussed more fully in a later section of thereport.

Lens and reflector deterioration of glasssealed beam lamps is practically non-existent.A very limited number of plastic headlamps havebeen introduced and drivers may encounter somelens abrasion depending on the quality of theprotective coating and the harshness of theenvironment.

Defects, whether in-operative lamps orlamps operated at less than design voltage area problem of unknown size. The Ford MotorCompany in their comments to the NHTSAestimated that 51.9 million headlamps werereplaced on passenger cars in the U.S. in 1979,of which 63% were due to burnout. It is notknown how many nighttime miles are travelledwith one lamp out before it is replaced. Withhalogen lamps, sealed or unsealed, it ispossible for a driver to use them even if thelens has cracked. Under this conditionmoisture and dirt can enter the lamp and reducethe road illumination.

Drivers have also been exposed tovisibility degradation at night because of theincreased use of tinting on windshields. Notonly do more cars use tinted glass, but thetransmissivity has decreased over the last 30years. The estimated percent of cars withtinted windshields is about 80%. Lighttransmittance through a' tinted windshield canbe as low as 65% depending on installationangle, compared to 82-84% transmittance for anon-tint windscreen.

ENVIRONMENTAL CHARACTERISTICS - The natureof the roadway enviroment to which drivers areexposed also affects night driving safety andthe performance of headlamps. Environmentalfactors include weather, roadway design, andcritical object visual characteristics.

The weather under which drivers mustoperate varies from region to region. Asummary of weather patterns has been prepared

by Thomas and»reported by Olson et al. (15).These data show that the incidence of fogvaries from practically never in much of thecountry to 10 to 20% of the days in coastalareas. The incidence of precipitation of 0.01inch or more varies from 20 to 50 days in theSouthwest, to 100 to 150 days in the Centraland Eastern regions, to 150 to 170 days in thefar Northwest and around the Great Lakes.Bhise et al. (10) estimated that the typicalU.S. driver will encounter 91 days of drivingannually with precipitation greater than 0.01inch. Thus, adverse weather cannot be ignoredin evaluations of headlamps performance.

The distribution of artificialillumination has been reported by Bhise et al.(10) and is reproduced in Table 2. These datashow that artificial illumination is rare onrural roads. This suggests that mosthigh-speed driving is done using only headlampillumination.

TABLE 2 - Frequency Of Fixed Roadway Liahtingon U.S. Roads (10)

100

ENVIRONMENT

InterstateSystem

Al 1 OtherFreeways

Divided Roads(Nonfreeways)

Undivided Roads(Over 2 Lanes)

Undivided Roads(2 lanes)

XIllu-minated

6.0

5.3

7.9

0.9

2.5

Rural

X NotIlluminated

94.0

94.7

92. 1

99.1

97.5

Urban

X illu-minated

91.1

85.5

87.8

87.8

55.2

X NotIlluminated

8.9

la. 5

12.2

12.2

44.8

Obstacles on the road must be brighter ordarker than their background in order to bedetected. The reflectivity of pavements, thatoften form the background for objects ofconcern to drivers, has seldom been measuredaccurately. Some data have been provided byBhise et al. (10) and are reproduced in Figure2. These data should be compared with those inFigure 3, also from Bhise et al. (10).Clearly, much clothing has reflectivecharacteristics similar to road pavements thatmight constitute a background in a potentialcollision situation. Thus, low contrastsbetween targets and background are common whenthe target is a pedestrian.

The exposure of drivers to hills andcurves can also influence headlampperformance. A beam pattern that provides goodvisibility distance under straight, flat roadconditions, may produce too much glare tooncoming drivers on curves or hills. A surveyof road topography was performed by Bhise etal. (10) and incorporated in the Ford headlampevaluation model, which is discussed later inthe paper.

Summer Clothing (Beiow Knee)Summer Clothing (Above Belt Line)Winter Coats

N=80 Subjects

10 20 30 40 50 60

Luminous Reflectivity (Percent)

70

Fig. 2 - Cumulative distribution ofpedestrian clothing reflectance(10)

JU

0)0

TOO0)

Per

cent

Ret

rore

f-*•

rv

0

0

C

•

t

'C'-%•

•'

• -V

• • *

•. •f

I "

•

*

—

200 400 600 800 1000

Distance trom Driver (feet)

Measured pavement retro-reflectivity characteristics (10)

In summary, the above types of factorsform the environment under which headlamps mustprovide visibility to drivers. Driver safetycan be reduced if headlamp design does not takethese environmental factors into account.

DRIVER CHARACTERISTICS - A number ofdriver related factors can affect thevisibility provided by headlamps. Thesecharacteristics include vision, fatigue,alertness, and alcohol. The distribution ofthese characteristics in the driving populationand their effect on headlamp performance isonly partially known.

A number of visual deficiencies can affecthow well drivers can see at night. Theseinclude glare sensitivity, lens yellowing,contrast sensitivity, and acuity. The degreeto which these factors can vary from individualto individual can be significant. They havealso been shown to vary with age.

One source of information on the variationon drivers' response to glare is by Pulling etal. (16). They measured the headlight glareresistance of 30 subjects in a drivingsimulator. They increased the intensity ofsimulated oncoming headlights until the subjectslowed down or otherwise gave evidence of beingimpaired. The headlight glare resistance wasmeasured as the logarithm of the ratio, at thethreshold for glare acceptability, between theilluminance from oncoming headlights and theambient illumination. The results are shown inFigure 4, A small scale survey was alsoconducted to assess the log glare ratios foundin typical driving environments. These-measurements showed glare ratios as much, as50:1 (log*!.7) on residents! streets and over500:1 (log=2.7) on an unlighted interstatehighway. Comparing these values to theheadlight glare resistance suggests thatincreasing age can have a significant impact onthe ability of drivers to drive under manytypical night conditions.

in 3.0uz2-5uaK2.0

0.5

20 40 60AGE

80

Fig 4 - Headlight glare resistance vs.age (16)

Another indication of the variability ofdriver visual capabilities with age is shown ina study by Sturgis and Osgood (17). Figure 5from that study shews how visual acuitydecreases as luminance is reduced and as ageincreases.

Of the many other factors that caninfluence driver performance, such as fatigueand alertness, little is known about theirfrequency of occurrence in the population.

-ig.

•0.5 0-0 0.5 1.0 15

LOG BACKGROUND LUMINANCE (CO/SQ M)

5 - Effect of background luminanceon visual acuity of three agegroups (17)

Considerable information does exist about thealcohol levels of the driving population.Blood alcohol concentration levels varydepending on time and day of the week, reachinga peak in the early morning hours of weekends.At such times, one driver in eight may have ablood alcohol level of .10% or more (18).

ENGINEERING DEVELOPMENT AND TESTING

BACKGROUND - Headlamps, like many vehiclecomponents, have evolved over a long period oftime. The history of this process has beenreviewed by a number of authors (19, 20, 21,22, 23, 24). A brief summary is presented here.

HEADLIGHTING IN THE UNITED STATES -Headlights first appeared about 1906. Thefirst headlights used acetylene and, like thefirst electric lamps which became availablelater, had a "beam pattern" much like a searchlight. About the time of World War I, aneffort was made to more adequately illuminatethe road by molding prisms into the lenses.Further developments continued in the 1920'sand early 1930's, resulting in substantialimprovements in light distribution intensity.

Unfortunately, this era was alsocharacterized by a proliferation of beampatterns, lamp sizes and shapes that not onlymade headlamp components expensive, but alsomade them difficult to replace when necessary.

In the middle 1930's work began toward thedevelopment of the sealed beam, which wasintroduced on 1940 model cars. The sealed beamis probably the most significant singledevelopment ever to occur in headlighting. Itsolved some serious problems associated withaging of the lamp unit, virtually guaranteeingconsistent, good quality headlightingthroughout the life of the vehicle. Headlampswere also standardized dimensionally as well,making them readily available at relatively lowcost.

Since the introduction of the sealed beam,some rather significant changes have beenintroduced to solve certain problems andimprove performance. For example, it was feltthat performance of the early lamps in rain,snow and fog was not as good as it might havebeen, due to illumination scattered abovehorizontal. In 1955 this scatter was reducedby the addition of a "fog cap" (a metal shieldpositioned near the filament).

Visual aiming of sealed beams was found tobe very difficult, due to the lack ofwell-defined edges in the pattern. Thisproblem was aided greatly by the introductionof mechanically aimable units in 1956.

In 1957 the first four-lamp system wasintroduced. By partially separating high andlow-beam functions, the need for compromise inlens design was reduced. Although in theorythe four-lamp system was developed to providedrivers with better illumination, no publishedresearch studies predating 1957 have been foundthat evaluated the visibility benefits of thissystem.

Further refinements in design resulted inan improved two-lamp system in 1959. Again, nocontrolled experimental studies have been foundthat evaluated the visibility performance ofthis development.

The 1970's saw the introduction ofrectangular two-and four-lamp systems. Theprimary benefit was in styling and the factthat they would permit better utilization ofthe space on the front of the vehicle.

Halogen technology was not applied tosealed beams until the late 1970's. The majorreason for this development was that FMVSS 108had been changed to allow a more powerful highbeam (from 75,000 cd to 150,000 cd), with thelimitation that power consumption had to staythe same. The low beam of these units have tomeet the same photometric requirements as theirtungsten counterparts.

HEADLIGHTING IN EUROPE - Europeanheadlighting systems are based on thedevelopment of the Graves "anti-dazzle" bulb,which was patented in 1920. The Graves bulbprovides a simple and inexpensive way ofgreatly reducing the amount of light projectedabove horizontal. A metal shield surrounds thefront, sides and bottom of the low beamfilament, preventing any light from beingprojected directly forward or to the lowerportion of the reflector. While this reducesthe efficiency of the lamp in terms of luminousflux/watt compared to the sealed beam system,it does produce significantly less glare.

