Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 3 CHAPTER 3 FINDINGS AND APPLICATIONS RESEARCH REVIEW AND SYNTHESIS In the present review and synthesis of human factors research associated with the perception of colored signal lights such as traffic signals, emphasis is placed on recent research on reaction times and missed signals, discomfort glare, and brightness perception. An introduction briefly summarizes current maintenance practices. The present review focuses on research published after 1998, when the first interim specification for LED traffic signals was published by the Institute of Transportation Engineers (ITE, 1998). LED traffic signal modules have gained popularity in the U.S. primarily because they consume far less energy than incandescent lamps - 85% less, on average (Iwasaki, 2003). Because of these energy savings, the U.S Environmental Protection Agency (EPA) recognized LED traffic signal modules as an ENERGY STAR product in 2000, and then Congress mandated that as of January 2006, all red and green traffic signal modules must meet the energy consumption specifications stipulated by the ENERGY STAR requirements (ENERGY STAR, 2003). In addition to energy savings, the use of LED traffic signals has the potential to increase safety at intersections. The reduced power required to operate LED traffic modules has allowed the use of low cost battery backup systems at intersections, increasing safety in the case of blackouts; a $3,000 backup system can power an intersection for two to four hours (Iwasaki et al., 2003). Also, since the late 1990s, LED modules typically reach end of life by reduced light output, rather than complete failure, so even âfailedâ modules usually give some signal information to drivers (Behura, 2007). Maintenance Behura (2007) reported on a survey of LED traffic signal maintenance practices. He observed that the use of LED traffic signals had been expected to reduce the lifetime cost per module compared with incandescent modules. In addition to reduced energy expenditures, the extended lifetime of the LED modules had been expected to reduce relamping material and labor costs by reducing the frequency these costs will be incurred. Behura (2007) stated that many agencies had not implemented appropriate maintenance programs for LED signal modules. The survey showed that: ⢠35% have no replacement program ⢠35% are complaint driven (despite the fact that LED modules typically reach end of life due to dimming rather than complete failure) ⢠24% implement routine, scheduled replacement
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 4 ⢠3% replace on vendor product life cycle ⢠3% replace based on in-service test results Rather than relying on a passive maintenance scheme, Behura (2007) suggested several maintenance schemes from most to least precise: ⢠Remove modules from service and test light output in a laboratory ⢠Field measurements might provide good results, if they could be done properly ⢠Statistical analysis based on time since installation ⢠Replace based on the warranty period Regardless of the maintenance practice, Behura (2007) recommended keeping a database of modules in the field including location, color, type, manufacturer and model, serial number, date of purchase, date of installation, and warranty end date. Behura also recommended that agencies clean lenses at intervals of one to two years (which would provide an opportunity for field measurements too). Visibility Perception and Reaction Time A study by Bullough et al. (2000) showed that under simulated daylight viewing conditions, LED and incandescent modules of the same nominal color, luminance, and onset time resulted in no statistically significant differences in mean reaction times, percentages of missed signals, color identification accuracy, and subjective brightness ratings. That study (Bullough et al., 2000) did find that reaction times and the percentage of missed signals decreased as luminous intensities (or luminances) increased, and that to obtain the same performance as a red traffic signal meeting then-current photometric requirements for luminous intensity for a 200-mm diameter module, yellow signals had to have a luminous intensity between 1.4 and 2.4 times higher than the red signal, and green signals had to have a luminous intensity between 2.4 and 2.8 times higher than the red signal. Freedman (2001) reported that yellow signals required a luminous intensity for the yellow about twice that of the red to obtain equal visual response, and that green signals required a luminous intensity of 1.3 times higher than red to obtain equal visual response. (Fisher and Cole [1975] recommended a 3:1 luminous intensity ratio for yellow:red and a 1.3:1 ratio for green:red). Taking these studies into account, the ITE recommended a luminous intensity ratio of 2.5:1 for yellow:red and 1.3:1 for green:red in its later specification for LED traffic signal performance (ITE, 2005). The only discrepancy among these studies is the higher ratio for green signals obtained by Bullough et al. (2000), which might be explained by the background light source used in their study, which had a correlated color temperature of about 3850 K, slightly lower than typical daylight illumination between 5500 and 6500 K (Wyszecki and Stiles, 1982). Regardless, since
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 5 the fundamental meaning of the green signal (i.e., "go") is different from that of the red and yellow signals (i.e., "stop"), it could be argued (Bullough, 2002) that equivalent reaction time and brightness for the green signal (relative to red) is not critical for driving safety. In the study by Bullough et al. (2000) lamp onset time was held constant (by using an electromechanical shutter). However, in the field, the onset time of incandescent traffic control signals is longer than that of LED sources (100 to 200 ms for incandescent, versus 13 to 32 ms for LEDs). Bullough (2005) reported the effect of this variation in onset time, and found that the differences in response time were very short. For example, the difference in response time for red traffic signals meeting the ITE (1998) specifications for luminous intensity, and having rise times of either 17 or 87 ms, was about 30 ms. The same comparison for yellow signals meeting ITE (1998) specifications