Lighting Research Center Lighting Research Center
 

Volume 4 Number 3
Copyright ©2000 Rensselaer Polytechnic Institute

La vita è bella

By John D. Bullough

It's not likely that Roberto Benigni will ever make any feature films extolling the virtues of long-life lamps and lighting systems. However, many in the lighting industry recently have been more than willing to discuss this aspect of lighting, and it appears that many more will do so in the future. A growing sophistication about the economics of lighting has emerged, and with it, so has an understanding that luminous efficacy, or lumens per watt, is not the only important variable in designing and maintaining cost-effective, quality lighting.

Lamp life is one of those other important variables. Life is increasingly a driver in the development of new lamps and lighting systems. More lamp manufacturers are using life to distinguish their own products from those of their competitors. But is life simply another ruler by which we can measure lighting? How important is it really?

Growing interest

In just the past few years, interest in the issue of life has grown. According to a survey of magazine articles in the LEXIS-NEXIS database from 1996 to the first half of 1999 (see Figure 1), lamp life is becoming a more popular topic. During the first half of 1996, fewer than 1% of these articles mentioned lamp life. By comparison, more than 3.5% of the articles in the first half of 1999 discuss lamp life. This steady increase demonstrates a growing awareness of life-related issues.

What's life?

We've become accustomed to lamp packaging that states the manufacturer's determination of lamp life, called rated life, usually in hours, of various light sources. The most straightforward interpretation of these ratings is arguably that they tell us how long the lamp will operate before it fails (“burns out”). But the definition of life is different for different lamp types. Furthermore, the conditions under which lamp life is measured can have a large impact on a lamp's actual service life. To make the issue even more complex, the standard testing conditions used to measure lamp life rarely correspond to the real-life conditions under which people use lamps. Even with these limitations, however, rated life provides important information for predicting the overall economic cost and impact of a lighting system.

Incandescent lamps

We test incandescent lamp life by operating a sample of lamps continuously in a specified position and at a specified voltage (within a range of 0.25%). The lamps may be occasionally removed so their individual photometric characteristics can be measured. The number of burning hours at which half the lamps have failed is considered the rated life of the lamps. Any two similar incandescent lamps with the same rated life, however, can have very different actual operating lives, even under identical conditions. The typical incandescent mortality curves in Figure 2 show that there can be a very large spread in hours between early-failing lamps and long-lasting lamps of the same rated life.

Reprinted with permission of the IESNA

As pointed out by the Illuminating Engineering Society of North America's IESNA Lighting Handbook, 9th edition (2000), incandescent life, not including failure due to impact or shattering, depends largely on evaporation of the lamp's tungsten filament. Incandescent lamps generate light by passing a current through a tungsten filament, which heats up and in turn emits light. In the process, the tungsten slowly and gradually evaporates. The glass bulb blackens as evaporating tungsten particles deposit on it, reducing light output. This evaporation thins the filament until it reaches a point where it breaks, disrupting the electrical circuit and resulting in a burned-out lamp. Typical incandescent lamp life ranges from several hundred to 1500 hours.

In the 1960s, manufacturers developed halogen lamps by adding iodine or some other halogen-family chemical to improve tungsten filament lamp life. The gas and filament are enclosed in another small bulb that allows the gas pressure to be increased. This higher pressure slows the tungsten's evaporation and reduces bulb blackening. Some tungsten-halogen lamps last more than 5000 hours.

Running incandescent lamps at a voltage different than the voltage the lamp is designed to use, called rated voltage, greatly affects their life. As shown in Figure 3, operating a lamp at 120% of its recommended voltage will cut its life down to 10 to 15% of its rated life. On the other hand, reducing the operating voltage to just 90% of the recommended voltage will increase life by a factor of 4. This is useful when designing lamps for traffic signals, indicators, and other applications where long life is important. Manufacturers can increase lamp life by designing lamps for a higher than necessary voltage. For example, they can increase the filament's resistance so that the lamp runs cooler. If a filament is longer than necessary for the voltage, its resistance is higher. The filament can also be thicker if it is made even longer. Due to the lower operating temperature and the thicker filament, the lamp life will be longer. The trade-off, however, is lower efficacy and light output than at the rated voltage. A traffic signal lamp has a rated life of 8000 hours because it is designed for a much higher voltage than the actual voltage used to power the lamp.