In 1953-54 a number of lighting tests werecarried out under the auspices of the CIE(International Commission on Illumination).These tests have been described by de Boer (25,26) and Jones (27). As part of this program,comparisons were made between American andEuropean lighting systems. The resultssuggested that visibility distances on the leftside of the road were comparable under mostconditions tested. However, since the American

low beam was asymmetrical (i.e., it directedthe most intense portion of the beam to theright), it produced greater visibilitydistances on the right side of the road. As aresult, it was recommended that changes be madeto the Graves shield to allow a greater amountof light to be projected along the right edgeof the road. This was accomplished by removinga portion of the shield on one side. The sharpcut-off characteristic was retained. However,instead of presenting a symmetrical appearancewhen projected against a wall of screen, it nowappeared flat on the left with a 15 degreeupward slant on the right. This revisedconcept became the European standard.

More recently a further modification hastaken place, with the high intensity portionsabove horizontal being cut off at +1 degree.This produces a shape approximating the letter"Z," instead of a shallow "V." This changereduces problems with glare on curves and intothe rear view mirrors of vehicles ahead.

Another advance in European headlightingcame with the introduction of iodine (halogen)sources. The first mention of these in theliterature occurs in the early 1960's (28, 29),although they did not appear on cars untilsometime later.

The use of iodine vapor inside a lightbulb makes possible a chemical reaction thatcauses vaporized tungsten to redeposit on thefilament itself rather than on the glassenvelope (30). Thus, the problem of bulbblackening is eliminated. It also makes itpossible to generate substantially moreluminous flux/watt and to use a smallerfilament. The smaller filament simplifies theproblem of focusing the beam. Because thefilament must be run at a much highertemperature in order to operate the chemicalreaction just described, it was necessary touse a quartz envelope on the bulb. It is forthis reason that such sources are termedquartz-iodine or quartz-halogen.

Substantial development has taken place inthe last several years since halogen lamps werefirst introduced. Earlier versions could useonly a single filament in the bulb, making itapplicable only for four-lamp systems.Present-day versions incorporate two filaments,so that both high and low beams can begenerated from a single source.

One of the major differences betweenEuropean and U.S. lamps is in construction.European lamps are not hermetically sealed andare typically of composite construction, i.e.,a glass lens glued to a metal reflector.Unsealed and composite lamps can be designed inshapes and sizes that may give them fulleconomy and styling benefits. However,European construction can lead to two potentialproblems that the sealed beam does not have:(1) accumulation of dirt and moisture oninterior surfaces due to the "breathing" actionof such units, and (2) oxidation of thereflector. One of the best surveys of thisproblem (31) found significant distortion of

beam patterns in as little as six months in asample of European lamps they tested.European safety inspection data (as quoted byEhrhardt, 30) have shown serious deteriorationas a common reason for failures, but suchsurveys are unsystematic and do notquantitatively measure lamp performance. TheEuropeans have worked hard to solve these lampdurability problems, but data on newer headlampdesigns are scarce. Although some Europeanlamps may provide good long term performance,the current standards do not set requirementsthat prevent poorer quality lamps from beingsold. For example, they do not specifycorrosion performance tests.

Another major difference between Europeanand U.S. headlamps is the method of aim.European lamps are aimed visually by projectingthe beam on a screen directly or on a screenthrough a condensing lens. Because theEuropean low beam pattern has a relativelysharp intensity gradient, aim can be adjustedby positioning the lamp until the gradientfalls in line with the reference markings onthe screen. Aiming accuracy depends in part onhow precisely the beam pattern can be alignedwith the reference markings. Because the U.S.beam pattern does not have intensity gradientsas distinctive as those in European lamps,visual aiming is more difficult. Visual aimingis used in the U.S. although it has been shownto be less accurate than mechanical aiming(32). Mechanical aiming checks alignment byreference to the aiming pads built into theheadlamp lens. A vertical plane through thesurface of these pads forms the aiming plane.Mechanical aimers are used to align the aimingplane vertically and horizontally using spiritlevels and a split image/mirror system. Thisaiming method is not affected by beam patternintensity gradients.

Questions of aim and durability arebecoming increasingly important in the U.S. asthe impetus grows to improve aerodynamics andincrease styling options. These motivationshave been the major force behind theengineering developments that seek to developnew lamp shapes, sizes, aiming devices,construction, and materials. The safetyimplications of these possible changes are onlybeginning to be understood.

Durability may become a problem becauseplastic lamps are being proposed. This type oflamp needs a protective coating on the lens toprotect it from abrasion. A scratched lens canalter the beam pattern and increase glare tooncoming drivers. The degree to whichperformance Suffers depends on the coating used

Durability can also be affected if thereflector materials corrode or the inside ofthe lamp becomes dirty, which may be morelikely if replaceable bulb systems areintroduced. Problems with European typesystems have been noted above. New materialsand techniques used in lamp design can improvethis problem but may have other durabilityproblems. In particular, if a lens cracks,

drivers may balk at paying the cost of lampreplacement which may be relatively high forlimited production lamp shapes. Because, thereplaceable bulb will still put light on theroad, some drivers may not realize that thecracked lens allows dirt, moisture, andpossibly corrosion to reduce lamp performance.Another potential problem with some replaceablebulb systems is that photometries can becomedegraded if the filament is not alignedprecisely when a new bulb is installed. Thisproblem can be solved by having precisetolerances for filament position and bulb'mounting.

Maintaining correct aim may become moredifficult if new lamp designs do not haveaiming pads compatible with current mechanicalaimers. The quality of aim in the vehiclefleet will thus be dependent on the extent towhich service shops purchase and can properlyuse adapters for these lamps.

The use of headlamp covers is anotherapproach with the potential to improveaerodynamics. Testing of glass covers wasconducted by the California Highway Patrol andreported in comments to NHTSA Docket 81-11.Their testing found that the covers reducedmaximum low beam intensity from 10-23 percentand increased glare intensity abovehorizontal. Their conclusion was that thetypes of covers tested degrade headlampperformance and should be prohibited.

Another new development with potentialeffects on lamp performance is the lowering ofpassenger car lamp mounting heights. Theimpetus comes from the desire for internationalharmonization, since Europe allows a slightlylower mounting height, and the desire to lowerhood lines for reduced drag. Over the yearsminimum headlamp heights have been lowered from30-32 inches to 24 inches from the ground tothe center of the lens. In practice the rangeof current passenger car headlamp heights isfrom 24-28 inches. The effect of loweringmounting heights on seeing distance was studiedin a test conducted by Roper and Meese (33).Figure 6 shows the results, which apply tolamps meeting SAE specifications for low beamand a target 16 inches square with 7 percentreflectance. These data show that about 10feet of seeing distance are lost for each inchthe lamp height is lowered. This figure canvary, however, depending on such factors asdriver eye height, beam aim, and beam pattern.

Changes are also being suggested that havethe potential for upgrading low beamphotometries. For example, four lamp systemsare being proposed which use filaments parallelto the longitudinal axis and single functionlow beam units. This development is said toallow higher light output in the lower rightquadrant for visibility with the same or lessglare intensity in the upper left quadrant ascurrent U.S. low beams. Also, composite lampswith replaceable bulbs are being proposed whichcould permit lamp designers to develop separateleft and right-side beam patterns for improved

visibility wtth less glare. Not all theengineering changes claim they will increaseseeing distance. Some of the new lamp designsbeing proposed may be limited in terms of theextent to which their photometries can beimproved although all claim to meet currentstandards.

NO ATTENTION FACTOR APPLIED

2*00 2000 !600 1200 900 400 0 400 800 1200 1600DISTANCE BETWEEN APPROACHING CARS DISTANCE BEYOND

• •FEET- MEETING POINT

Fig. 6 - Effect of lamp mounting heighton visibility distance (33)

MOTORCYCLE HEADLAMPS - The main focus ofthis report is automobile headlighting.However, a great variety of special lightingsystems have been developed for certainapplications. Motorcycles are such a case, andare of interest here because they operate inthe same environment as automobiles. Thespecial problems to lighting design posed bymotorcycle use are:

a. The vehicle rolls and the lamp issteered in cornering, distorting thepatterns and changing its illuminationdistance.

b. There is usually only one lamp.c. Many bikes (especially smaller ones)

have limited electrical poweravailable.

The standards controlling motorcyclelighting are less restrictive than thosecontrolling automobile lighting. As a resultthere are many varieties of motorcycleheadlamps, making them potentially moreexpensive and difficult to find when areplacement is required. Many of these lamps,especially thoje designed for mid-size andsmaller motorcycles, have considerably lessoutput than a car headlamp.

This problem has been explored in detailin two recent NHTSA-sponsored studies (34,35). Among their findings was thatconsiderable improvement has resulted from theintroduction of halogen sealed beams.

ENGINEERING DEVELOPMENT - Lightingdevelopment work in the United States has beengenerally described by Kilgour (23). It is acomplex process, often initiated by a singleperson or company. This is typically followedby laboratory and field testing, using

procedures which are standardized on a company,not an industry basis. If modifications toindustry standards are sought, the problemmoves to an appropriate committee of theSociety of Automotive Engineers (SAE). In thecase of beam development, hardware is usuallyfabricated and appropriate demonstrations arestaged for committee members. Ultimately, whenit appears as though consensus is likely, theissue is balloted to the committee. Some lampmodifications may require changes to FederalMotor Vehicle Safety Standard 108 before theycan be implemented.

While this method of headlamp developmentis not "scientific" in the usual meaning of theterm, it cannot be said to have produced a poorlighting system. However, there is a problemin that the process is not documented in theopen literature. This make it difficult toreview and critique the rationale for proposedchanges and to quantify the improvements insafety that these changes may bring.

LABORATORY AND FIELD TESTING AND COMPUTERMODELING

In view of the difficulty of relatingheadlamp design factors to accidents and thelimited information on the safety relatedaspects of engineering developments over theyears, the most descriptive and detailedinformation on headlamp safety problems comesfrom experimental studies and computermodeling. The dependent variables used inthese tests and analytic analyses serve asproxy measures of the degree to which headlampcharacteristics can affect night driving safety.

VISIBILITY - The most basic function ofheadlamps is to provide drivers with thevisibility for vehicle control, for detectionand assimilation of road sign information, aswell as detection and identification of roadwayhazards (e.g., pedestrians, stalled vehicles,and pot holes). Most headlight visibilityresearch has been concerned with the roadwayhazard problem.