yielded a response time difference of about 25 ms. Nor did rise time affect the consistency with which a signal was detected, so it was concluded by Bullough (2005) that onset time has little practical consequence for traffic signals. Cohn et al. (1998) confirmed that the visibility of red LED modules and red incandescent signal modules was about equal under daylight conditions, and concluded that the pixilated appearance of some LED signals might actually provide a visual benefit. This conclusion was confirmed in a subsequent study by Bullough et al. (2002) who found response times to a signal light consisting of an array of point sources, but with equivalent far-field luminous intensity as a diffuse signal light, were shorter than to the diffuse signal light. While it could be argued that performance metrics such as reaction time and missed signals are most important when considering traffic signals, some studies examined subjective metrics such as brightness and visibility. As described above, Bullough et al. (2000) found no statistically significant difference in perceived brightness between incandescent and LED lamps of the same luminance (for red, yellow, and green) under simulated daylight conditions, although the number of brightness judgments made in that study was relatively small. A study by Bullough et al. (2007) of green LED versus green incandescent signals viewed under nighttime conditions found that the LEDs appeared to be 1.4 to 1.7 times brighter, which is attributed to their saturation.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 6 Figure 1. Luminous efficiency functions for color-normal and protan observers, and spectral distribution from red incandescent and LED traffic signal modules (Andersen, 2002). Color Vision Deficiency According to a review by Cole (2004) of 124 journal articles, color-deficient drivers: ⢠Have longer reaction times to signals ⢠May confuse signal lights with street lights ⢠Have shorter recognition distances, especially against a bright sky ⢠Can mistake red lights for yellow (with greater luminance correlated with a greater error rate; this is because color-deficient drivers tend to rely on relative luminance to distinguish between red and yellow signals, so increasing the luminance of a red signal makes it more likely to be interpreted as being yellow) Because approximately 4% of the population has color deficient vision, significant attention has been paid to this issue when specifying signal properties. Andersen (2002) noted that the specifications were particularly important for the long wavelength cutoff for red signals because the long wavelength end of the luminous efficiency function for protan observers overlaps only with the short wavelength end of the red LED spectrum as shown in Figure 1. He found that shifting the red LED loci by only 8 nm toward the longer wavelengths can result in a 21% difference in the visual signal provided to protan observers.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 7 a. b. Figure 2. a) Average reaction times and interquartile ranges for color-normal subjects; b) average reaction times and interquartile ranges for protan subjects (Huang et al., 2003). Huang et al. (2003) found in a series of experiments that for red and yellow traffic signals using LED and incandescent sources, protan observers had longer reaction times (Figure 2), a greater number of missed signals, and a greater number of color misidentifications than color- normal observers. Huang et al. (2003) tested red LED signals with different dominant wavelengths and concluded that the shorter dominant wavelengths improved detection among protan observers, but decreased the rate of correct color identification. The results suggested that the color boundaries specified by the ITE (1998) for each signal color might be improved if boundaries more consistent with recommendations from the Commission Internationale de l'Ãclairage (CIE) were used; and this is the case for the current ITE (2005) specification for LED traffic signal colors. Starr et al. (2004) conducted a field study of green LED traffic signals when viewed under direct sunlight. When traffic signals are viewed under these conditions, they can appear to be lighted even when they are not (the sun phantom effect). Both color-normal and color- deficient observers can misread a signal under these conditions, but it is more common among the color-deficient group. Starr et al. installed fourteen green signals along a route in Minnesota. One of the signals was incandescent, while the rest were LED modules that varied by brand, lens type (tinted or clear), and LED technology (old technology with high LED count versus new technology with lower LED count). Subjects observed the modules while direct sunlight fell on them. The results showed that few (< 4%) color-normal reported that a green signal was on when it was not, but that many more (~25%) of the color-deficient participants falsely reported that a green signal indication was on. While there were variations in results between modules, no clear-cut advantages were identified among the signal modules tested by Starr et al. (2004). Fog Kurniawan et al. (2008) conducted a laboratory study of apparent brightness when subjects viewed LED lamps through a fog of water droplets in the laboratory. Subjects viewed LEDs of various colors through fogs of various water droplet sizes and reported the observed brightness level. The authors found that apparent brightness decreased as the fog droplet size increased, and that all colors were affected about equally.