Page 1 of 16 FLASH PUSLE WIDTH EFFECTIVENESS IN NOTIFICATION APPLIANCES Ken Savage TYCO Safety Products January 25, 2011
Page 2 of 16 ABSTRACT Today s fire alarm system notification equipment standards (NFPA 72, UL1971) allow a visible notification appliance to have a maximum light output pulse width of 0.2 seconds with a maximum duty cycle of 40% at a flash rate between 1 and 2 flashes per second. This implies that any pulse width at a specific candela rating that conforms to the standards is equally effective in notifying people in a building. To be an effective notification appliance, the light pulse needs to be capable of attracting the attention of people performing daily tasks. We performed a controlled experiment to determine if different pulse widths at the same candela output were equally effective in notifying people. In a room that had its ambient light level controlled, four test units with equal light output and distribution patterns were used to determine how quickly a test subject would be notified while performing a task. The four test units flashed at 1Hz and had pulse widths of 0.5, 25, 50, and 100 milliseconds respectively. The test subject was seated at one end of the room and the test unit, mounted at 80 inches from the floor, was advanced from the other end of the room towards the test subject. When the subject noticed the signal, the test fixture was stopped and the distance to the subject was measured. The test group was a broad, equally weighted mix of Males and Females of varying ages, with and without glasses. The prime observation from this testing was that there was marked difference between detection of the signal at different pulse widths.
Page 3 of 16 INTRODUCTION All current fire alarm notification strobes are based on Xenon tube technology, and the requirements for light output for visible notification appliances were developed using that technology. Recent advances in Light Emitting Diode (LED) technology brings the possibility that LEDs may be used as an alternative light source for Xenon flash tubes in fire alarm notification strobes in the near future. LED s by nature generate their light output differently than Xenon tubes, that is an LED puts out less light for a longer duration of time to achieve equivalent output. We could find little data on the relative efficacy of that type of light output. As a result Tyco Safety Products conducted tests to evaluate the relative effectiveness of different flash pulse widths. BACKGROUND Fire Alarm systems use visible appliances to notify the occupants, and especially hearing impaired occupants, in a protected facility of a fire hazard. These appliances are typically strobes that flash approximately once per second, with the text FIRE displayed on the body of the unit. The purpose of the flash is to attract the attention of occupants, in particular hearing impaired occupants, from what they are doing, and notify them of the fire hazard. Today the fire detection industry exclusively employs Xenon flash tubes to create the flash, and has for many years. Based on research on the effectiveness of strobes in notifying the hearing impaired, The National Fire Alarm Code (NFPA 72) and the Americans with Disabilities Act (ADA) have developed requirements for the output of strobes needed to signal a given area. This research was done using Xenon tubes as the light source used to notify people. The light output of strobes are measured in candela, and the equipment standards produced by the major listings agencies (UL/ULC) describe how to measure the candela rating. The formula used for candela in UL1971, is based on the Blondel-Rey equation and extracted from the standard (next page). The assumption implicit in the equipment standards is that any form of light generation that meets the equation is equally effective in notifying people in a building. We were unable to find any research that verified this assumption.
Page 4 of 16 Figure 1 Formula for determining candela output from UL1971 As mentioned, when LEDs are used as light sources in a fire alarm notification appliance, they have a very different output profile than a Xenon tube. The power capability of even the newest high brightness LEDs is limited relative to that of the Xenon tube, with current ranges in the 4 to 10W area for single element packages. This means that the LED cannot flash at the same intensity found in the peak of the xenon tube flash, and in fact is orders of magnitude less bright. Available LEDs can achieve candela ratings in the range required in mainstream fire notification applications by lengthening the pulse width again by orders of magnitude. This creates a qualitatively different type of light output pulse. When considering the use of LEDs in fire alarm notification devices we were concerned about our inability to verify that light sources, compliant with the equation, but substantively different in nature than Xenon tubes, were effective. To determine if there was any cause for concern with the ability of different flash pulse widths to attract attention, we conducted a test program using Xenon tubes and LEDs. The test program was limited, and intended only to highlight if the units performed equivalently. It also did not explore a broad variety of notification situations.