One way to evaluate the seriousness of theroad hazard problem is to compare vehiclestopping distance to driver seeing distance.If the seeing distance is greater than thestopping distance, then the driver would have apositive safety margin and should be able toavoid hitting the obstacle.

Many studies have been run to determineseeing distances and most have used a somewhatsimilar methodology. For example, the targetsare either small (0.5 meter square is typical)or man-size. They are painted a uniform darkgray (e.g., 7% reflective), or covered withdark-colored fabric. They are typically flat,but are occasionally three-dimensional. Thesubjects ride in an automobile that is drivenat a set speed on a straight, flat road.Subjects are typically under 30 years old,alert, and unfatigued. One or more targets areplaced at known positions along the right edgeof the road, and a glare source, representing

an oncoming car, is placed to the left. Thesubjects are instructed to indicate when eachtarget is detected, usually by pressing abutton. The results are reported in terms ofmean detection distances for the variousconditions of interest. Occasionally,distributional data are provided as well.

Table 3 is a summary listing oftwenty-five headlighting studies that have incommon the detection and/or identification oflarge and/or small, low-contrast targets placedon the right side of the road using standard

U.S. and European headlamps. Most used amethodology similar to that just described(called "standard" in the table). Those thatused a different method are briefly described.

The differences in visibility distanceamong studies are impressive. For example,with no opposing glare car, mean visibilitydistances to a large target range from 14meters [46.2 feet] (Johansson and Rumar, 45) to120 meters [396 feet] (Hemion, 43), adifference of about 8.5:1.

TABLE 3 - Means (X) And Standard Deviations (sd) of Visibility Distance to Low-Contrast TargetsReported In Twenty-Five Field Studies

AUTHOR(S)

Adler andLunenfeld (36)

Bhise, Farberand McMahan (37)

DeBoer (25)

DeBoer andMorasz (38)

Graf andKrebs (39)

Halstead-Nusaloch et al-

(40)

Helmers andRumar (Al)

Helmers andRumar (42)

Hemion (A3)

Hemion (44)

Johansson andRumar (45)

„ ,

Johansson et at*(46)

Kazenmaier (47)

Rumar (53) '

Rumar, Helmersand Thorell (54)

Schmidt-Clausen (55)

Yeatman (56)

Zechnall (57)

T

YEAR

1973

1976

1955

1956

1976

1979

1973

1974

1968

1973

1963

1963

_

1956

1970

1973

1982

1968

1963

i

METHOD

Standard

Standard

Standard

Standard

Non-alerted subjecta, normaldriving environment* Eye

Subjects drove car on publicroads, encountering"natural" targets atunexpected locations

Standard

Standard

Standard

Standard

Subjects drove toward man-sized obstacle - braked assoon as they cov'd see it.

Subject Car Stationary.Glare car moved towardsubjects, measured distancewhen target disappeared.

Standard

Standard

Standard

Standard (European Lamps)

Standard

Standard

LOW BEAM VISIBILITY DISTANCE (Meters)

NO GLARE

Large Tgt

X

107

35

30

120

122

14

51

sd

10

2.5

6

Small Tgt

X

61

58

65

80

36

75

70

25

no

45

125

65

69

95

sd

20

6

6

MAXIMUM GLARE

Large Tgt

X

76

107

sd

10

Small Tgt

X

65

53

48

60

23

50

40

100

38

50

sd

15

7

10

TABLE 3 - (-Continued)

AUTHOR(S)

Lindae (48)

Moore (20)

Mortimer andOlson (49)

Roper andHoward (SO)

_

Roper andMeeae (33)

Roper andScott (51)

Rumar (52)

1

YEAR

1

1962L

19581

1974

1938

1963

1939

1965

METHOD

'

StandardL t

Standardf

Standard method, exceptsubjects had to identifyorientation of target.

r TDriver approached unex- 20pected man-sized target in[mphcenter of the lane. Measureddistance from target when! 50subject lifted foot from Imphaccelerator.

_[

Standard

Standard

Standard

LOU BEAM VISIBILITY DISTANCE (Meters)

NO GLARE

Large Tgt

X

53

—

30

69

sd

Small Tgt

X

1OO

61

88

145

45

sd

.

8.6

MAXIMUM GLARE

Large Tgt

X

55

sd

__ — _-J

Small Tgt

X

60

50

70

114

.]38

sd

7.3

Some of the reasons for these differencesin mean visibility distances are known. Forexample, the studies which report the shortestdistances used methods which introduceduncertainty about target position (e.g.,Johansson and Rumar. 45), the nature of thetarget (e.g., Halstead-Nussloch et al., 40), ordeceived the subject about the nature of thetest in an effort to determine the visibilitydistance of unalerted subjects (e.g., Roper andHoward, 50; and Graf and Krebs, 39).

In some cases mean visibility distancesvary from study to study because the testconditions vary. For example, targets may havedifferent shapes, reflectances and locations.Beam intensities and patterns may vary, andsubjects may have different visualcapabilities. The ambient lighting and targetbackground may not be the same from study tostudy. All of these factors can affect thevisibility distances measured.

Sometimes the reason for the differencesamong studies cannot be determined becauseimportant information is often omitted. Forexample, many investigators do not providephotometric data on the headlamps, which iscritical because of the manufacturing anddesign variability from lamp to lamp.Background reflectivity, which is as importantin detection as target reflectivity, is rarelymeasured.

Campbell (58) has calculated automobilestopping distances for level roads, and brakingin the skid mode. Figure 7 shows thesecalculated stopping distances as a function ofinitial velocity, two road frictions, and twodriver reaction times. These curves can beconsidered to be conservative, minimumestimates of stopping distance.

By comparing the range of seeing distancesto the estimated stopping distances, themaximum safe speed can be determined. Becauseof the variability in both seeing distance andstopping distance, the maximum safe speed canvary from about 20 mph to about 55 mph, underoptimum conditions without an oncoming glarecar. With an opposing glare source seeingdistances are less, and maximum safe speedswould be correspondingly lower. Thus, sincemany of the seeing distances and stoppingdistances were conservative estimates and didnot account for the limit conditions associatedwith many accidents, one can conclude thatunder real world conditions many drivers cannotsee far enough on low beams to stop and preventa collision with a roadway hazard.

In order to determine ways to reduce thisnegative safety margin, researchers haveconducted many tests to determine the extent towhich various factors increase detectiondistance. One of the key factors in lampdesign that can be changed to increasevisibility distance is beam intensity

BEAM INTENSITY - Data provided by Roperand Howard (50) serve as an illustration of therelationship between beam intensity andvisibility distance for three levels of targetreflectivity. (See Figure 8.) These data arefor a pedestrian target, no opposing glare car,unalerted subjects and car speed of 50 mph. Asa reference, current U.S. low beams woulddirect about 18-30,000 candlepower toward aright side target 200 feet down the road; highbeams, about 60,000 to 120,000.

The data indicates that gains invisibility distance associated with increasinglamp output are relatively rapid up to 25,000candlepower, and flatten out considerably after

11

200 300

STOPPING DISTANCE IN FEET

Fig. 7 - Automobile stopping distances(58)

100 125 150BEAM CANOLEPOWER

Fig. 8 - Visibility distance as afunction of beam candlepower andtarget reflectivity (50)(Candlepower values are X 1,000)(Visibility distances are X 10feet)

that. Obviously, it is not easy to obtaingreat improvements in seeing distance over whatis provided today simply by using more powerfullamps, even ignoring the problem of glare. Forexample, the recent increase in output allowedfor high beams (from 75,000 cd to 150,000 cdmax.) resulted in about a 20% improvement invisibility distance, according to Figure 8.Some of the reasons for the relatively poorgain in visibility per unit increase inintensity are as follows:

First, illumination reaching a surfacevaries as the square of the distance from thesource. That is, the illumination reaching asurface is equal to the luminous intensity ofthe source directed toward that surface dividedby the square of the distance. This means, forexample, that to double the distance between asource and surface yet maintain' the latter atthe same level of illumination, the sourceintensity will have to be increased fourfold.

Second, the key to target detection is notsimply target luminance, but the contrastprovide by the target and its background.Typically, the headlamps are illuminating bothsurfaces. If the source is moved further away,but its output is increased to provide the sametarget luminance, contrast will actuallydecrease because the background is nowrelatively closer to the target. Thus, if thetarget was at threshold before, it becomessub-threshold under the revised conditions.Required threshold contrast varies inverselywith level of illumination, so the onlysolution is to increase source intensityfurther to a point where the reduced contrastwill be at threshold again.

Third, the atmosphere has a significantscattering effect. Except for extremeconditions such as fog, this problem haslargely been ignored until recently. However,measurements reported by Huculak (59) show thatatmospheric scattering can be quitesignificant. Huculak offers one example inwhich the retroreflectance of the road surface

12

at about 200 meters appeared to be 50% higherthan it actually was, due to atmosphericscatter. And this was reported to be a clearnight. What this means is that if we use morepowerful lamps to try to see further,atmospheric scattering works against us byreducing the contrast of more distant targets.

In sum, a great deal of evidence aboutvisibility leads to the following conclusions:

a. When considering low-contrast targetssuch as dark-clad pedestrians, animalsand unlighted vehicles, present-daylow beams headlights do not offersufficient visibility for safeoperation at the national limit (55mph or about 90 km/hr). There is somedebate about the degree of shortfall,but under real world conditions themaximum safe speed is probably nohigher than 70 km/hr (45 mph).

b. The output of present-day low beamheadlamps is at the point that largeincreases in visibility distancesrequire very great increases inintensity. The main restriction tofurther increases is glare.

GLARE CONTROL - In the presence ofoncoming vehicle glare, visibility distance canbe reduced significantly compared to open roaddriving. Table 3 previously shown, lists theseeing distances with oncoming vehicles undermaximum glare conditions and shows thereduction in distance compared to the no glarecondition. The negative safety margin forunopposed driving is even lower in the presenceof oncoming traffic. Glare that reducesvisibility distance is called disabilityglare. Glare can also have the effect ofcausing discomfort to. the driver withoutreducing visibility. This is called discomfortor psychological glare. If low beam intensitycould be increased without increasing glare,then seeing distance could be increased, atleast a little. Efforts to improve seeingdistance by controlling glare have concentratedon beam pattern shaping and polarizedheadlighting.