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 8 In their study of signal light brightness, Bullough et al. (2007) found that viewing signals through fog reduced the brightness enhancement of LED signals relative to incandescent signals under nighttime viewing conditions, primarily because scattered light from different light sources is superimposed on the signal images, reducing differences among different signal lights. Discomfort Bullough et al. (2001) studied the visual discomfort that results from viewing LED traffic modules at night. Based on their results and the 1998 interim LED traffic signal specifications (ITE, 1998), about 40% of the population would be expected to experience discomfort when viewing yellow and green LED signals at night, while red signals would not be expected to produce discomfort. Using the current specifications (ITE, 2005), the percentages for yellow and green would be reduced to about 20%, and for red would remain 0%. Reductions in luminous intensity for yellow and green signals by about 30% at night would be expected to reduce discomfort glare almost completely, while having little impact on the visibility of the signals (Freedman et al., 1985). Potential Future Research The research summarized above point to a number of areas where traffic signals could be improved through additional research and development. Several of these concepts have been patented, but are not in widespread practice. Behura (2007) indicated that LED signal modules should be replaced based on when they become too dim. To streamline the module testing process, a low cost photosensor could be built into each module. A signal indicating the luminous intensity could be transmitted, such as over a dedicated signal line to the control box or using radio frequency transmissions, to the maintaining agency. Behura (2007) also indicated that the power circuitry is now more likely to fail than the LED light engine itself. This indicates that research on methods to construct low cost, durable power circuitry and devices would be a fruitful way of decreasing the lifetime operating costs of LED modules. Extensive research has shown that color-deficient drivers have some difficulty detecting and correctly identifying traffic signals under all conditions and that color-normal drivers have difficulty when modules fall under direct sunlight (sun phantom effect). Shape-coding of traffic signal modules has been suggested as a countermeasure for helping overcoming difficulty in viewing by color-deficient observers, and the use of a flashing display is another (Whillans, 1983). Visibility during sun-phantom conditions could be improved by simply increasing (on a temporary basis) the luminous intensity during such conditions and poor ambient weather. Dynamic control of LED intensity results in smaller chromaticity shifts than can be achieved with incandescent lamps, and of course, reductions in intensity can be performed at night to reduce viewer discomfort in accordance with ITE (2005) specifications (provided this does not increase the potential for conflict monitors to have difficulty identifying when the dimmed signals are switched on).
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 9 Finally, the research by Starr et al. (2004) shows it would be useful to develop LED modules that do not permit (or at least reduce) the sun phantom effect. In addition to visors, this may be possible through controlling lens properties or the albedo of the back surface of the module. PHOTOMETRIC MEASUREMENT TECHNIQUES As described by Behura (2007), luminous intensity of LED traffic signal modules can degrade over time. In accordance with the ITE (2005) specifications for the photometric performance of LED signal modules, the ITE recommends that LED traffic signals be replaced when the intensity of the fixture no longer produces the minimum specified luminous intensity. In order to determine the luminous intensity, the ITE suggests monitoring signals over time using a calibrated light meter. In this manner, light measurements from the same signal could be compared over time to determine a percentage of degradation. The ITE (1998) points out that these relative measurements may not provide an accurate measure of absolute intensity. It can be difficult for transportation agencies to determine the performance of a traffic signal both in the field and in the shop, because most agencies do not have photometric measurement equipment and sending modules to a laboratory for testing can be expensive and time consuming. For this reason, the project team investigated several simple methods for measuring LED traffic signal luminous intensity. Two different types of light measuring instruments, an illuminance meter and a luminance meter, were used to estimate the luminous intensity of red and green traffic signal modules under laboratory conditions and under field conditions. Note that the same signals were used throughout the study so that the various methods could be readily compared. Illuminance Test Method Illuminance is a measure of the density of light falling on a surface (Rea, 2000). The luminous intensity of a traffic signal module can be determined by measuring the illuminance falling on a light meter (at a sufficient distance from the module) and then applying the inverse square law to determine the luminous intensity needed to produce the measured illuminance. This is the basis for the following test method. Materials Red and green 300-mm traffic signal modules, an illuminance meter calibrated to measure narrow-bandwidth spectra, a black-painted room capable of being completely dark, a tape measure, a tripod, and a level were used.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 10 Procedure The traffic signal module was secured to a table approximately 1.5 m above the floor and leveled such that the face of the signal module was perpendicular to the floor. The signal module was turned on for at least 30 minutes in order for the signal to thermally stabilize to the room temperature environment. The tape measure was extended along the floor from a point directly below the face of the signal to the furthest measuring distance used (approximately 15 m). The illuminance meter was attached to the tripod and the face of the illuminance meter was adjusted so that it was level with the center of the traffic signal. The tripod was then moved to each desired measuring distance. The lights of the room were turned off so that the only light in the room was the light being produced by the signal. A measurement of illuminance was then recorded, and then the lights were turned back on. This process of turning off the lights and recording the illuminance was repeated for every test distance listed in Tables 1 and 2. Using the measured illuminance from the signal modules at each distance, the luminous intensity of the signal could be determined by applying the inverse square law: Luminous Intensity (cd) = Illuminance (lx) à Distance² (m) Results Tables 1 and 2 summarize the results of the illuminance test method. Discussion The luminous