Page 5 of 16 TESTING FLASH PULSE WIDTH EFFECTIVENESS When conducting the tests, we wanted to have the test units produce the equivalent candela rating (per UL 1971) and distribution pattern with the only variable being the flash pulse width. This was accomplished using a simple parabolic reflector along with some diffusing. Test subjects were given a task and the strobe was activated behind them. This is felt to be roughly equivalent to the common occurrence in buildings where the strobe is not in the direct field of view, and notification is accomplished via reflected light. The test group was a broad, equally weighted mix of Males and Females of varying ages, with and without glasses. When we looked at the overall data, those factors (age, sex, or glasses) had little bearing on the results. LIGHT OUTPUT PROFILES Figure 2 X Plane Dispersion of Units Under Test
Page 6 of 16 Figure 3 - Y Plane Dispersion of Units Under Test The tests were conducted in a brightly lit classroom, with semi-opaque shades drawn down over windows to minimize the effects of outdoor conditions on test results. Ambient lighting was from 2 rows of fluorescents on the ceiling. The classroom floor was carpeted. Test subjects were seated at a table, facing a wall, with the chair/table edge at 2.5 from the wall.
Page 7 of 16 Ambient light level was monitored to maintain same conditions from test to test. The sensor and meter were located on the table in the area (marked on table) where the test subject was reading. The sensor and meter were removed during the test and returned to the marked location in between test runs. The UUTs were positioned at a wall opposite the subject, and moved slowly closer to the subject until they were detected. The test units were mounted on a fixtured rolling cart that held the UUTs at a height of 80 from the floor. A linear tape along the travel line from the UUT starting point to the wall facing the subject measured the distance from the UUT to the test subject. The maximum distance from UUT to test subject is 28. Figure 5 shows the overall room layout.
Page 8 of 16 21' Projector screen is drawn down below top surface of table to provide a matte surface in front of the white board surface Whiteboard 7' 12' Projector Screen Entrance Table/Chair 30' UUT path along floor is marked in ½ increments. 0 is the front (chair) edge of table 35' UUT Window (shades are down) 12' Figure 5 - Room layout
Page 9 of 16 TEST PROCEDURE The test subjects were seated at the table, and were instructed to read from provided newspapers while maintaining the reading material on the table s front surface near the front edge of the table. They were further told to stop reading, and raise their hand or speak up when they first noticed a flash or became distracted. Once the subject began reading, the UUT was turned on and moved slowly towards the subject. The UUT was aligned such that the UUT was in line with the test subject s head so that light was centered about the test subject. The closing rate was approximately 1 inch per second and reduced to approximately ½ inch per second at a point about 10 feet from the subject. Steps to remove variation in the test: Several newspapers were used as reading material as they have more of a matted, non reflective surface. Subjects were told to choose what they wanted to read in an effort to maintain their interest in reading. Subjects wore a pair of sound muffling ear phones, since the Xenon tube makes a pinging sound when it flashes. Test subjects were told to place the newspaper in a specific location on the table on the front surface and to keep the paper flat. This was done to get everyone s focus in the same location to minimize peripheral vision effects of seeing reflections off the wall in front of the subject. Light levels were checked in between test subjects to make sure light levels remained the same. To reduce learning effects, tests were conducted in the following order (3 runs 12 tests total): o First run UUT order 4, 3, 2, 1 o Second run UUT order 2, 3, 4, 1 o Third run UUT order 4, 3, 2, 1 The same operator was used to conduct all tests The test group was comprised of 8 catagories: Male over 40 yrs with glasses, Male over 40 yrs without glasses, Male under 40 yrs with glasses, Male under 49 yrs without glasses, Female over 40 yrs with glasses, Female over 40 without glasses, Female under 40 yrs with glasses, Female under 40 yrs without glasses. Each category had 7 test subjects for a total of 56. Test subjects with long hair had hair pulled back so not to affect peripheral vision Factors for age, sex, glasses, no glasses, jewelry, Dept (job function) were noted. UUTs looked the same when mounted. UUTs had same candela and light pattern UUT stand and devices were verified every day with a level to ensure mounting remained the same.