The major beam patterns that have beenstudied and compared have been the U.S.tungsten sealed beam pattern and the Europeansharp cut-off pattern. There is no clearevidence that one pattern is any better thanthe other. The results depend on thecircumstances under which the beams wereevaluated, the evaluation criteria selected,and the particular lamps that were chosen to berepresentative of the beam pattern. Forexample, U.S. lamps typically show higherseeing distances for right side targets;European beams because of their reduced glareshow longer seeing distances for left sidetargets. Furthermore, a beam pattern with asharp intensity gradient like the European beamthat may provide good performance on flatstraight roads, may not provide as qood avisibility distance on curves, hi l l s , and withless than perfect aim. Another caution to note

in making such comparisons is that the seeingdistances were usually measured with the glarelamps the same as the visibility lamps.Different results would be obtained if unlikebeams were used. Thus, the driver of a carwith European lamps meeting a car with U.S. lowbeams might not have seeing distances as goodas the opposing driver using a U.S. beam.

Because of the difficulty of substantiallyimproving seeing distance with conventionalsystems, polarized headlighting wasinvestigated in a number of studies. In thissystem polarized filters are placed in front ofthe headlamps and in front of the driver'seyes. The axes of polarization are oriented sothat when the driver looks through the filter,light from the opposing lamps is blocked out.Because of the significant glare reductionpossible with the system, it permits extremelamp intensities and reduces the need for highstandards of aim and maintenance. However,because of concerns over cost and proper use ofthe system by drivers, it was never implemented.

Another approach to lengthen seeingdistance through beam pattern changes, was thethree-beam system. The third beam was designedto increase seeing distance in certain meetingsituations. One study found that targetdetection distance was extended 29 percentfurther in two-lane meeting situations thanstandard low beams. (60). However, the mid-beam was too bright to be used in meetings oncurves and following closely or overtaking andpassing preceeding cars. Because of concernover proper driver use of the beams and thecriticality of proper aim for the mid beamlamp, this system was never implemented.

Ongoing research at the Unversity ofMichigan is investigating whether moreconventional approaches to low beam patterndesign can result in increases in seeingdistance in meeting situations. The extent ofthe increase possible depends on thelimitations imposed by discomfort anddisability glare.

DISABILITY GLARE - Light entering the eyeis to some extent scattered by the optic mediaso that random illumination is superimposedover the image of both target and background.As a result, background luminance iseffectively increased and contrast is reduced(61). The effect has been shown to be the sameas that produced by supplying veiling luminancedirectly over the target and background (62,63).

A systematic evaluation of therelationship between illuminating and glaringintensities and seeing distance has beenprovided by Grime (64). His data arereproduced in Figure 9. To provide a frame ofreference, present U.S. low beams, in thesituation described for Figure Q, would produceabout 18-30,000 cd of illumination towards Aright side target 200 feet down the road andabout 1500 cd or less of glare towards anopposing car about 200 foot away.

Beam Intensityon Object—

Candle-Power

3,000 4,000

Fig. 9 -Glare Intensity—Candle-Power

The relation between seeingdistance, beam intensitydirected at an object andglaring intensity (64) [Objectposition 10 feet to side ofopposing lamp and 10 feet behindit; object reflectance, 7percent; height of object, 18inches; car speed, 30 mph]

The main point from Figure 9 is that thedisabling effects of glare are ndn-linear. Lowlevels of glare produce more degradation perunit increase than higher levels.Extrapolating from the relationships shown inthis figure indicate that doubling current lowbeam intensity towards the target increasesseeing distance by about the same percentage asreducing glare by 50%. The advantage ofreducing glare is that discomfort is reduced.However, this approach requires a very sharpintensity gradient that may be difficult toachieve in practice. The European beampatterns ire designed to produce lower glarebut in doing so they sacrifice intensitytowards roadside objects and also lampefficiency. The U.S. sealed beams rely on lensprescription changes to control the beampatterns but this approach cannot produce sharpintensity gradients. Hence, glare levelscannot be easily reduced without loweringoverall output.

There are also problems with the approachof increasing intensity to increase seeingdistance. The effects of glare as representedin Figure 9 assume straight, flat road meetingsituations. Because of misaim and various roadgeometries, the eyes of oncoming drivers aresometimes exposed to portions of the beam otherthan the "design" glare zone. Being hitsuddenly with the high intensity portion of thelow beam pattern when cresting a hill orrounding a curve is an unpleasant experience atbest. It can lead to severe, if brief, lossesin visibility. More intense lights wouldexacerbate this problem.

The solution to these seeminglyirreconcilable tradeoffs may be to selectively

increase intensity in the direction of hazardson the'right side of the road. Glareintensities directed at oncoming drivers mayneed to be raised slightly if conventionalsealed beam technology is used. However, newtechnology (e.g., single-function,axial-filament low beams) may allow glareintensities to be lowered. The key todetermining how much intensity can beselectively increased is the discomfort glareacceptable to drivers.

Discomfort Glare. It is common knowledgethat glaring lights can produce a sensation ofdiscomfort. While the sensation is realenough, the mechanism(s) involved are notknown. Lacking an understanding of whatphysical responses are involved, it has beennecessary to approach discomfort glare as aproblem in subjective scaling.

Most headlight discomfort evaluations haveemployed the following 9-point scale describedby DeBoer (1965):

unbearable1.2.3.4.5.6.7.8.9.

disturbing

just acceptable

satisfactory

just noticeableThe DeBoer scale has been used by a number

of different investigators. Of specialinterest is the meticulous laboratory workreported by Schmidt-Clausen and Bindels (66).Their results are reproduced in Figure 10.Based on these data, the peak glaringintensities of U.S. low beams in a meet on astraight-flat road would be rated about "4,"suggesting that a further increase in glare isundesirable. Some verification of this wasobtained by Bhise et al. (10), based on adimming request study. Their results arereproduced in Figure 11, and show that requestsfor dimming began increasing rapidly when theglare experienced by the opposing driver wassuch as to produce a rating of "4" or less.

Mortimer and Olson (67) also used theDeBoer scale to obtain glare ratings forseveral lighting systems in a full-scalemeeting situation. Their data are reproducedin Figure 12. The considerable effect of glareangle, as determined by where the subjects weresearching for targets, should be noted. Thedata also suggest that ratings obtained in arealistic meeting situation may be differentfrom those derived from laboratory studies.Assuming the driver chooses to look" ahead or tothe right rather than at the approachingheadlamps, it appears that U.S. low beams couldbe rated "5" or better, rather than the "4"predicted by the DeBoer equation.

A problem with ratings of the typedescribed has been pointed out by Lulla andBennett (68). They note that the scale valuesassigned a given glare level depend to some

14

10

O)S 5acc.

\

\— 3 — 2 — 1 0 1 2

Log glare illuminance log Eg (lux)

Fig. 10 - DeBoer ratings (W) as a functionof glare illuminance (66)

NotNoticeable 9 • •

5C3

B

isto

Satisfactory

JustAcceptable

Disturbing

Intolerable

5 • •

3 • •

2 - •

LeftTarget

-f-U.S.Low

Mid-A Mid-B

Beam

ECE ECE-U.SLow Mid

Fig. 12 - Mean ratings of maximum glarediscomfort in meetings with fivebeams when searching for targetson the right and left side ofthe lane (67)

1001

80

60

S 40

20

t6 -

105,000 CP

1

Unbearable Disturbing jus! AcceptaDie

Discomfort Glare Index (W)

Fig. 11 - Request for dimming fromapproaching drivers as afunction of calculated glarediscomfort rating (10)

extent on the range of stimuli presented. Thissuggests that: (1) the ratings cannot betreated as absolutes, and (2) if thecomposition of glare to which people areexposed in everyday driving changes, theirjudgment of what is "just acceptable" willchange as well. Some additional work on thisphenomenon is currently being carried out atthe University of Michigan.

This work was carried out in two stages.Stage 1 was a laboratory study very much likethat of Schmidt-Clausen and Bindels. Theresults were identical. Stage 2 used thelaboratory approach and rating scale, but wascarried out in on a roadway with automobiles.The stage 2 results are shown in Figure 13. Itis apparent from Figure 13 that the subjects inthe field setting rated the same glare levelsmore comfortable than did the subjects in thelaboratory setting (DeBoer prediction).

Discomfort glare is one of the principallimiting factors in attempting to gain improvedvisibility by increasing lamp output. Thereare only a few rigorous studies of discomfortglare in a setting applicable to headlighting.

15

&3

FULL RANGEPARTIAL RANGE

-3 -Z -I 0LOG LUX

Fig. 13 - Comparison of mean ratingsassigned each glare condition inthe field compared withprediction based on use ofDeBoer equation (Olson andSivak, unpublished)

Unfortunately, as just noted, the results arenot in agreement. Given the critical nature ofthis work to headlamp design, further work isdesirable to arrive at more definitiveconclusions.

Complicating the glare issue, both fromthe point of disability and discomfort, is thefact that it is experienced not only fromvehicles in the forward field, but fromfollowing vehicles, via the rearview mirrors.The problem of rear view mirror glare has beenstudied in some depth (69, 70, 71). Brieflysummarized, the findings are that both kinds ofglare are a problem with the present U.S.low-beam system. Obviously, any increases inintensity will worsen the rear view mirrorproblem. However, Olson and Sivak (71) arguethat this is a problem for which there is apotential solution, namely lower reflectivityfor interior and exterior mirrors. However,there is some uncertainty about what thisreflectivity should be, especially for outsidemirrors.

CONDITIONS DEGRADING LIGHTING SYSTEMPERFORMANCE - Headlamps are designed to providecertain performance levels. Since they areimportant safety systems, they should come asclose to the design performance as possibleunder normal operating conditions.