Reprinted with permission of the IESNA

Fluorescent lamps

Fluorescent lamps require a different life test than incandescent lamps because different factors affect their life. Fluorescent lamps are tested while operating at a specified temperature (25°C/77°F) on a continuous 3-hour-on, 20-minute-off cycle, with a standard ballast circuit that controls the current. As with incandescent lamps, rated life is the elapsed number of operating hours at which half of the lamps in a sample have burned out, which means that the actual life of individual lamps varies. The typical rated lives of fluorescent lamps are between 7500 and 20,000 hours significantly longer than incandescent lamps.

The length of the on-off cycle used in fluorescent lamp life testing is extremely important. Unlike incandescent lamps, where operating the lamp usually causes it to burn out, the main factor in the end of life for fluorescent lamps is loss of the electrodes' emissive coating. While the emissive coating evaporates slowly as the lamp operates, starting the lamp speeds up the process. When a fluorescent lamp starts, the ballast applies a high voltage to the lamp electrodes, eroding the electrode coating. Frequently switched fluorescent lamps lose their electrode coatings more rapidly and fail to start sooner than infrequently switched fluorescent lamps. Figure 4 shows the increases in fluorescent lamp life if a lamp operates on longer switching cycles than the standard 3-hour-on, 20-minute-off cycle. Figure 5 shows reductions in life for compact fluorescent lamps caused by operating cycles much shorter than the 3-hour-on, 20-minute-off testing cycle. These were the results of tests supported by the Electric Power Research Institute, the New York State Energy Research and Development Authority (NYSERDA), and the National Lighting Product Information Program (NLPIP).

Reprinted with permission of the IESNA

Because of the relatively long life of fluorescent lamps and the amount of time required to test a lamp's life (a 20,000-hour lamp can take nearly three years to test), recent research examined whether accelerated fluorescent lamp testing can be useful in predicting lamp life and other lighting system performance characteristics. A test of several fluorescent lamp and ballast combinations found that for rapid-start fluorescent lamps operated on a 5-minute-on, 5-minute-off cycle, an important parameter in predicting lamp life was the ratio of the hot electrode resistance (RH) to the cold electrode resistance (RC). When this ratio, RH/RC, was close to 4.25, the lamps lasted longer (see Figure 6 from “Compatibility Testing of Fluorescent Lamp and Ballast Systems”; by Yunfen Ji, Robert Davis, Conan O'Rourke, and Edmund Chui of the Lighting Research Center, published in The Proceedings of the Institute of Electrical and Electronic Engineers [IEEE] 32nd Industry Applications Society Annual Meeting, 1997). RH/RC indicates whether the lamp and ballast are compatible, which is an important—but not the only—lamp life predictor. The Empire State Electric Energy Research Corporation (ESEERCO), NYSERDA, and NLPIP sponsored this research.

©1997 IEEE

The RH/RC ratio does not apply to instant-start fluorescent lamps because their lamp electrodes are not pre-heated. Tao Yin of the Lighting Research Center showed in her graduate thesis that lamp life for instant-start lamps that are operated on a 5-minute-on, 5-minute-off cycle correlated with the integrated starting voltage. Integrated starting voltage is calculated by measuring the electrode voltage during the time between the application of power and the lamp start (see Figure 7).

Each of these studies used the same lamp types with different ballasts and found lamp life differed with individual ballasts. This demonstrates the importance of lamp-ballast compatibility, that is, matching the lamp's characteristics with those of the ballast that starts and operates it. The same lamp operated on different ballasts can have very different life characteristics. This latest research will likely influence design of tests for fluorescent lamp life using rapid-cycle switching. Such tests would allow more rapid information dissemination about lamp life and compatibility issues between fluorescent lamps and ballasts. However, there is currently no way to determine lamp and ballast compatibility without taking these measurements for individual lamps and ballasts.