Page 10 of 16 TEST RESULTS Summary The initial test plan was to test 16 subjects and monitor the results to see if a larger sample was needed. Monitoring of the data showed that we needed a larger sample so testing was expanded to 32 test subjects over 4 days. In reviewing the data it appeared that glasses vs. no glasses may be a factor, so a second round of testing was performed to round out the sub groups to (7) in each category. 24 more people were tested in the second week for a total of 56 test subjects. Distance data varied from person to person but there was one common theme: the shorter the pulse width the better. The hierarchy of detection by an individual was Xenon, 25msec LED, 50msec LED, 100msec LED. Discussion From the raw data we have the following results: Xenon 0.5msec (UUT1) Average distance for detection: 167.87 Standard deviation of distance: 40.22 Range of all subjects: 76.75-303.25 25msec LED (UUT2) Average distance for detection: 140.46 Standard deviation of distance: 37.02 Range of all subjects: 64.25 267.25 50msec LED (UUT3) Average distance for detection: 76.99 Standard deviation of distance: 35.25 Range of all subjects: 0 178.25 100msec LED (UUT4) Average distance for detection: 8.19 Standard deviation of distance: 16.63 Range of all subjects: 0 71.25
Page 11 of 16 Looking at the overall data, distances varied by individual test subject and other factors for age, sex, glasses or jewelry had little bearing on the results. The Xenon tube is on for about 500usec, LED s are 25, 50, and 100msec. So when we look at the data as an overall group for each UUT, we can see that as pulse width increases, detection decreases. In most cases the test subjects failed to notice the 100msec pulse. Looking at the overall data normalized to the Xenon flash we get: LED Average Median 25msec 85.12% 84.46% 50msec 46.63% 45.65% 100msec 4.62% 2.63% The spread or range of detection distance also varied by test subject without regard to age, sex, glasses or jewelry. To get an overall view the spread for a particular test subject and UUT, the spread was taken as a percentage of average distance of three runs. The results from all 56 test subjects were then averaged together. From this you can see that the shorter the pulse width the smaller the variation in detection. The 100msec average numbers get skewed from the large number of no detections, the median result shows it better. Average Median Xenon 0.5msec 16.82% 16.70% LED 25msec 21.02% 21.84% LED 50msec 54.21% 41.58% LED 100msec 78.29% 15.00% Box plot data shows the range of distance detection for each unit without regard to other factors. Shorter pulse width gives earlier detection. 300 Boxplot of Distance-2 250 Distance-2 200 150 100 50 0 1 2 UUT-2 3 4
Page 12 of 16 CONCLUSIONS These tests were set up to have all the test units at the same candela rating per the current accepted standards. When selecting test subjects a conscious effort was made to keep each test group broadly represented (for example for the over 40 group tried to get a distribution of 40 s, 50 s, 60 s). Also an effort was made to get test subjects from many different job functions as different jobs to some degree utilize different traits. From this testing two major relationships were apparent for an individual: The shorter the pulse width the earlier the detection The shorter the pulse width the smaller the detection variation With regards to the current industry standards a troubling observation from the data was that as pulse width increased detection decreased. Even though the strobe units that were being tested all had the same light output and distribution, the strobes with the longer pulse width were much less detectable. The 100msec was essentially invisible to the test subjects. This is troubling as the current standard allows up to 200msec for a pulse width. There were even a few cases where the 50msec pulse width went undetected. There seems to be an inherent trait for human test subjects, when the total light output is the same, to notice when that light is presented as shorter bursts of light energy than longer ones. The current standards for visible appliances should be changed to address the lack of effectiveness for longer pulse widths.
