Open Access
Add to BibSonomyAbstract
In a psychophysical forced-choice experiment, observers’ ability to detect illuminance differences was found to be 7.4% of the initial illuminance. When matching the illuminance of one space with another, observers’ acceptance of illuminance differences was 17.8% to 19.1%. Lighting control systems with resolutions between 14.8% and 17.7% were found to have greater usability than others. A new approach to step-dimming leverages knowledge of the detectability and acceptability of illuminance differences, as well as usability, to reduce lighting energy consumption. This method can reduce lighting energy consumption more than continuous dimming.
© 2016 Optical Society of America
1. Introduction
Previous research has investigated strategies for reducing the energy consumed by lighting by reducing the illuminance of electrically lit spaces [1, 2]. Research suggests that occupants cannot detect illuminance reductions of up to 20% [3]. Although earlier research results were technology-neutral, much of this research was conducted with fluorescent lights. Dimmable fluorescent products in the market are fluorescent luminaires with electronic dimmable ballasts. The relative high cost of fluorescent dimming ballasts prevented widespread adoption of those research findings in real illumination applications. Dimmable fluorescent systems were historically used in high-end commercial spaces.
With the development of dimmable light-emitting diodes (LEDs) and digital control technologies, dimming in commercial spaces has become easier and more economical. LED dimming technology is well-established and commercially available, as long as users choose LED lamps, drivers and control systems compatible with each other. Dimmable LED lighting systems have already been used in a large number of lighting application areas to provide general lighting. Additionally, the increasing need for sustainable building design encourages increased use of dimmable LEDs with automatic control systems and daylight harvesting systems. Those advanced control systems could be designed with very high resolution, with “resolution” referring to the smallest change that can be produced by a lighting control system (LCS). Many lighting products are compatible with the DALI (Digital Addressable Lighting Interface) control protocol, which provides 256 levels of brightness between off and the maximum level [4].
If a lighting control system was to automatically reduce illuminance slightly, such that the change was either undetectable or at least acceptable to the user, energy consumption would be reduced in two ways: the power driving the light source would be reduced and the LEDs would exhibit higher efficacy, from reduced current and thermal droop. LED current droop, which is the reduction of luminous efficacy with increases in current, means that LEDs operate with relatively low efficacy at the relatively high drive currents needed for illumination applications [5]. Since increased LED junction temperature also leads to a reduction in internal quantum efficiency, reducing the heat generated by the LEDs would also increase efficacy [6]. When users of lighting control systems increase illuminance in a space, the luminous efficacy of the LED lighting system is reduced. When users decrease illuminance, the efficacy of the lighting is increased.
Thus, attention has been drawn back to illuminance reduction again for energy conservation. However, the development of lighting control systems that operate in this manner requires a better understanding of users’ ability to detect differences in illumination, their acceptance of illuminance differences, and their satisfaction with LCSs of various resolutions. This project combines psychophysical experiments with a questionnaire to investigate the detectability and acceptability of illuminance differences and applies the results in the design of lighting control systems, by proposing a series of dimming curves.
2. Previous studies
2.1 Detectability
Numerous previous works explored the human visual system’s ability to detect differences in brightness. Initial research, dating back to as early as the 1940s, in this field mainly focused on human visual perception [7, 8]. Most research directly applied to lighting on this topic is more recent. Akashi and Neches proposed a solution for load shedding by reducing illuminance. They investigated four dimming curves and reported that “dimming curvature,” which refers to the shape of the dimming curve, “has little influence on the detection of illuminance reduction” and that 50% of their experimental participants could detect a 15% reduction of the initial illuminance [1]. In their experiment, participants were asked whether the illuminance increased, decreased or remained the same. This method, although direct and efficient, relies heavily on the participants’ subjective reports of what they perceived. When participants reported the detection of change, they were usually very confident with their judgments. This means that an observer’s ability to detect differences was likely underestimated with this method. Psychophysicists have found that, by using forced-choice method, many observers could detect very weak signals, which they claimed they could not detect [9].
Recently, a study conducted by Shikakura’s team, on the “minimum ratio of perceiving fluctuation” systematically examined the detectability of illuminance differences [10]. They reported that the detectability of illuminance decreases and illuminance increases were 8% and 6% respectively, when no visual task was involved in the experiment. The first experiment in this project suffers from two issues: firstly, a neutral response option - “cannot determine”- was provided, meaning that participants were not required to make a positive or negative judgment. However, psychophysicists have found that some stimuli, which observers claim cannot be detected, can in fact be detected correctly in forced choice experiments. The authors pointed out that the stimulus detection threshold was probably overestimated by offering the neutral option. They suggested the use of a classical forced-choice psychophysical method, in which participants would be forced to choose a positive or negative response without a neutral option [9]. Another issue is the possibility of the false detection of stimuli in Shikakura’s experiment. When participants reported changes in illuminance without indicating whether it was increased or decreased, there was no way to determine whether observers’ reports were representative of their percepts or the result of response errors. Shikakura’s team noticed this issue and improved their second experiment by asking participants whether illuminance was increased or decreased. All false detections were treated as errors when analyzing the data.