The lighting industry has attempted to overcome some of the inherent limitations of the fluorescent lamp electrode system by developing several electrodeless lamps that are now on the market. Philips Lighting's QL lamp, OSRAM SYLVANIA's Icetron, and GE Lighting's Genura lamp are examples of electrodeless lamps. They operate via electromagnetic energy to excite the gas fill, rather than an electric field created between two metal electrodes as in a conventional fluorescent lamp. This creates a mercury discharge that in turn excites phosphors and generates light in the same way as a fluorescent lamp. Such lamps, because they have no electrodes, are immune to the emissive coating losses of conventional fluorescent lamps, and manufacturers have claimed rated lives of more than 50,000 hours and as high as 100,000 hours.

Electrode life is not the only important factor in long-term operation of fluorescent lighting systems. The phosphors that coat the inside and the glass that makes up the lamp tube degrade after long use. Several factors affect phosphor and tube degradation, including current density, phosphor type, ultraviolet flux, wall temperature, and wall material. As a result, the light output of a fluorescent lamp will decrease to between 60% and 90% of its initial light output after 10,000 hours and will continue to decrease. Most lighting designers factor this characteristic, known as fluorescent lamp lumen depreciation, into their designs, but installations that are irregularly maintained or relamped may include several lamps that appear to be working fine, but in reality are producing close to half the amount of light originally intended. This could be especially true for electrodeless lamps, which have no electrodes that can fail earlier.

High-intensity discharge lamps

Life testing for high-intensity discharge lamps such as mercury, metal halide, and high-pressure sodium is much the same as for fluorescent lamps, except that the standard operating cycle is different. Because these lamps are often used in applications such as industrial facilities, warehouses, or parking lots, where they turn on at the beginning of their use period and stay on for 8 hours or more, the switching cycle is 11 hours on, 1 hour off. For metal halide lamps, which increasingly appear in conventional indoor lighting applications because of their good color-rendering properties, an 11-hour-on, 1-hour-off cycle may be inappropriate for predicting lamp life because of frequent switching.

As with fluorescent lamps, high-intensity discharge lamps often fail due to losses in the electrodes' emissive coating. Metal halide lamps have a particular obstacle, because the electrodes need to be compatible with the chemicals in the metal halide arc stream. Because metal halide is not compatible with the materials used to coat electrodes, metal halide lamps generally use bare tungsten electrodes, which erode more rapidly than coated tungsten electrodes. High-pressure sodium lamps also experience electrode failure as they approach end of life and often cycle on and off before they finally fail for good. Typical mercury lamp rated life is 20,000 to 24,000 hours; metal halide lamp rated life is usually much shorter, from 7500 to 15,000 hours; and high-pressure sodium lamps have a rated life of around 24,000 hours.

All three types of high-intensity discharge lamps experience lamp lumen depreciation from deposits of electrode material on the wall of the arc tube and, in the case of metal halide lamps, changes in the chemical makeup of the arc stream. This is why many metal halide lamps have been reported to undergo noticeable shifts in lamp color as they age. This may be important if uniform color appearance is desired in a particular lighting installation: If a single lamp is replaced, it may visibly differ from older lamps. Because all high-intensity discharge lamps experience some lumen depreciation, a regular maintenance and relamping schedule should be followed. As with fluorescent lamps, a high-intensity discharge lamp may not be performing well even when it appears to the casual observer to be on.

Light-emitting diodes

Light-emitting diodes (LEDs) are solid-state semiconductor devices that produce light. In the past, they were used primarily in applications such as indicator lights that require just a small spot of light, but more recently, they have been used for applications such as exit signs and traffic signals. Because these light sources are relatively new to the lighting industry, standard definitions of lamp life don't exist. They don't fail in the sense that other sources do. Over time, however, their light output decreases until they are no longer useful. Thus, no life test has yet been determined for LEDs. As shown in Figures 8 and 9, LEDs produce only 50 to 70% of their initial light output by 100,000 hours under specific conditions. The light output of the first LEDs used in exit signs and traffic signals, which used aluminum gallium arsenide (AlGaAs) materials, was cut in half after just 15,000 hours in some cases.

Courtesy of Agilent Technologies

More recent technology using aluminum gallium indium phosphide (AlGaInP) and indium gallium nitride (InGaN) have proven more stable for long-term light output. These LEDs still must be closely monitored for light output, especially in exit signs and traffic signals, whose visibility is critical to people's safety.