intensity of a signal light in a particular direction is invariant as a function of distance from the signal light. Therefore after applying the inverse square law for every illuminance and distance combination in one direction, the luminous intensity value should be the same. The results show that this is the case. The reason this is important is that in order for the inverse square law to be applied correctly the light source must approximate a point source with light diverging from the source. A photometric rule of thumb (Rea, 2000) is that in order for the inverse square law to be applied with low error (<5%), the illuminance measurement distance from the source to the detector should be at least five times the maximum dimension of the source. In the case of a 300-mm traffic signal module this would be 1.5 m; however, a traffic signal module produces partially collimated light that does not necessarily diverge like a point source at 1.5 m. Therefore the measurement distances for this test method were much larger than 1.5 m, and therefore started at more than 6 m from the signal. This starting distance was necessary so that the light would be seen as diverging from the signal like a point source.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 11 Table 1. Measured illuminance and calculated luminous intensity values for different measurement distances (red signal module). Red Traffic Signal Distance (m) 6.10 7.62 9.14 10.67 12.19 Illuminance (lux) Min 6.48 4.21 2.96 2.11 1.69 Illuminance (lux) Max 6.55 4.23 3.00 2.16 1.70 Illuminance (lux) Average 6.52 4.22 2.98 2.14 1.70 Intensity (cd) 242 245 249 243 252 Table 2. Measured illuminance and calculated luminous intensity values for different measurement distances (green signal module). Green Traffic Signal Distance (m) 6.10 7.62 9.14 10.67 12.19 Illuminance (lux) Min 36.00 22.60 15.50 11.75 9.30 Illuminance (lux) Max 36.50 22.80 15.80 11.80 9.30 Illuminance (lux) Average 36.25 22.70 15.65 11.78 9.30 Intensity (cd) 1347 1318 1309 1340 1382 The smaller the test distance needed to provide reasonably accurate results, the easier it will be to find a space to accommodate traffic signal measurements. The differences among the luminous intensity results for the green signal module were less than 6%, and for the red signal module, the differences were less than 4%. This means that the luminous intensity of the signal module could be estimated in a dark room capable of accommodating a 6 m measuring distance with reasonable accuracy. Nonetheless, there can be differences in the optical design of different modules, so distances longer than 6 m should be employed whenever possible. Luminance Test Methods Luminance is the measure of light intensity in the direction of an observer per unit projected area of a source (Rea, 2000). It is analogous to the visual response known as brightness. Using a calibrated luminance meter, it is possible to measure the luminance of a traffic signal module directly and then divide the measured luminance by its projected area to determine the luminous intensity. Several different methods based on measuring the luminance of the traffic signal have been investigated. Before discussing each test method it is important to describe how a luminance meter is used. Many luminance meters look like a kind of mini-camera with a viewing port, a display, and a measurement trigger. In order to take a luminance measurement, the user should look through the viewing port and aim a marker (usually circular or rectangular) located inside the view port at the object being measured. (The image should also be in proper focus.) A marker inside the viewing port represents a fixed acceptance angle, typically subtending 1 degree in visual angle,
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 12 for light to enter the instrument and be measured. After aiming the marker, a trigger is pulled and a luminance measurement is taken, and the value of luminance is displayed. When a luminance measurement is taken, the light entering the instrument is integrated over the area indicated by the circular or rectangular marker. Because the angle subtended by the marker is fixed, as the distance of the meter from the traffic signal module increases, the marker would cover a larger area of the signal module face. For the present measurements, an instrument with a circular marker nominally subtending 1 degree was used. Using luminance test method 1, the luminous intensity of the traffic signal module was estimated when the circular marker remained smaller than the angle subtended by the signal module (whereby each measurement represents only a portion of the signal module's face). Using luminance test method 2, the signal module luminous intensity was estimated by measuring the intensity of the traffic signal module when the circular marker was larger than the traffic signal module, and the surrounding area around the signal was black (i.e., when the measurement was made in a dark environment). Using luminance test method 3, the luminous intensity was estimated by determining the luminance of the traffic signal module under field test conditions during the day when the circular marker was larger than the angle subtended by the traffic signal. Color Correction Luminance meters are calibrated by recording the luminance of a calibration standard having a known luminance value. The calibration standard is typically an incandescent light source producing white light of a known luminance (Rea, 2000). When a luminance meter is used to take measurements of light sources that differ greatly in color from the reference standard, a color correction factor might be necessary to account for deviations in the instrument's sensitivity from the photopic luminous efficiency function at localized wavelength regions, due to limitations of the meter's spectral response system. The color correction factor for a colored light source can be found by measuring a colored light of known luminance and then dividing the measured result by the known value. For this study the color correction factor for the red and green traffic signals are: ⢠Red = .882 ⢠Green = .961 These color correction factors were applied to all luminance measurements described below (Tables 3, 4 and 5). Luminance Test Method 1
