Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-04-133
Date: December 2005

Enhanced Night Visibility, Volume II: Overview of Phase I and Development of Phase II

PDF Version (687 KB)

PDF files can be viewed with the Acrobat® Reader®


Research and development of the UV–A and fluorescent technology began in Sweden in the late 1980s. Saab and Volvo® formed a joint venture called Ultralux® to develop and promote the technology. Prototype UV–A headlamps were manufactured and used in field tests and demonstrations. These headlamps incorporated components from a number of European manufacturers: 50-watt (W) high intensity discharge (HID) lamps with enhanced UV–A output from Philips®, reflectors and housings from Hella® and Valeo, filters from Schott, and ballasts from Hughes Power Products®. In the end, though, this technology was not adopted in Europe, largely because of a perceived lack of market; by the late 1990s, the European UV–A effort had been largely abandoned. The situation required that a major activity area of the ENV project be the fabrication and assembly of UV–A headlamps to help develop of a UV–A headlamp specification.

Before commercial UV–A headlamps could become a reality, specifications would need to be developed by the Society of Automotive Engineers, approved by the National Highway Traffic Safety Administration (NHTSA), and incorporated in Federal Motor Vehicle Safety Standards (49 CFR 571.108), as is the case for conventional headlamps. The specification parameters need to differentiate UV–A headlamp performance from that of conventional headlamps. Because UV–A radiation is not visible, it cannot be measured by the usual photometric equipment in units of luminous flux (such as candela per square meter (cd/m²) at a specified distance). Instead, it must be measured by UV–A-sensitive radiometric equipment in terms of radiant flux (such as microwatts per square centimeter (cm²) at a specified distance). Equally critical, the spectral characteristics of UV–A headlamps must be specified.

Research was needed to determine the following minimum specifications:

  • Required distribution of UV–A radiant intensities (isowatt/steradian diagrams).

  • Desired spectral power distribution, including peak wavelength, of the headlamp output.

  • Upper and lower cutoffs for filtering the UV–A.

  • Aiming test points and procedures for UV–A headlamps.

  • Operational design constraints and guidelines that ensure there is no potential biohazard (e.g., no UV–A radiation when the vehicle is traveling at less than 42 km/h).

Obviously, these would concern vehicle manufacturers, but some additional design and safety considerations also required resolution. Only industry could undertake the research and development necessary to address these issues because they were beyond the direct scope of the analytical and empirical research to be conducted as part of the contract. Nonetheless, it was hoped that the ENV project, through its work with stakeholders, would help foster and coordinate these proprietary efforts:

  • UV–A headlamp efficiency needs to be maximized to provide the required UV–A output while minimizing power requirements.

  • UV–A headlamp optics require materials that transmit rather than screen out UV–A, while allowing styling flexibility.

  • Auxiliary UV–A high beam use may allow the Society of Automotive Engineers to adopt the sharp cutoff low-beam pattern used by the Economic Commission for Europe. This would yield safety benefits (resulting from lower glare to oncoming traffic) and also offer manufacturers greater economy of scale.

  • UV–A headlamp safety needs to be addressed and resolved for manufacturers' liability concerns.

  • UV–A headlamps must be capable of being frequently turned off and on (e.g., when vehicle is going slow or is stopped) without significantly shortening the life of the lamp.

  • UV–A headlamps in normal use would supplement conventional low beams: however, for use in fog or mist conditions, the benefits of UV–A (absence of backscatter) may be substantially reduced if a conventional low beam is also on. Yet if only the UV–A headlamps are on, the vehicle would not be adequately visible to oncoming traffic, and the driver would not be able to see the road surface. Therefore, appropriate visible-light fog lamps need to be developed to enhance the visibility of the UV–A vehicle without creating undue backscatter.


From the outset of the ENV project it was clear that obtaining appropriate UV–A headlamps would be a critical issue. Early in the process the research team contacted vehicle and headlamp manufacturers, only to discover that there was no source for readymade UV–A headlamps; the Ultralux prototype components were no longer available, and the headlamps from North American Lighting® had much lower UV–A output. Fortunately, Labino AB manufactured UV–A lighting for other applications. Thus, bulbs, ballasts, and filters were readily available to support the testing and demonstration activities.

Project requirements for UV–A headlamps were considered in the following phases:

  • Year 2: To carry out tests of the UV–A and fluorescent technology on the Virginia Smart Road (appendix A), the project needed to assemble or fabricate a small number of UV–A headlamps. These headlamps would be considered preprototypes rather than true prototypes in the sense of a product being readied for market. The team planned to use the Labino high-output UV–A 35-W bulbs and ballasts. If necessary, as many as four to six UV–A sources may have been placed on a vehicle to attain a high UV–A condition. A high-beam reflector (such as for a halogen (i.e., tungsten-halogen) headlamp) would be needed, and it might have required customization to handle the UV–A bulbs. The Labino Trac-Pack housing may have been appropriate, but the housings may have required fabrication by VTTI. The Schott filters used in the Ultralux units worked well, and the team planned to obtain more of these and use the Labino lamps. The protective front glass needed to transmit UV–A.