Some indication of .how well lightingsystems actually function under real-worldconditions has been provided by Yerrell (72).In this study measurements were taken of theglare and, illuminating intensities of a largenumber of passing vehicles at several sites inGreat Britain. These data are reproduced inFigures 14 and 15.

4 6

Illumination intensity (cd)

8 x 1Q3

Fig. 14 - Frequency distribution ofheadlamp illuminationintensities in Great Britain(72)

100

6 8 10 12 14

'Glare' intensity (cd)

16 x 10'

Fiq. 15 - Frequency distribution of glareintensities in Great Britain(72)

16

Based on*1dealized British isocandeladiagrams, and the photometric target placementsused by Yerrell, the "expected" glare andillumination values were about 1600 and 4000cd, respectively. The 4000 cd illuminationvalue falls slightly below the median in Figure14, indicating it was met by fewer than half ofthe vehicles in the sample. About 10% of themeasured illuminating intensities wereone-quarter or less of the expected value.Glare, as shown in Figure 15, was also quitevariable, although most measurements (about85%) were less than the expected value.

Thus, the available evidence indicatesthat field headlamp performance varies greatlyand that a large fraction of vehicles (about25%, according to Yerrell) are running withilluminating intensities about one-half or lessof levels expected based on design criteria.

The reasons for the state of affairsillustrated in Figures 14 and 15 are many andprobably not fully known. But, research hasbeen carried out to identify some of the moresignificant ones. In the next portion of thisreport these factors will be reviewed, togetherwith other variables affecting systemperformance not considered in the Yerrell study.

Aim - The first systematic work on the aimproblem was carried out at the Transportationand Road Research Laboratory in Great Britain(73, 74, 75, 76). They identified the majorsources of misaim as improper setting byservice facilities, changes in the adjustingmechanism, and changes in vehicle attitude dueto load.

With the implementation of vehicleinspection programs in several states, headlampaim was noted as the most common "defect," andmisaims far in excess of accepted standardswere commonly found (77, 78, 79, 80).

Hull et al. (32) measured the extent ofmisaim in the motor vehicle population bysurveying cars in Texas, Missouri, and Kansas.Figure 16 shows low beam vertical aim forTexas, a State with vehicle inspection. Figure17 shows low beam vertical aim for Kansas andMissouri, states without lamp inspection. As abaseline, SAE standards specify limits of ̂ 4inches at 25 feet. One degree equals 5.2inches at 25 feet. About 68% of the lamps werewithin the SAE limits for low beam vertical aimin Texas compared to 59% in the other twoStates.

Olson and Mortimer (81) surveyed theliterature to identify the major sources ofmisaim and estimate their magnitude.Disregarding vehicle attitude changes, the dataindicate that a large fraction of the cars onthe road at any given time can be expected tohave one or more lamps significantly misaimed.The sources of aim error are summarized inTable 4. Adding to these sources the problemof vehicle load effects, the misaim problem isseen to be very significant.

5"iioS•- •Si- «

i.w&

2

LOWED SEAMVERTICAL

-• -4 -t 0 I 4 « t 10INCHES DOWN (-1 OH UP AT 29 FT

Fig. 16 - Distribution of headlight aim ina vehicle inspection state(Texas) (32)

I"

LOWER BEAMVERTICAL

MTh^-• -• -4 -2 0 2 4 • • 10 12

IMCHU DOWN I-) OR UP AT 23 FT

Fig. 17 - Distribution of headlight aim innon inspection states (Kansasand Missouri) (32)

TABLE 4 - Headlamp Aim Variances From AllSignificant Sources Excluding Load(81)

SOURCE

LampMlsorient aiming planeDifference between beam

and mounting planeChange in use

VehicleLamp mountingDog tracking*Matchboxing**

AimingLong axis alignment

PhotometricVisual

AimingVisual screenVisual screenPhotometricMechanical

Human Factors

VARIANCE (Degrees)

Horizontal Vertical

.05

.92

.77

.12

.19

.11

.521.06

.29

.96

.04

.02

.58

.05

.92

.77

.38

.08

.29

.02

.04

.58

*Condition when rear axle is not perpendicular to the longaxis of the car.

**Condition when headlamps are not positioned in a lineperpendicular to the long axis of the vehicle.

17

Relatively minor discrepancies betweenwhere a lamp is actually aimed and where it wasintended to be aimed can produce significantperformance changes. By way of illustration,consider Table 5, taken from Hull et al. (32).These data show target detection distancesunder various beam and misaim conditions.Changes of as little as one degree can resultin very large changes in visibility, either dueto altered illumination or glare.

Detailed measurements of the effect ofdirt on photometric performance have beenreported by Cox (84). Data from two lampsmeasured by Cox are shown in Figure 19. Thetop one was pre-judged "moderately dirty," thebottom one "very dirty." The moderately dirtycondition reduced output on both beams by about50%. According to Rumar, this is about theleast reduction which the drivers can beexpected to notice.

TABLE 5 - Visibility Distance As A Function OfAim On Observer And Meeting Cars (32)

TestNo.

1

2

3

4

5678 .9

1011

Observer's CarLamp Aim

Low beam aimedcorrectlyLow beam aimedcorrectlyLow beam aimedcorrectlyLow beam aimedcorrectlyL.B.1°D-VL.B.1°D-VL.B.1°U-VL.B.1°U-VH.B. aimedcorrectlyH.B.1-1/2°U-VH.B.1-1/2°D-V

Glare CarLamp Aim

No glare care-

Low beam aimedcorrectly..

L.B.1°U-1°R

L.B . 1 -1 /2^U-1 -1 /2°L

l.B. correctNo glare carL.B.1°U-VNo glare carNo glare car

No g8are carNo glare car

Target DetectionDistance, ft.

397

345

265

250

208219444697729

642594

•Location referenced to "correct" aim position.

For some misaim problems there areavailable solutions. For example, vehicleattitude can be controlled with several typesof self leveling devices ( 82). Other sourcesof aim error are more difficult to solve.

Dirt - As headlamps accumulate dirt,intensity and seeing distance become degraded.Some indication of the extent of this problemin Sweden has been provided by Rumar (83),whose data are reproduced in Figure 18.

20 40 60 80LIGHT REDUCTION (%)

Fig. 18 - Proportion of cars at gasstations having various degreesof light reduction in thecentral part of the high beamcaused by dirt under three roadconditions (83)

Tran

sver

se d

islrib

ulio

n o

f int

ensi

lM

ambe

am

•^.-^

/

/.

///

s

V 40.000 cc

\

\20.C

\\xX.

100 CO

Clean — .Dirry —

^^15' 10° 5' Let! 0« Right 5'

§ 20000I

5 10000

_ — -

— '

1//

V30.000 cc

\\

20.000 cc

CleanDuly — -

v~^^M

5- Leu 0- H.gt« 5* 10- 15-

Fig. 19 - Measurements of attenuation dueto dirt, taken in the horizontalplane for a lamp judgedmoderately dirty (top) and verydirty (bottom) (84)

Some indication of what various levels ofattenuation mean in terms of loss of visibilityhas been provided by Rumar (83), whose data arereproduced ^n Figure 20. It will be noted thatlight dirt layers can actually increasevisibility distance slightly. This is due toscattering that causes additional illuminationto be directed upward. Rumar's data indicatethat a 50% reduction in output would reducevisibility distance by 10-15%.

Based on .the information described, dirtlooms as a significant factor in reducedheadlamp performance. Several lampwashing/wiping systems have been suggested assolutions and have been marketed in othercountries. In the U.S., the use of wiper

18

systems is limited because Federal regulationsprohibit wiper blades from obstructing lightfrom the lamp. The aiming pads on U.S. lampsmake it difficult for wiper blades to "park"off the lens. Further study to develop themost cost-effective option is needed.

zo

20oLUsr

CDCO

Fig.

OPPOSINGLOWBEAM >'

BEAM

- 60LIGHT REDUCTION (%)

8O

20 - Visibility reduction caused bylight attenuation due to dirtfor three conditions obtained incontrolled experiments (83)

Weather - Headlamps must be designed to beminimally affected by adverse atmosphericconditions. Conditions suoh as fog, snow andrain cause the illumination projected from aheadlamp to be scattered. This reduces theillumination actually reaching a target ofinterest. Some of the illumination isscattered upward, causing the atmospherebetween the driver and target to light up, withthe result that target contrast is reduced (7).

There have been some efforts to quantifythe effects of rain and fog (7, 85, 10) with aview toward designing lighting systems to bemore effective in adverse weather. The resultsindicate that performance is improved by usinglamps which project relatively littleillumination above horizontal, and mountingthem as low as possible. Performance is notimproved by tinting the lens a yellow color.

COMPUTER MODELS - Head lighting is acomplex topic, and full-scale lighting studiesare expensive to set up and run. The resultsof these experimental studies are onlyapplicable to the limited conditions of thetest. It is thus difficult to evaluatelighting system parameters systematically. Forexample, experimental results measured understraight, level road meeting situations may notprovide a realistic evaluation of performanceon curves or when misaimed. Because computermodels permit lamp performance to be readilyevaluated under many possible conditions, theyare valuable tools in lighting research.

The first serious effort to develop acomputer seeing distance model was that of Jehu(86, 87). This model was written first tosimulate a meeting of two cars on astraight-flat road, and then extended toconsider curves. The model was used toevaluate various lighting systems in use atthat time.

An extensive research effort wasundertaken at the University of Michigan in theearly 1970's which led to the development of acomputer model that computes seeing distance.The model was developed based on extensivefield tests (49) and originally applied tomeeting situations on flat-straight roads(88). The model was subsequently expanded toinclude hills and curves (89).

The Unversity of Michigan model has beenused extensively over the years for evaluationof alternative lighting systems (90), and as ameans of screening candidate lighting systemsfor later field evaluation (40, 34, 35).

One problem that was addressed recentlyusing the U of M model is lamp mounting height(71). Current regulations allow mountingheights to vary from a minimum of 24 inches toa maximum of 54 inches. There was concern thatthe greater heights, common on trucks, causedexcessive glare for drivers.