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 13 The luminance of a traffic signal was measured at increasing distances with the circular marker remaining no larger than size of the signal. Materials. Red and green 300-mm traffic signal modules, a calibrated luminance meter, a tape measure, a tripod, and a level were used. Procedure. The traffic signal module was secured to a table approximately 1.5 m above the floor and leveled such that the face of signal was perpendicular to the floor. The signal module was turned on for at least 30 minutes in order for the signal to thermally stabilize. The tape measure was extended along the floor from a point directly below the face of the signal module to the furthest measuring distance (approximately 15 m). The luminance meter was attached to the tripod and the height of the luminance meter was adjusted so that the entrance lens was even with the center of the traffic signal. The tripod was then moved to the desired measuring distance. The lights of the room remained off during this experiment. Luminances were recorded for each distance, and the luminous intensity was calculated as follows: Luminous Intensity (cd) = Luminance (cd/m²) à Traffic Signal Area (m²) Table 3. Luminous intensity values estimated from luminance values using luminance test method 1 (red signal module). Red Traffic Signal Distance (m) 3.0 6.1 7.6 9.1 10.7 12.2 13.7 15.2 Luminance (cd/m2) Min 2434 3743 4533 4660 4265 3793 3332 2914 Luminance (cd/m2) Max 6139 5202 4551 4690 4439 3811 3430 3050 Luminance (cd/m2) Average 4190 4549 4545 4680 4322 3804 3373 2965 Signal area (m2) 0.073 0.073 0.073 0.073 0.073 0.073 0.073 0.073 Intensity (cd) 306 332 332 342 316 278 246 216 Table 4. Luminous intensity values estimated from luminance values using luminance test method 1 (green signal module). Green Traffic Signal Distance (m) 3.0 6.1 7.6 9.1 10.7 12.2 13.7 15.2 Luminance (cd/m2) Min 16299 17903 18922 20191 20931 20969 20911 15847 Luminance (cd/m2) Max 29608 22929 19643 20537 21219 21882 21478 19585 Luminance (cd/m2) Average 22533 20783 19303 20392 21075 21212 21248 18301 Signal area (m2) 0.073 0.073 0.073 0.073 0.073 0.073 0.073 0.073 Intensity (cd) 1645 1517 1409 1489 1538 1549 1551 1336 Results. Tables 3 and 4 summarize the results of luminance test method 1. Discussion. Calculating the luminous intensity using this method requires making an assumption that light emitted from the face of the traffic signal is uniform. This is, however, not the case; rather, the face of the signal module is typically made up of many LEDs, each
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 14 producing light and dark patterns over the face of the signal due to their distance from each other and their individual beam patterns. When the luminance meter was closest to the traffic signal (i.e., at a distance of about 3 m), the circular marker would surround only 4 or 5 individual LEDs. Due to the ratio of the luminance of the LEDs to that of the surrounding background, luminance measurements of different portions of the signal module would be highly variable at such a close distance. As the meter was moved further back, the luminance measurements for different portions of the signal module face were less variable. Moving further back helped to better approximate a uniformly emitting surface. The difference in variability of the measurements with respect to distance can be noted by observing that as distance increased, the minimum and maximum luminance values became closer to each other until about 15 m, when the circular marker started to become larger than the traffic signal module face. For this method, a distance of 9 to 12 m produced the least variable results. The distance had a large effect on the measurement of luminance. The luminous intensity results were not as consistent for this luminance test method as they were for the illuminance test method. Luminance Test Method 2 The luminance of traffic signal modules was measured when the circular marker of the luminance meter subtended an angle much larger than the signal module's angular size, with a black surrounding background. Materials. A green 300-mm traffic signal module, a calibrated luminance meter, a tape measure, a tripod, and a level were used. Procedure. The traffic signal module was secured to a table approximately 1.5 m above the floor and leveled such that the face of the signal module was perpendicular to the floor. The signal was turned on for at least 30 minutes in order for the module to thermally stabilize. The tape measure was extended along the floor from a point on the floor directly below the face of the signal module to a distance of 28 m. The luminance meter was attached to the tripod and the height of the luminance meter lens was adjusted so that it was even with the center of the traffic signal module. The tripod was then moved to the measuring distance of 28 m. The lights of the room remained off in the room during this experiment. Luminance measurements were recorded. Calculation and Result. The following interim calculations were used to assess the luminance, and resulting luminous intensity, of the signal module:
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 15 Distance = 28 m Area of traffic signal module = 0.073 m2 Acceptance angle of luminance meter (measured) = 0.93 degrees Acceptance half angle = 0.465 degrees Area of circular marker = Ï Ã [tan(Acceptance half angle) à (Distance in m)]² Scale factor = (Area of circular marker in m²)/(Area of traffic signal in m²) Luminous intensity (cd) = Luminance (cd/m²) à Traffic signal area (m²) à Scale factor Luminance (measured) = 12000 cd/m² Resulting luminous intensity = 1954 cd Discussion. Compared to the illuminance test method (which tended to provide consistent measurement results), luminance test method 2 overestimates the intensity of the traffic signal moduleâs luminous intensity. Luminance Test Method 3 The luminance of a green traffic signal was measured outside during the daytime in order to simulate a field measurement that might be made by a person from the roadway up at a signal (Condition 1), from the sidewalk up at a signal (Condition 2), and from a bucket truck at a similar height as the signal module (Condition 3). Materials. A green 300-mm traffic signal module, a calibrated luminance meter, a tape measure, and two tripods were used. Procedure. There