Because LEDs require much smaller voltages of direct current, another factor that reduces the apparent long life of LEDs is the need for auxiliary electronics and equipment to house and operate these sources. Because electrical power commercially available in the U.S. is in the form of alternating current, LEDs require direct current converters. Such devices may have rated lives significantly shorter than the LEDs with which they are used, so specifiers need to consider the rated life of the whole product or system rather than just the potentially promising long rated life of the LEDs. Higher voltage and high temperatures can also increase lumen depreciation in LEDs.

Economic implications

Life is an important factor in determining long-term economic impact of a system. Consider the simple concept used by the IESNA known as the “cost-of-light.” This metric estimates the cost of providing lighting and is given in units of dollars per million lumen-hours, which considers the amount of light required and the cost of both operating lighting systems for that period of time and replacing lamps at the end of their life. According to the IESNA Lighting Handbook, 9th edition, the cost of light can be given by the following simple equation:

U = (10/Q) x [(P + h)/L + WR]
Where:
U is the cost of light per million lumen-hours in dollars
Q is the mean lamp lumens
P is the price of the lamp in cents
h is the labor cost to replace a single lamp in cents
L is the average rated lamp life in thousands of hours
W is the input power of the lamp in watts (including ballast power if appropriate)
R is the cost of energy in cents per kilowatt-hour

Imagine two hypothetical lamps that are identical in every way except for their rated life: Lamp 1 has a rated life of 1000 hours, and lamp 2 has a rated life of 10,000 hours. In this example, let's consider the cost of light for a lighting system using the following parameters:

Q = 3000 lumens
P = $5 or 500 cents
h = $5 or 500 cents
L = 1 for lamp 1 (1000 hours)
L = 10 for lamp 2 (10,000 hours)
W = 40 watts
R = 10 cents per kilowatt-hour

For the first lamp, with a rated life of 1000 hours, the cost of light per million lumen-hours is $4.67. For the second lamp, with a rated life of 10,000 hours, the corresponding cost of light is $1.67 —slightly more than one-third the cost of light for the first lamp. Although this example and the cost-of-light calculation in general do not consider the time value of money, the effects of inflation, or other variables that could drastically affect the economics of a lighting installation, it does show how life can be significant even without considering energy efficiency (remember, our example used hypothetical lamps with equivalent luminous efficacies).

In general, however, it is much more difficult to realistically predict potential maintenance savings. A survey of municipalities conducted by the California Energy Commission about traffic signal use (City and County Responses to LED Traffic Signal Survey, April 1999), found that relatively few municipalities identified potential maintenance savings as a reason for using LED traffic signals rather than incandescent signals. The maintenance costs associated with traffic signals are a direct consequence of the different lives of the LED and incandescent lamp technologies used in these types of signals.

Compared to the 80% to 90% energy reduction achievable by using LED traffic signals as replacements for incandescent signals, any additional savings due to decreased maintenance requirements might seem to be merely frosting on the cake. However, estimated maintenance savings from several cities, including St. Paul, MN, and Boulder and Denver, CO, range from 51% to 88% of the actual savings in dollars that are attributable to energy savings alone.

What next?

What does the future hold with respect to lamp life? As Figure 1 indicates, lighting specifiers' and the lighting industry's interest in this issue will grow. This interest will fuel research and marketing efforts to enhance the benefits possible from increased life, including
  • Developing fluorescent lighting systems— lamps and ballasts, not just lamps, whose starting parameters (RH/RC ratio for rapid-start or starting voltage-time characteristics for instant-start) will encourage longer life.
  • Improving metal halide lamp technology to increase both operating life with improved electrode materials and stability of light output and color throughout operating life.
  • Refining the design of lighting products that use LEDs, so the LEDs and the electronic gear that drives them will become more integrated and life will incorporate the whole system.
  • Greater sophistication and accuracy in predicting the economic benefits of lighting systems with longer life, so specifiers will more explicitly understand life as an important parameter to consider when selecting lighting.

Certainly the phrase “life is beautiful” has a great deal of validity in the lighting industry!




Rensselaer Polytechnic Institute
LRC Intranet Web mail Lighting Research Center