In dealing with this question, the modelwas run to simulate a number of possiblemeeting situations using an experimental, highintensity low beam. Some of the data whichresulted are shown in Table 6. This tableindicates the gains in visibility distanceassociated with greater mounting heights, aswell as disability and discomfort glareeffects. Clearly, greater mounting heightsprovide both more visibility and glare. Basedon this and other analysis, it was concludedthat reducing the upper limit on mountingheight is unwarranted.

TABLE 6 - Effects Of Lamp Mounting Height OnVisibility And Glare (71)

Mounting Heights(Inches)

24

40

54

VISIBILITY DISTANCE

Maximum(No Gl«re)

292

315

338

Minimum(Max Glare)

259

253

249

DISCOMFORT GLARE

Maximun.Rating*

4.7

4.5

4.4

Distance at5.0 or worse

(feet)

550

600

650

•Based on a 9-po1nt scale where 5 • just acceptable, 1 • unbearable, and9 • just noticeable.

While computer models of the typedescribed above are a considerable help inevaluating certain aspects of beam performance,they are limited in the number of conditions(target type, road topography) that can be

19

considered at one time. Since lamp performancevaries as a function of conditions tested, theinvestigator still has the problem of weightingoften conflicting results to arrive at ameaningful assessment of performance.

Investigators at Ford set out to produce acomputer model which would provide acomprehensive evaluation. This work has beensummarized by Bhise et al. (10). As a firststep the investigators developed a seeingdistance model, something like the others whichhave been described. A simulated standardizedtest route was then assembled. This routeconsists of a series of highway sections in theform of environmental parameters which arethought to have an influence on visualperformance while driving at night. Itincludes such factors as pavement, lane lineand target reflectance, road geometry, laneconfiguration, ambient illumination, and glarefrom fixed lighting and%traffic.

When headlighting systems are run in themodel, the output is a figure of merit. Thisfigure of merit is the percentage of thedistance travelled on the standardized testroute during which the seeing distance topedestrians and pavement lines and thediscomfort glare levels experienced by opposingdrivers simultaneously meet certain acceptancecriteria.

In one analysis using the model, a numberof different headlamp systems were compared,including 5-3/4 inch round, 7-inch round, and4-x 6-1/2 inch rectangular US systems; andEuropean Bilux 2-lamp, and Halogen H4, 2-lampsystems. The figures of merit were 74.1, 74.0,73.9, 67.3, and 71.4, respectively. The lowerratings for the European systems wereattributed to their lower intensity low beamand poorer performance in wet road driving andat speeds over 40 mph. However, the Europeansystem discomforted from about 2-6% of thedrivers compared to about 10% for the U.S.beams.

Other analyses conducted with the Fordmodel have shown that the figure of merit ismuch more sensitive to environmental variablesthan headlamp factors. The largest differencebetween two headlamps had been about 10percentage points, while differentenvironmental conditions (e.g., road type, roadsurface condition) have produced differences aslarge as 60 percentage points. This impliesthat changing the beam patterns will not haveas much effect on system performance aschanging roadway environment factors. However,to a large extent this approach is notpractical, since certain key variables (e.g.,road geometry) cannot really be changed.

Another implication of the model is thatthe intensity of the current U.S. low beamappears to be close to optimum. Some beampatterns did show slight increases in thefigure of merit over U.S. low beams. However,it is not clear how much weight to attach tothe relatively small differences that are foundbecause the relationship of the figure of merit

to real-world accidents is unknown. Thus, tounderstand the meaning of the figure of merit,it is important to understand the performancecriteria built into the model. For example,pedestrian visibility is determined for"unalerted drivers," and discomfort glare isacceptable if the DeBoer glare index is greaterthan 4 or if it is less than the discomfortglare of a U.S. low beam. Greater figures ofmerit and larger differences between beampatterns could be achieved if, for example,some pedestrian encounters were with alerteddrivers and greater discomfort glare could betolerated by drivers.

The Ford model is the only approachpresently available for evaluating headlightsfrom a systems point of view. It can be ofgreat help in serving as a standardized meansof evaluating the relative consequences ofchanges in headlight parameters. What isneeded is general agreement among the lightingresearch community on the logic and criteriaused in calculating the figure of merit.

SUMMARY

Driver recognition and detection errorsoften occur at night because of the obviousvisibility degradation that occurs indarkness. Although present-day low beams areadequate for many tasks involving the detectionof high contrast objects, they are less thanadequate in revealing low contrast objects.The available illumination may not besufficient to permit drivers to see obstacles,signs, and the direction of the road. Theextent of this insufficiency cannot.be assesseddirectly but has been inferred from accidentstatistics, driver reports, experimentalstudies, and analytical evaluations.

In examining the broad accidentstatistics, nighttime accident rates are foundto be three times the day rate. Although someof this difference can be attributed to greateralcohol involvement, fatigue, and young driverproblems, inadequate lighting can be assumed toplay a role as well. Even drunk and fatigueddrivers might benefit from improvedvisibility. A number of studies have shownthat improved roadway illumination using fixedlighting reduces accidents. However, for themajority of vehicle miles in this country, theonly practical means of increasing roadwayillumination is to improve headlight systems.

One way to quantify the safety margindrivers have at night is to compare thevisibility afforded by headlights with thestopping distance of the car. The measuredseeing distance with low beam headlamps variesas a function of many factors, including driverage, target characteristics, beam intensity,and test methodology. For pedestrian typetargets, seeing distance on typical U.S. lowbeams has been shown to vary from about 30meters (99 ft) to 120 meters (396 ft). Inmeeting situations, the seeing distance rangesfrom about 25 to 80 meters (83-264 feet).

20

The maximum safe speeds for these stoppingdistances can be calculated for variouscombinations of road friction and assumeddriver perception reaction time. For example,on a dry pavement the maximum speed for a 396foot stopping distance is about 55 mph,assuming a 2.5 second driver reaction time. Inopposing traffic the 264 foot stopping, distancecorresponds to about a 42 mph safe speed.Since these maximum visibility distances arelikely to be considerably less in the realworld, the margin of safety for many driverswill be negative. For example, the 83 footvisibility distance measured in some teststranslates to about 28 mph under the most idealdriving conditions (i.e., dry pavement and onesecond driver perception reaction time).

The problem of nighttime driving has beenrecognized for many years by researchers whohave sought to develop improved headlightingsystems. Progress has been slow because of thedifficulty of resolving the many tradeoffsinvolved with headlight design. The primarytradeoff involves illuminating intensity versusglare intensity. Seeing distance in open-roaddriving can be increased somewhat by simplyincreasing the lamp intensity; but in vehiclemeeting situations the number of driverscomplaining about glare would also increase andseeing distance would be impaired. Thepolarized headlighting system had been proposedas a solution to this tradeoff. However, costand implementation problems caused researchersto seek simpler and more practical approaches.

The research on improving the low beam hasconcentrated on how to change beam intensityand distribution. Such an .approach has shownthat only relatively small increases invisibility can be made because of thelimitations of current lamp design andmanufacturing techniques and because of thelaws of physics and response of the humanvisual system. However, even small increasesin visibility are important to help counter themany environmental, driver, and vehicle factorsthat degrade low beam performance.

Some of the vehicle design factors thatcan reduce driver visibility have become moresignificant in recent years because of basicchanges in the characteristics of the vehiclefleet. Although basic low beam technology hasnot changed significantly since the 1950's, thevehicle fleet has changed in a number of waysthat can reduce headlamp performance including1) increased use of tinted windshields, 2)lower lamp mounting height in passenger cars,and 3) lower driver eye height. In addition,the increased percentage of smaller cars raisesthe question of whether there is increased lampsusceptibility to misaim due to vehicleloading. Currently, a number of proposals havebeen advanced which would significantly changethe technology of lamp construction. Whilethese technologies may offer some styling andfuel economy benefits to consumers, their longterm road illumination performance is uncertain.

Headlamp performance is also affected bythe extent and nature of the limit conditionsimposed by various environmental and driverfactors. For example, weather and roadtopography have been shown to influenceheadlamp illumination. Also, the increasingproportion of older drivers could furtherreduce the level of visibility provided byheadlamps.

While each of the above factors may have arelatively small effect on nighttime visibilitywhen considered individually, the combinedaction of all of them suggest that the need tomaintain and improve low beam performance ismore critical than ever.

One of the ways to evaluate the effects ofthese conditions on headlamp performance isthrough the use .of a figure of merit such asthat developed by Bhise et al. The figure ofmerit provides a systematic and comprehensiveevaluation of the many interacting factorsinfluencing night driving safety.

One of the factors affecting the figure ofmerit is discomfort glare. Discomfort glare isan issue which divides American and Europeanlighting engineers. There are relatively fewstudies on discomfort glare. Some of these,especially those conducted in laboratorysettings, suggest we are at the limit forglare. Others, especially those conducted infield settings, indicate further increase inglare may be possible. And there are otherdata that indicate that people's judgment ofwhat is objectionable varies depending on thecontext in which the judgment is made.

Other components of the figure of merit areperformance degrading factors such as misaimand dirt. However, existing data on the extentand nature of these factors is not up to datefor the current U.S. vehicle fleet.

AN AGENDA FOR FUTURE ATTENTION - Theinformation discussed in this literature reviewsuggests a number of problem areas that needfurther attention:

1) Because headlamp performance has beenshown to vary as a function of testconditions, a standardized approach toevaluation of headlamp performanceneeds to be developed. This approachshould be based on the performance ofheadlamps under the hazardousconditions that can lead toaccidents. The figure of merittechnique used in the Ford model canbe employed as a basis for suchevaluations. A figure of merit forheadlamp system performance can beused to evaluate the safetyimplications of changes to the vehiclefleet that affect nighttime vision.The components of the figure of meritdeveloped by Ford comprise many of thedriver, vehicle, environment, andhazard characteristics associated withaccidents caused by inadequatenighttime visibility. A review of the

21

model is needed to assess the modelcomponents and performance criteriathat affect the figure of meritvalue. It would also be worthwhile toexplore the possibility of relatingchanges in the figure of merit toaccident rates for various subfleetseven though such a task may prove ,obe very difficult.