Page 13 of 16 References UL1971, Signaling Devices for the Hearing Impaired, 3 rd Edition, 2006 revision CAN/ULC-S526-07, Visible Signal Devices for Fire Alarm Systems, including Accessories, 3 rd Edition
Page 14 of 16 APPENDIX UNIT UNDER TEST DESCRIPTIONS Xenon tube based strobe Based on a SimplexGrinnell 4906-9102 110 CD Ceiling-Mount Strobe. A ceiling mount strobe was selected because the PCB layout is such that the Xenon tube mounts on the opposite side of the electronics, thus making it easier to add a different reflector and other modifications to match light output distribution with the LED. In addition to the parabolic reflector and diffusion layer, an area of the flash tube is masked off with black electrical tape to make the tube more of a point source and to locate the point source at the bottom of the reflector.. The Xenon tube unit is mounted parallel to the surface of the wall. The Xenon unit puts out the rough equivalent of 14.6CD. See the unit output graphs figures 2 & 3. This unit is referred to as UUT1.
Page 15 of 16 High Brightness LED Prototype LED used: Nichia NCSW119T. This LED has a light output distribution pattern that varies as the cosine of the angle formed to a line perpendicular to the mounting surface of the package. This light output pattern is referred to as Lambertian. The LED source is mounted in the same reflector/diffuser arrangement as the Xenon tube. The LED is mounted at the bottom of the reflector and the diffuser layer is slightly modified to facilitate a light distribution pattern more closely matched to the xenon tube. The LED unit is mounted parallel to the surface of the wall, and support electronics were designed to regulate the current through LED and vary the pulse width at about 25msec/1.4A, 50msec/0.7A, and 100msec/0.35A respectively. The settings are switch and jumper selectable. One characteristic of the LED is that light output drops to near zero as the angle of viewing approaches 90 0 from the perpendicular. This characteristic has been negated with the selected reflector/diffuser choice which keeps the bulk of the light energy between +/-20. The overall distribution goes out to +/-65 however this does not meet the UL criteria. The scope of this test does not include creation of lensing or reflectors that would redirect light to address this deficiency. The unit has an adjustable pulse width that can be set for 25ms, 50ms, and 100ms pulse widths of the prototypes were set as follows: UUT2 (LED Setting #1) Flash Pulse width of 25ms rough equivalent of 14.7 CD UUT3 (LED Setting #2) Flash Pulse width of 50ms rough equivalent of 14.6 CD UUT4 (LED Setting #3) Flash Pulse width of 100ms rough equivalent of 14.7 CD The candela rating profiles of the 4 test settings can be seen in figures 2 & 3.
Page 16 of 16 EQUIPMENT SETUP An International Light IL1700 Radiometer was used to monitor ambient light levels and establish UUT light output ratings. A Simplex 4905-9914 Sync Cube provides the flash trigger for the xenon test unit (UUT1) A timer circuit is used to control the LED Flash rate and pulse width. Pulse width is varied but remains at 1Hz. Lab Power Supplies provide 24V and 8V unit power for the Xenon and LED UUTs respectively. Kenwood PR36-3A Regulated DC supply Kenwood PR36-3A Regulated DC supply 24VDC 8 VDC - + - + Selectable pulse width 25msec, 50msec, 100msec. Fixed 1Hz cycle LM3485 LED driver demo Selectable LED drive 350mA, 700mA, 1.4A + - LED Reflector assy. UUT 2-4 On/Off SimplexGrinnell 4905-9914 Sync Cube + - On/Off UUT1 Modified SimplexGrinnell 4100-9102 Ceiling mount strobe Figure 4 - Block Diagram - Xenon and LED UUT