  • Years 3 and 4: The research team expected that for more extensive and naturalistic testing and demonstration using privately-owned vehicles, only two (or possibly three) prototype UV–A headlamps with sufficient output would be required per vehicle. These headlamps could still be attached as add-ons, so they would not need dynamic styling. Ideally, these headlamps would have been developed and assembled for the project with the help of USCAR, Labino, or another private industry.


Establishing a basis for UV–A headlamp specifications required both analytical and empirical research. A limited number of field studies with UV technology in Sweden and the United States had demonstrated the potential for large gains in nighttime visibility: however, these were not parametric studies because the levels of UV–A output, fluorescent efficiency, and resulting luminance were not measured and specified as independent variables.

The performance of the UV–A and fluorescent technology as a system depends on the combined performance of the headlamps and the fluorescent materials. Until the UV–A output is fully specified, it is not possible to say what level of fluorescent efficiency is required of the materials to yield a given increase in luminance and visibility distance. Likewise, without knowing the fluorescent performance characteristics of the materials, a headlamp designer would not know how much UV–A is enough.

To address these issues, the team planned a two-stage approach. The first stage included development and implementation of a computerized visibility model supplemented by measurements of prototype headlamps and materials to analytically test various hypothetical UV–A headlamps in concert with various hypothetical materials. The second stage included field studies to empirically verify results of the model and test proposed specifications for UV–A headlamps and fluorescent materials.

Procedures and Methods

UV–A Headlamp Measurements

Samples of prototype UV–A headlamps were to be obtained and measured, including the Ultralux, Labino, and North American Lighting units and unassembled components (lamps, lenses, filters) as they became available. The FHWA’s Photometric and Visibility Laboratory at the Turner-Fairbank Highway Research Center (TFHRC) was to make measurements according to the following plan:

  • UV–A beam pattern: The unfiltered, UV–A-enhanced HID headlamp would be affixed to a computer-controlled, three-axis goniometer. A complete set of luminous intensity measurements would be made at a standard array of geometries (about 1,200 readings), and isocandela diagrams would be generated. The UV–A filter would then be put in place, and a spectroradiometer would measure the UV–A output in the direction that yielded the highest photometric reading. A conversion factor would then be applied to scale the photometric readings into radiometric estimates of UV–A output, and isowatt diagrams would be generated.

  • Relative spectral power distribution (SPD): The relative SPD of the UV–A headlamps would be measured with a spectroradiometer.

  • Spectral transmissivity of the lens and the UV–A filter: These would be measured with a spectroradiometer.

Measurement of Fluorescent Materials

The evaluation of fluorescent infrastructure materials is described in chapter 3, activity 2.

Model Pavement Marking Visibility

A computerized model of visibility for retroreflective pavement markings was to be developed to predict the visibility of retroreflective and fluorescent pavement markings in halogen low beams, HID, and low beams supplemented with UV–A headlamps. The model was to serve as a tool to determine target visibility over a wide spectrum of parameters with relative ease. One major advantage of the computer model would be its ability to examine headlamp and pavement-marking scenarios that were unavailable for field evaluation.

The proposed computer model was to provide the following outputs: pavement marking luminance, road surface luminance, luminance contrast, threshold contrast, and visibility distances. The user would be able to specify the following input parameters: driver age, exposure time, probability of detection, ambient sky luminance, headlamp intensity distribution (for low-beam, HID, and UV–A headlamps), and pavement marking material retroreflectivity and fluorescent efficiency. The user could also vary the spectral composition of the headlamps and the spectral response of the fluorescent markings. Headlamps, observer, and pavement markings would have been implemented in three degrees of freedom only (translation only). The model would handle any number of headlamps attached to the vehicle at any location (three translation degrees of freedom).

The model was to be implemented in MatLab® and C++ for Microsoft® Windows® (Microsoft Windows 95, Microsoft Windows 98, and Microsoft Windows NT®). The preliminary design stage was to include definition and documentation of the required mathematical equations, definition and documentation of a thesaurus for the data structures and functions, development of an overall (global) data structure and function structure, and definition of the input and output data formats. The main design stage was to accomplish refinement of the overall design of the individual functions and local data structures. Modeling of fluorescent pavement markings was to include the spectral sensitivity. A brief final report describing the Phase I model and the results of the test scenarios were to be submitted.