2) Because the most recent data on theextent and nature of headlamp misaimis over 10 years old, a survey ofcurrent aim practices is needed. Thissurvey will help to evaluate theseriousness of the aim problem andidentify the sources of aim variancethat need to be remedied. The extentof the headlamp dirt problem alsoneeds to be surveyed for the samereasons.

3) Because of the significance ofdiscomfort glare to the evaluation ofheadlamps, further work is needed toincrease our knowledge of thesubject. Additional testing underactual field conditions with a broadrange of subjects is required. Also,evaluations of the exposure of driversto different glare levels is needed todetermine frequency of exposure toglare and the intensity of theexposure. Such information can beused in assessing the tradeoffsbetween a beam pattern which mightprovide intense glare in relativelyfew meeting situations versus a lampwhich provides lower glare levels in alarge number of meetings.

4) Because many new technologies forheadlamp construction are beingdeveloped, a better understanding oftheir potential advantages anddisadvantages is needed.

From a broader perspective, a betterunderstanding of the visibility conditionsassociated with accidents needs to bedeveloped. A knowledge of the visibilitycharacteristics of struck obstacles orpedestrians or road delineation could helppinpoint the major safety problems of nighttimedriving and provide a better understanding ofappropriate countermeasures. Given what weknow now about the limited seeing distanceavailable to drivers, it might also beworthwhile to try to answer the question of whydrivers maintain speeds so far above theirability to stop or maneuver to avoid acollision.

REFERENCES

1. Gallagher, V.P. and Janoff, M.S.Interaction between fixed and vehicularillumination systems. Franklin InstituteResearch Laboratories, Philadelphia, Pa, ReportNo. F-C2873/ FHWA-RD-72-51, July, 1972

2. Scott, P.P. The relationship betweenroad lighting quality and accident frequency.The Transport and Road Research Laboratory,Digest LR929, 1980.

3. Merritt, J.O., Helander, M., Abrams,C.,and Miller, J.C. The effect of wide-angleheadlight illumination on driver performance.Human Factors Research Inc. Contract No.DOT-HS-7-01724, June 1982

4. Treat, J.R. and Romberg, R.A.Tri-levelstudy: Modification. Task 1: Finalreport on potential benefits of variousimprovements in vehicle systems preventingaccidents or reducing their severity. Finalreport. Indiana University, Bloomington,Institute for Research in Public Safety, ReportNo. DOT-HS-805 094, 1977.

5. Hale, A. and Blomberg, R.D."Conspicuity for Pedestrians and Bicyclists -Task 2 Interim Report: Operational Definitionof Conspicuity and Accident Analyses" NHTSAContract, DTNH22-80-C-07052, February, 1982.

6. Burger, W.J., Smith, R.L., Queen,J.E., and Slack, G.B. Accident and nearaccident causation: The contribution ofautomobile design characteristics. Dunlap andAssociates, LaJolla, California, Contract No.DOT-HS-802-714, 1977.

7. Koth, B.W., McCunney, W.D., Dverk,D.P.,Janoff, M.S., and Freedman, M. VehicleFog Lighting: An Analytical Evaluation.Franklin Institute Research Laboratory. ReportDOT-HS-803-442, March 1978.

8. Haghighi-Talab, D., "AnInvestigation into the Relationship BetweenRainfall and Road Accident Frequencies on TwoCities" Accident Analysis and Prevention - Vol5, No. 4, 1973 pp 345-349.

9. OECD Road Research Group, "AdverseWeather, Reduced Visibility, and Road Safety"The Organization for Economic Cooperation andDevelopment, August 1976.

10. Bhise, V.D., Parker, E.I., Saunby,C.S., Troell, G.M., Walunas, O.B., andBernstein, A. Modeling Vision With HeadlightsIn a Systems Context. Ford Motor Company,Automotive Safety Office, Dearborn, Michigan.Society of Automotive Engineers, InternationalAutomotive Engineering Congress and Exposition,28 Feb-4 March 1977, Report No. SAE 770238.

11. Devaux, P. Unified European PassingBeam and Yellow Light. International RoadSafety and Traffic Review, 4, 1956, 33-39.

12. Christie, A.W., Ashwood, J.E., andSymons,R.D.H.. Visual Acuity In YellowHeadlights. Road Research Laboratory,Crowthorne, England, Report No. RRL Report LR156, 1968.

13. Schreuder, D.A. White or yellowlight for Vehicle Headlamps? Institute forRoad Safety Research SWOV, Voorburg (TheNetherlands). Report No. 1976-2E, 1976.

14. Lee, R.L. and Scott, D.P. "Driver EyeHeight and Vehicle Sales Trends" FederalHighway Administration, Office of HighwaySafety, January, 1981.

22

15. Olson, P.L., Jorgeson, C.M., Sturgis,S.P., Thomas, J.K., and Domas, P.A. FactorsInfluencing the Effectiveness of AutomotiveRear Lighting Systems. The Highway SafetyResearch Institute, University of Michigan,Report No. UM-HSRI-HF-75-4, October 1975.

16. Pulling, N.H., Wolf, E., Sturgis,S.,Vaillancourt, D.R., and Dolliver, J.J.,"Headlight Glare Resistance and Driver Age"Human Factors 1980, 22(1), 103-112

17. Sturges, S. and Osgood, D.,"Measuring the Quality of Drivers' NightVision" Proceedings of the Human FactorsSociety 25th Annual Meeting - 1981

18. Hauser, E.W., Moore, R.P., andHomeyer, P.G. Macklenburg CountyDrinking/Driving Survey, 1970. ResearchTriangle Park, North Carolina. ResearchTriangle Institute, Project No. SU-582-4,January 1971.

19. Nelson, J.H. Design and use ofheadlamp meeting beams. Proceedings of theInstitute of Mechanical Engineers, AutoDivision. 1954-55, 3, 100-107, 117-129.

20. Moore, R.L. Headlight design.Ergonomics,1958, 1(2), 163-176.

21. Roper, V.J. Four headlamps forbetter seeing. Traffic Engineering, 1957, 27,171-175.

22. Devaux, P. State-of-the-art -signaling and lighting. InternationalCommission on Illumination (France). 1970International Automobile Safety ConferenceCompendium, SAE, Report No. SAE 700386, 1970,555-580

23. Kilgour, T.R. Some results ofcooperative vehicle lighting research.Chrysler Corporation. National ResearchCouncil, Highway Research Board Bulletin, 1960,255. 92-100

24. Meese, G.E. Vehicular Lightingsystems fortwo-lane rural highways. GeneralElectric Company, Cleveland, Ohio. Workshop onHighway Visibility, Highway Research Board,Committee on Visibility, 1972.

25. DeBoer, J.B. A "duplo" headlightwith asymmetric passing beam. Light andLighting, 1955, XLVIII(4).

267 DeBoer, J.B. Progress in automobilelighting as a result of internationalvisibility tests. Road Safety and TrafficReview. Winter, 1956,~ 18-23.

27. Jones, K.J. Comments on vehiclelighting following international visibilitytests. Joseph Lucas (Electrical) Ltd.,Birmingham, England. International Road Safetyand Traffic Review, pp. 24-25, 41, Winter 1956

28. Bailly, F., Cibie, P., and Devaux, P.The use of iodine lamps in car headlights.NNHZWF Motor Industry Research Association,Lindley, England. Ingcnieurs de 1' Auto, June1963, J36(6), 317-330.

29. Roper, V.J. and Morgenstern, L.A.Quartz-iodine headlamps slated for Europeandebut. General Electric Company, Cleveland,Miniature Lamp Department, Ohio, SAE Journal,1964, 72(6), 30-35.

30. Ehrhardt, R.Q. Halogen sealed beamhead-lamps. General Electric Company,Miniature Lamp Department, Cleveland, Ohio.Society of Automotive Engineers Congress andExposition, 26 Feb- 2 March 1979, Report No.SAE 790200.

31. Zaccherini, F. and Thulin, A. Asurvey of motor vehicle headlights. Stockholm,P66-332, Final Report, Lab C 12, 1969.

32. Hull, R.W., Hemion, R.H., and Cadena,D.G. Guidelines for improving the stability ofheadlamp aim. Final report. SouthwestResearch Institute, San Antonio, Texas, ReportNo. AR829/ DOT/HS 800 739, August 1972.

33. Roper, V.J. and Meese, G.E. Morelight on the headlighting problem. GeneralGlectric Co., Miniature Lamp Department,Cleveland, Ohio, 1963.

34. Sturgis, S.P. Motorcycleheadlighting research. University of MichiganHighway Safety Research Institute, Report No.UM-HSRI-HF-75-3, Contract No. DOT-HS-4-00907,August 1975.

35. Olson, P.L. and Abrams, R.A.Improved motorcycle and moped headlamps.University of Michigan Highway Safety ResearchInstitute, Report No. 82-18, May 1982.

36. Adler, B. and Lunenfeld, H.Three-beam headlight evaluation. AirborneInstruments Laboratory, Deer Park, N.Y. ReportNo. 0276-1/DOT-HS-800 844, April 1973.

37. Bhise, V.D., McMahan, P.B., andFarber, E.I. Predicting target-detectiondistance with headlights. TransportationResearch Record. 1976, No. 611, 1-16.

38. DeBoer, J.B. and Morasz, W.Berechnung der Schweite aus der Lichtverteilungvon Automobilescheinwerfern. N.V. Philips'Gloeilampenfabrieken. Light-TechnicalLaboratory, Eindhoven, Netherlands,Lichttechnik, 1956, 8(10), 433-437.

39. Graf, C.P. and Krebs, M.J. Headlightfactors and nighttime vision. Final report.Honeywell, Inc., Systems and Research Division,Minneapolis, Minn. Report No.76SRC13/DOT-HS-802 102, November 1976.

40. Halstead-Nussloch, R., Olson, P.L.Burgess, W.T., Flannagan, M.J., and Sivak, M.Evaluation of the feasibility of a single-beamheadlighting system. Final report. HighwaySafety Research Institute, Ann Arbor, Michigan,Report No. UM-HSRI-79-91, December 1979.