were three measurement location geometries used; one was 5 degrees below the normal from the signal, one was 5 degrees below and 5 degrees to the left of the signal, and one was directly ahead of the signal. Luminance measurements were taken with the signal on and off. A distance of 23 m was measured and the beginning and end of the distance measurement line were marked with chalk. The traffic signal module was mounted to a large tripod and raised such that the center of the signal was 3 m above the ground. The face of the signal was adjusted to be perpendicular to the ground. The tripod containing the signal was placed at the beginning of the 23 m measurement line. The luminance meter was placed on a tripod and raised so that its lens was 1 m above the ground and then placed at the end of the 23-m distance mark, in line with the center of the traffic signal. Under this condition, the signal was 2 m above the luminance meter lens (corresponding to 5 degrees of angular distance). The luminance meter was aimed toward the signal and measurements were taken from this position. Then, the luminance meter was moved to the left by 2 m, re-aimed toward the signal, and another set of measurements was taken. Finally, the signal was lowered to 1.5 m above the ground, and the luminance meter was raised so that its lens was also 1.5 m above the ground. With this measurement geometry, a final set of luminance measurements was taken. Calculation. The following interim calculations were used to estimate the luminous intensity of the signal module:
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 16 Distance = 23 m Area of traffic signal module = 0.073 m² Luminance (cd/m²) = (Luminance On) â (Luminance Off) Acceptance angle of luminance meter (measured) = 0.93 degrees Acceptance half angle = 0.465 degrees Area of circular marker = Ï Ã [tan(Acceptance half angle) à (Distance in m)]² Scale factor = (Area of circular marker in m²)/(Area of traffic signal in m²) Cosine correction factor = 1/cos(5 degrees) = 1.004 Luminous intensity (cd) = Luminance (cd/m²) à Traffic signal area (m²) à Cosine correction factors à Scale factor Table 5. Luminous intensity values estimated from luminance values using luminance test method 3 in an exterior, daytime environment (green signal module). Green Traffic Signal Module Condition 1 Condition 2 Condition 3 Signal on/off on off On off on off Distance away (m) 23 23 23 23 23 23 Distance below (m) 2 2 2 2 0 0 Distance to the left (m) 0 0 2 2 0 0 Average luminance (cd/m2) 13267 6893 13954 5313 9559 540 On-Off luminance (cd/m2) 6632 8991.5 9384.15 Cos correction for 5 degrees below 1.003854802 1.003854802 na Cos correction for 5 degrees side na 1.003854802 na Area signal (m2) 0.073 0.073 0.073 Scale factor 1.482 1.482 1.482 Luminous intensity 692 942 976 Results. Table 5 summarizes the results of luminance test method 3. Discussion. Compared to the illuminance test method, luminance test method 3 underestimates the intensity of the traffic signalâs intensity. It is possible that scattered light from the daytime sky contributed to measurements resulting in an underestimation of the true luminance difference between the on- and off-signal module measurements. Overall Discussion of Traffic Signal Photometric Measurement Methods The most reliable method for characterizing the luminous intensity of the traffic signal modules used in the preceding series of measurements was the illuminance test method. If a dedicated space for making such measurements and a test jig for mounting the signal module and illuminance meter were able to be set in a fixed location (preferably, with at least 12 m of measurement distance to minimize errors associated with applying the inverse square law [Rea, 2000]). Field measurements of traffic signal intensity based on luminance are prone to variability caused by measurement geometry, non-uniform luminance of module faces, and/or scattered light.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 17 Devices for measuring the luminous flux (or the relative luminous flux) produced by a traffic signal module have been developed whereby a receiver is fitted over the circular module and the luminous flux is gathered into an integrating chamber (Miller and Zaidi, 2002). Such devices can be used for estimating the relative change in luminous intensity for a module that has been previously characterized using far-field photometry. If modules are designed such that individual LED failures might result in a angular-specific degradation of luminous intensity, rather than a uniform reduction of intensity at all angles, then they will not be able to estimate the luminous intensity for a particular direction. LED TRAFFIC SIGNAL FAILURE MODES In this study, analyses were conducted on 49 failed LED traffic signal modules representing all three signal colors and three manufacturers. Four modes of failure were commonly found: failure of the startup ("boot strap") circuit in the driver integrated circuit (IC), heat produced by a power resistor degrading adjacent LEDs, failure of a Schottky diode, and general LED failures. Four suggestions are made for designing LED traffic signal modules to lengthen their life: reducing the complexity of the startup circuit, specifying higher-rated or ceramic capacitors, redesigning the array of LEDs, and moving high power components off the main circuit board. Purpose The purpose of this study was to identify common causes of failure among LED traffic signal modules. To accomplish this, the project team solicited failed LED modules from transportation agencies across the country. A total of 61 failed modules were sent to the project team for analysis by the City of Los Angeles, Nebraska, New Jersey, New York State, and Wisconsin DOTs. Of these, 32 were green, 23 were red, and 6 were yellow. Fifty-three were 300-mm modules, 4 were 200-mm modules, and 3 were 125-mm in diameter. (The 125 mm modules were not analyzed as they appear to be used for purposes other than traffic control at signalized roadway intersections.) Three manufacturers were represented among the samples, which are referred to as Brands A, B, and C. Thirty-four were Brand A, 18 were Brand B, and 9 were Brand C. Failure analysis was conducted on 49 of the 61 modules; a preliminary examination led us to believe the remaining modules had similar failure modes. The following steps were taken to examine the modules: ⢠The modules were visually inspected for external damage. ⢠Power was applied via the external wire leads and the moduleâs state of operation (if any) was noted. For example, some modules failed to light at all, while others flickered.