Issues Addressed

The researchers planned to use the model to seek preliminary answers to several issues raised by the vehicle team:

  • The required distribution of UV–A radiant intensities: UV–A radiant intensity distributions (isowatt/steradian diagrams) for all available UV–A prototype headlamps were to be entered into data matrices. A particular prototype would then be selected as an input parameter, which could be multiplied by a scalar vector to increase or decrease the UV–A output. First, the model would position and aim each prototype to maximize pavement marking visibility. Each prototype would be subjected to a sensitivity analysis to determine the lowest scalar multiple of its UV–A output that would achieve a criterion for adequate driver preview time.

  • The desired spectral power distribution, including peak wavelength, of the UV–A headlamp output and the upper and lower cutoffs for filtering: The desired spectral power distribution would be determined in part by the action spectra of available fluorescent materials because the fluorescent efficiency of a material depends on how well the action spectrum of a given material matches the spectral output of the UV–A headlamp. In addition, the wavelength cutoffs for filtering are affected by the biohazard potential at the lower wavelength end and the potential for creating glare at the upper wavelength end. The model would make it possible to easily examine what effect these spectral parameters had on pavement marking luminance and preview times. Specifically, the model would run sensitivity analyses using the spectral power distributions from the available UV–A bulbs; action spectra from all available fluorescent materials; 320, 330, or 340 nm as a lower bound; and 370, 380, or 390 nm as an upper bound.


This model would be limited to straight and level roadway situations. Only the visibility of pavement markings was to be predicted. Glare and backscatter caused by fog was not to be considered. Because these are significant limitations, the research team recommended that the model eventually be enhanced to allow for the following additional analyses:

  • Visibility and legibility of TCDs at various levels of fluorescence in various illumination conditions.

  • Visibility of pedestrians with clothing at various levels of fluorescence in various illumination conditions.

  • Effectiveness of UV–A headlamps in various ambient illumination conditions.

  • Effectiveness of UV–A headlamps in various glare conditions.

  • Effectiveness of UV–A headlamps in various levels of fog. It should be noted that existing visibility models consider only the extinction part of fog. Attempts would be made to accurately model the effects of backscatter (veiling luminance) caused by fog as well.

The Model

Work on the model began in January 1999. A limited working prototype of the model was to be released to the project team as soon as practicable to allow for feedback from the model users. Availability of fluorescent pavement-marking data was not critical for the development of the computer model itself; the development could use hypothetical pavement marking fluorescence efficiency matrices for the time being; however, such data would be necessary for the final validation of the model and for performing the model runs. Both the final validation and the model runs were to occur when pavement marking fluorescent-efficiency data became available.

Field Tests

Field testing was to be conducted on the Smart Road and developed in coordination with the driver/pedestrian and infrastructure teams. Visibility measures were to be made for various headlamp and TCD conditions. The prototype UV–A headlamp with the highest radiant output was to be used. It was desirable that visibility be tested with at least two levels of UV–A intensity. Because the UV–A output of the preprototype headlamps might be less than optimal, a high-UV–A-output condition was to be achieved by using up to six of these headlamps per vehicle. The low-UV–A-output condition was to be implemented using fewer of these headlamps.

Equally important was the issue of what standard headlamps to use for the field tests. Until the time of the project, it had been appropriate to use a common tungsten-halogen headlamp. In view of the increased visibility claimed for the newly-developed metal halide, HID headlamps, it seemed important to include these as a basis for comparison. The use of a high-beam condition for comparison against the UV–A condition was also considered. While it would have been of interest to know how well the UV–A headlamps compared to a visible high beam, this comparison did not translate to a real-world option. That is, even if a visible high beam could outperform a UV–A system, drivers would simply be unable to make use of these high beams in the real world in most driving situations because ofthe presence of other traffic. The research team proposed that the field testing include the following six headlamp conditions:

  • Tungsten-halogen low beam alone.

  • Tungsten-halogen low beam with auxiliary high-UV–A-output headlamps.

  • Tungsten-halogen low beam with auxiliary low-UV–A-output headlamps.

  • HID low beam alone.

  • HID low beam supplemented with auxiliary high-UV–A-output headlamps.

  • HID low beam supplemented with auxiliary low-UV–A-output headlamps.

Comparisons of visibility distances, subjective ratings, and measurement of glare in these six conditions (as part of activity 4) were planned to help answer the following questions:

  • How do these dependent measures vary as a function of the UV–A output?

  • Does UV–A-enhanced lighting offer advantages over HID and tungsten-halogen alone?

  • How do field data compare with predictions from the visibility model? (The model could have been recalibrated, allowing more accurate predictions for a wider range of conditions than could have been field tested.)

  • How do the glare measures of the six headlamp conditions compare? In particular, do participants experience increased discomfort or disability glare in the UV–A conditions? If so, this may indicate a need to lower the upper-wavelength bounds for the UV–A headlamp filter.


Previous    Table of Contents    Next
United States Department of Transportation - Federal Highway Administration