41. Helmers, G. and Rumar, K. Obstaclevisibility in rural night driving as related toreflective qualities of the road surface.Uppsala University, Department of Psychology,Sweden. Report No. 150, 1974.

42. Helmers, G. and Rumar, K. High beamintensity and obstacle visibility. UppsalaUniversity, Department of Psychology, Sweden.Report No. 150, 1974.

43. Hemion, R.H. Interchangeability ofvehicle headlamps. Southwest ResearchInstitute, Report DOT HS-800966, August 30,1973.

23

44. Hemion, R.H. and Hull, R.W. Optimumtwo-lamp headlighting system. Final report.Southwest Research Institute, San Antonio,Texas, Report No. AR-891/DOT-800 890, July 1973.

45. Johansson, G. and Rumar, K.Available braking distances in night driving.Uppsala University, Department of Psychology,Sweden, Report No. 13, November 1963.

46. Johansson, G., Bergstorm, S.S.,Jansson, G., Ottander, C., Rumar, K., andOrnberg, G. Visibility distances in simulatednight driving conditions with full and dippedheadlights. Uppsala University, PsychologicalLaboratory, Sweden. Ergonomics. 1963, 6(2),171-179.

47. Kazenmaier, A. InternationaleVergleichsversuche and Abblendlichtern vonAutomobilescheinwerfern des amerikanischen undeuropaischen Typs. (International comparativeexperiments with the meeting beams of Americanand European motor-car headlights.) RobertBosch GmbH, Stuttgart, Germany. DeutscheKraftfahrtforschung undStrassenverkenhrstechnik. Heft 94, 1956.

48. Lindae, G. Licht am Fahrzeug - einBeitrag zur Verkehrssicherheit. Robert BoschGmgH, Stuttgart, Germany. ATZ, 1962, (64 (5),152-158. —

49. Mortimer, R.G. and Olson, P.L.Development and use of driving tests toevaluate headlamp beams. Final report.Highway Safety Research Institute, Ann Arbor,Michigan, Report No. UM-HSRI-HF-74-14, March1974.

50. Roper, V.J. and Howard, E.A. Seeingwith motorcar headlamps. IlluminationEngineering, 1938, j!3, 412-438.

51. Roper V. and Scott, K.D. Silhouetteseeing with motor car headlamps. GeneralElectric Company, Cleveland, Nela ParkEngineering Department, Ohio. IlluminatingEngineering Society Transactions, 1939, 34,1073-1084

52. Rumar, K. Visible distances in nightdriving with misaligned meeting dippedheadlights. Uppsala University, Department ofPsychology, Uppsala, Sweden. August 1965.

53. Rumar, K. Halogen and conventionalcontinental European headlights; a comparisonof visibility distances. Uppsala University,Department of Psychology, Sweden, 1970.

54. Rumar, K., Helmers, G., Thorell, M.Obstacle visibility with European halogen H4and American sealed beam headlights. UppsalaUniversity, Traffic Safety Research Group,Sweden, Report No. 133, 1973

55. Schmidt-Clausen, H.J. The visibilitydistance of a car-driver in driving situation.Society of Automotive Engineers, InternationalCongress and Exposition, February 22-26, 1982,Detroit, MI. Report No. SAE 820416.

56. Webster, L.A. and Yeatman, F.R. Aninvestigation of headlight glare as related tolateral separation of vehicles. IllinoisUniversity, Engineering Experiment Station,Urbana, Report No. Engineering ExperimentStation Bulletin 496, 1968.

57. Zechnall, R. Influence of lightdistribution of headlamps on seeing distances.Ingelstam, E., ed., Lighting Problems inHighway Traffic. MacMillan Co., 1963, 53-72.

5 8 1 C a m p b e l l , W.F., "AutomobileStopping Distances for Judging HeadlightPerformance" National Research Council Canada,LTR-ST 595, February, 1973.

59. Huculak, P. Ascertaining the effectsof atmospheric factors during visual detectionexperiments in automobile headlighting.National Research Council, Ottawa, NationalAeronautical Establishment (Canada), Report No.LTR-ST-987, May 1978a.

60. Hull, R.W., Hemion, R.H., Cadena, D.G.and Dial, B.C., "Vehicle Forward LightingPerformance and Inspection Requirements " NHTSAContract FH-11-7342, July 1971.

61. Cole, B.L. Some observations ondisability glare. In: Proceedings of GlareSeminar. Austrailian Road Research Board,Victoria, 1977

62. Holladay, L.L. Action of alight-source in the field of view in loweringvisibility. Journal of the Optical Society ofAmerica. Vol. 14, 1927, 1-9.

6~J. Stiles, W.S. The effect of glare onthe brightness difference threshold.Proceedings of the Royal Society. B104, 1929,322-250.

64. Grime, C. The performance ofhead-lamp meeting beams. Institution ofMechanical Engineers. Automobile Division.Proceedings. 1954-55. 3. 108-129.

65. DeBoer, J.B. Quality criteria forthe passing beam of headlights. Paperpresented at the CTB meeting in Walldorf, WestGermany, September 1973.

66. Schmidt-Clausen, H.J. and Bindels,J.T.H., Assessment of discomfort glare in motorvehicle lighting. Lighting Research andTechnology. 1974. j6(2), 79-88.

67. Mortimer, R.G. and Olson, P.L.Evaluation of meeting beams by field tests andcomputer evaluation. Final report. HighwaySafety Research Institute, Ann Arbor,Michigan. Report No. UM-HSRI-HF-74-27,December 1974.

68. Lulla, A.B. and Bennett, C.A.Discomfort glare: range effects. Journal ofthe Illumination Engineering Society. 10(2),January 1981, 74-80.

69. Miller, N.D., Baumgardner, D., andMortimer, R.G. An evaluation of glare innighttime driving caused by headlightsreflected from rearview mirrors. Society ofAutomotive Engineers, Automobile EngineeringMeeting, October 21-25, 1974, Toronto, Canada.Report No. SAE 740962.

70. Rowland, G.E., Moretti, E.S., andPatton, M.L., Evaluation of glare fromfollowing vehicle headlights. Final report.Rowland and Company, Inc., Haddonfield, N.J.,Report No. DOT/HS 805 837, January 1981.

24

71. Olson, P.L. and Sivak, M. Improvedlow-beam photometries. University of MichiganHighway Safety Research Institute, Report No.UM-HSRI-81-4, February 1981.

72. Yerrell, J.S. ' Headlamp intensitiesin Europe and Britain. Road ResearchLaboratory, Crowthorne (England), Report No.RRL LR 383, 1971.

73. Jehu, V.J. Seeing with foglamps.Light and Lighting, May 1953, 46(5), 174-178.

74. Jehu, V.J. Headlight setting. PartI. An investigation of commercial testers.Part II. - An improved headlight tester.Automobile Engineer. Part I, April 1954, 44(4)159-164. Part II: 44(5), May 1954, 213-215.

75. Jehu, V.J. Vehicle headlamptesting. Road Research Laboratory, Crowthorne(England), 1958.

76. Harris, A.J. The meeting beams ofheadlights; effects of deterioration andmisaim. Illuminating Engineering SocietyTransactions. 18(8), 1953, 207-220

7T.Finch, D.M., Dunlop, D.R., andCollins, D.M. Headlamp survey program.University of California, Berkeley. Report No.HP-47, September 1969.

78. Finch, D.M. Controlling headlampaim. California University, Berkeley, Collegeof Engineering, 1970.

79. Heath, W.M. Problems in passengercar headlamp aim. Society of AutomotiveEngineers, International Automotive EngineeringCongress, Jan. 8-12, 1973, Detroit, Michigan,Report No. SAE 730009.

80. Terry, R.M. Automotive headlampaiming - intentions and results. Society ofAutomotive Engineers, International AutomotiveEngineering Congress, January 8-12, 1973,Detroit, MI, Report No. SAE 730007. '

81. Olson, P.L. and Mortimer, R.G.Analysis of sources of error in headlamp aim.Automotive Engineering Congress, February 25 -March 1, 1974, Detroit, MI, Report No. SAE740312.

82. Yerrell, J.S. The performance of twoself-level!ing headlamp systems. Road ResearchLaboratory, Crowthorne (England), Report No.RRL LR 378, 1971.

83. Rumar, K. Dirty headlights -frequency and visibility distances. UppsalaUniversity, Department of Psychology, Sweden,1970.

84. Cox, N.T. The effect of dirt onvehicle headlamp performance. Report No. RRLReport LR 240, Road Research Laboratory,Crowthorne (England).

851 Jehu, V.J. A survey of vehicleheadlights. Illuminating Engineering SocietyTransactions, 1953, 18(9). 249-259.

86. Jehu, V.J. A method of evaluatingseeing distances on a straight road for vehiclemeeting beams. Road Research Laboratory,Crowthorne, England. Illuminating EngineeringSociety Transactions. 1955, 20(2), 57-67.

$7.Jehu, V.J. A method of evaluatingseeing distances on a curved road and itsapplication to headlight beams in England.Illuminating Engineering Society. London.Transactions, 1957, .22(3), 69-83.

88. Mortimer, R.G. and Becker, J.M.Development of a computer simulation to predictthe visibility distance provided by headlampbeams. Highway Safety Research Institute, AnnArbor, Michigan, Report No. UM-HSRI-HF-73-15,July 1973a.

89. Mortimer, R.G. and Becker, J.M.Computer simulation evaluation of current U.S.and European headlamp meeting beams, and aproposed mid beam. Highway Safety ResearchInstitute, Ann Arbor, Michigan. AutomotiveEngineering Congress, 25 Feb-1 March 1974,Detroit, Michigan. Report No. SAE 740311.

90. Mortimer, R.G. and Becker, J.M.Development of a computer simulation to predictthe visibility distance provided by headlampbeams. Highway Safety Research Institute, AnnArbor, Michigan, Report No. UM-HSRI-HF-TM-73-4,May 1973b.

The viewsexpressed in this paper are those ofthe authors and not necessarily those of theNational Highway Traffic Safety Administration

25