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 18 ⢠The modules were opened and the mechanical state of the inside of the module was inspected. ⢠Attempts were made to get the module to work properly or to identify why it would not, using techniques such as substitution of known good components and application of external power supplies to the LED arrays and/or driver circuits. Based on data sheets for labeled, commercially-available IC chips used in many of the modules, it was possible to predict where supply voltages should be present. This allowed determination of whether the board's power supply was functioning. In absence of schematics for the proprietary electrical designs of the modules, two resources proved helpful. As mentioned above, when the driver ICs in the units were marked with their part numbers, the manufacturer's (of these components) data sheets provided guidance as to how the circuit worked. These IC data sheets also provided the waveform and voltages that were expected on various pins. Second, we also requested and received a few functioning modules, which provided a standard for comparison. Figure 3. Block diagram of a typical LED signal module design. The block diagram in Figure 3 shows the operation of a typical functioning traffic signal module. The LED array is driven by a power factor corrector (PFC) which both corrects the power factor and generates the constant current for the LEDs by controlling its output voltage. In steady-state operation, the circuit self-generates the IC supply voltage from the output of the PFC. During startup conditions, there is a boot strap circuit that provides the initial supply voltage until the IC is running, at which point the IC generates its own voltage. The circuit is efficient in providing both power factor correction and constant current without requiring multiple ICs. Results Two failure modes were common among the modules from two manufacturers (Brand A and Brand C):
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 19 ⢠Fourteen of the 34 Brand A and Brand C modules that were examined (out of the total of 43 Brand A and C modules received) had problems with the boot strap circuit or IC power supply circuit. Typically the failure included one of the resistors overheating and failing, but it is likely that other failures within the power circuitry led to the failure of the resistors. For example, in several units a capacitor failed first, which was the likely cause of the resistor's failure. Many of the modules illuminated if an external power supply was used to power the driver IC. ⢠Twelve of 34 Brand A and Brand C modules that were examined showed issues caused by a large power resistor physically located within the LED array. The heat given off by this resistor appears to have caused the LEDs located above it to fail, which resulted in a cascade of LED failures. It appears that this issue was recognized by signal module manufacturers, because modules with later date codes had this resistor located on a separate, daughter board. In addition to these failure modes, it was noticed that the plastic lenses of many modules often had broken screw mounts, indicating that the impact resistance of the plastic might be degraded. There were two common failure modes of Brand B modules: ⢠These modules included a Schottky diode in the power circuitry, and eight of the 18 modules examined failed because of a malfunction of this component. ⢠Five of the 18 modules exhibited general LED failures. In these cases it appeared that the LED lamp itself failed without an (obvious) external cause. Discussion While a sufficient number of failed modules were examined to have confidence that common failure modes were identified, the causes of failure of these modules might not be representative of failures among the current installed population of traffic signal modules, for several reasons. First, all of the modules examined were older than five years (they were supplied because their warranty periods had expired). Presumably, manufacturers conducted their own failure analyses on returned modules under warranty and modified their designs subsequently based on the results. As mentioned above, some design modification in response to early failures is evident in the modules examined here. Second, it is to be expected that the rate of module failure would follow the standard "bathtub" curve (Wilkins, 2002): relatively high failure rates initially followed by a period of low failure rate followed by increasing failures due to degradation. Modules that failed soon after installation were sent back to the manufacturer under warranty, so causes of those infant failures, if different than the late-stage degradation failures, would be underrepresented in this study. Third, it is possible that the samples of modules that were received from agencies were not necessarily representative samples of the modules that failed within their territories, even though this is what was requested. We believe that making several design changes could have prevented many of the failures observed in this study:
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 20 First, modules from two manufacturers used a complex boot strap circuit to start up the LED driver IC. The complexity of the design might have been due to an effort to make the module turn on quickly. Reducing the complexity of this circuit would likely increase the reliability of the module by reducing the number of parts that could fail. For example, one could provide the supply voltage to the driver IC by using a conventional power supply with a transformer, diode, and capacitor. Alternatively, all of the drive electronics could be replaced with a rectifier, capacitor, and current limiting resistor. This would result in an increase in reliability but a reduction in efficiency, however. Second, many modules failed due to the malfunction of electrolytic capacitors used for the power supplies of various ICs in the circuit. Replacing these with capacitors rated to higher temperatures and voltages or switching from electrolytic to ceramic capacitors would eliminate many of these failures. Figure 4. Arrays of 2 Ã 20 and 4 Ã 20 LEDs in parallel. Third, the LED arrays could be redesigned to reduce the likelihood of cascading failures. The LEDs in modules from one manufacturer were arranged as 2 Ã 20 (sections of 2 LEDs in parallel) and 4 Ã 20 arrays (sections of 4 LEDs in parallel) which themselves were in parallel, as shown in Figure 4. If one of the LEDs within a 2-LED-section failed, then the remaining LED in that section would see much higher current, leading to a cascade of failures. If the array were reconfigured as a 6 Ã 20 array, then a failure of an LED would be less likely to cause adjacent LEDs to fail. Fourth, high power components should be removed to the back of the main circuit board or to a daughter board. Higher temperatures lead to shorter life for LEDs (Bullough, 2003), so power components should be kept away from LEDs. In this study, a number of LEDs were
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 21 observed that had been discolored due to the heating effects of adjacent power components and smoke traces up the board were observed above power resistors. Sample Failure Analysis Reports Following are four failure analysis reports selected from those completed to illustrate the four common failure modes identified in the study. Failure Analysis #1 Brand: C. Size: 300-mm. Color: Red. Illustrating: Failed resistor in boot strap circuit. External mechanical inspection: No signs of damage. Initial application of power: Applied 120 volts alternating current (AC). Unit did nothing. External fuse was intact. Internal mechanical inspection: ⢠There were two integrated circuits: a power factor corrector and a divider chain. ⢠The drive electronics are mounted on a separate printed circuit board. ⢠Resistor R12 showed signs of excessive heat. Analysis: ⢠Resistor R12 measured as open. It was replaced with an array of resistors totaling 520 ohms at 18 watts. ⢠Replaced electrolytic capacitors on driver board. Unit does not run. ⢠Applied 12 volts direct current (DC) to the power factor corrector IC to simulate functioning power supply, Unit does not run. ⢠Drove the LED array from an external source, LED array worked 100%. Conclusion: The LED drive circuit is not functioning. The failure mode involved overheating resistor R12 (associated with the boot strap circuit). The exact cause could not be determined. General observations: The drive circuit is quite complex. Without a schematic it is quite difficult to further debug the failure mode.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 22 Failure Analysis #2 Brand: A. Size: 300-mm. Color: Green. Illustrating: Failed LEDs due to proximity to power resistor. External mechanical inspection: No signs of damage. Initial application of power: Applied 120 volts AC. Unit did nothing. External fuse was intact. Internal mechanical inspection: ⢠There were two integrated circuits: a power factor corrector and a divider chain. ⢠The drive electronics were mounted on the LED board. ⢠A power resistor (associated with pulse circuit, R30 in Figure 5) showed signs of excessive heat. There were smoke traces up the board. ⢠The LEDs above the power resistor were yellowed. There were smoke traces above the resistor. Figure 5. Power resistor and smoke residue on nearby LEDs. Analysis: It was found that four LEDs on the board were non-functional. When they were shorted to complete the circuit the unit functioned.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 23 Conclusion: The LEDs above the power resistor are four units in parallel. Two of the four had failed so the remaining two were running at twice the expected current. There were two LEDs that failed in the upper portion of the board. The circuit there had two LEDs in parallel. Thus the paired LEDs were also running at twice the expected current. It would be expected that this unit would quickly progress to 100% failure as these overstressed LEDs failed. Two LEDs exhibited significant color changes, perhaps indicating that they had been stressed or were operating at higher temperature. No other reason for the color change could be observed. It could be lot variation of the emitted color. Failure Analysis #3 Brand: B. Size: 300-mm. Color: Green. Illustrating: Failed Schottky diode. External mechanical inspection: No signs of damage. Initial application of power: Applied 120 volts AC. Unit does nothing. Power drain about 2 watts. Internal mechanical inspection: No apparent damage. Electronics were mounted on secondary card. Analysis: Replaced several capacitors with no effect. Replaced a Schottky diode and the unit worked. Attached scope to monitor current and voltage to the diode; peak voltage was about 100 volts, peak current about 1 ampere. Diode was rated at 3 amperes, 200 volts. No apparent reason why it failed.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 24 Figure 6. Defective Schottky diode. Conclusion: The Schottky diode illustrated in Figure 6 failed, causing the drive circuit to malfunction. The reason for its failure was not obvious. The diode appeared to be running well within specs. The replacement diode did not get hot. There were no transient events upon turn- on. Failure Analysis #4 Brand: B. Size: 300-mm. Color: Green. Illustrating: General LED failures. External mechanical inspection: Broken plastic near screw, later determined not to be related to the failure. Initial application of power: Applied 120 volts AC. Unit ran with some LEDs out and some LEDs flickering. Internal mechanical inspection: No apparent damage. Electronics were mounted on secondary card.
NCHRP Web-Only Document 146: Replacement Processes for Light Emitting Diode (LED) Traffic Signals 25 Figure 7. Locations of failed LEDs. Analysis: ⢠There was one string of three LEDs out, indicated in Figure 7. The LEDs were pulled. Two were functional, one had failed. ⢠There was one series string of two LEDs that was flickering, indicated by âFâ in Figure 7. The LEDs were pulled. One was functional. The other was thermally intermittent at 25 milliamperes, and flashed. Conclusion: Two individual LEDs failed or went thermally intermittent (causing the flashing). This caused the associated three LEDs to appear to fail since they were in series strings.