2.2 Acceptability
Unlike the classic studies of detection, which date from over a hundred years ago, research on users’ acceptance of various lighting conditions gained more attention in recent years. Researchers reported that illuminance reductions of up to 30% could be accepted by occupants [1–3]. Akashi and Neches suggested that 50% and 80% participants accepted illuminance reductions of up to 40% and 20%, respectively. They stated again that the shape of dimming curve had little impact on the acceptability of illuminance reductions [10]. In a field study on office lighting, author suggested that illuminance reductions of up to one third of the initial illuminance had no long-term impact on users’ satisfaction [2]. However, most results about user acceptance use verbal reports of participants or subjective ratings. Well controlled psychophysical experiments would strengthen the conclusions that could be drawn.
Resolution, which is the brightness difference produced by each control input action, is essential for every LCS. For an efficient and user-friendly LCS, the value of resolution should be between the users’ illuminance difference detection threshold and acceptance threshold. Therefore, user acceptability of the illuminance differences, which is constrained the LCS resolution, is of equal importance. The concept of usability is widely used in studies of human-computer interaction (HCI) and can be used to measure user acceptance of the resolution of LCSs.
According to ISO 9241 Part 11-Guidance on usability, usability is defined as “the extent to which a product can be used to achieve goals in three areas: effectiveness, efficiency and satisfaction” [11]. It is a property of any interactive digital technology, revealing whether the system is simple and pleasant to use. Accuracy of task completion can be used to measure effectiveness, while efficiency can be evaluated by multiple parameters, such as task completion time, learning time, etc [12]. Some studies only take effectiveness and efficiency into account because these two variables are quite objective and relatively easy to measure. However, subjective impressions and preferences are also useful for evaluating users’ satisfaction. Subjective and objective measurements of the usability are not always highly correlated [13].
3. Methods
When users start controlling luminaires with a LCS, regardless the interface of the control systems, they may look at luminaires directly if the luminance is relatively low. However, when the luminance or the contrast with background is extremely high, they often look at the illuminated surfaces. To develop a general design guide of control systems, both illuminance and luminance resolution need to be explored. As the first part of a systematic study, this project focused on illuminance, investigating the detectability and acceptability of illuminance differences, as well as the impact of LCS illuminance resolution. This study of detectability employed a classic psychophysical method, the two-alternative-forced-choice (2AFC) method, while the methodology used to investigate users’ acceptance of illuminance differences was another psychophysical method – the adjustment method. The acceptance of LCSs of different illuminance resolutions was evaluated by the usability of systems with various resolutions. Specifically, usability was assessed by the effectiveness and efficiency of matching tasks with the given LCS, as well as users’ satisfaction, expressed by subjective ratings in questionnaires.
3.1 Experimental setup
In the Lighting Lab at the University of Sydney, two identical lighting booths were constructed, as shown in Fig. 1 and Fig. 2. Each booth was constructed using four matte white wall panels with a reflectance of 84%, measured by a Konica Minolta CM-2600d spectrophotometer. Since this experiment studied control resolution for horizontal illuminance, a rear wall panel was not included to limit the distraction of vertical illuminance. Two warm white LED light engines were mounted in the middle of the ceiling of each 1.0 m x 1.0 m x 1.4 m booth. The LED light engines were powered by two programmable power supplies. Mean illuminance was measured at numerous drive currents and this data was used to generate an equation that described the relationship between illuminance and current. Participants sat in front of the pair of booths, using a button box to express their response or to control the brightness of the lights. Their feedback was recorded by MatLab.
Fig. 1 Experimental setup. Author in participant seat only for illustration; only naïve observers participated in experiment.
3.2. Detectability experiment (2AFC method)
In each trial of the detectability experiment, one lighting booth was randomly illuminated with one of the five reference illuminances (from 100 lx to 500 lx, in 100 lx steps). The other booth was illuminated with the test illuminance, which was the reference illuminance +/− one of seven difference levels (5.7%, 11.4%, 17.1%, 22.9%, 28.6%, 34.3%, 40% of the reference illuminance). The maximum difference level of 40% was chosen to ensure that the difference in illuminance between the reference and test booths could be easily detected by all users. Seven testing levels, which were the maximum difference level divided into seven evenly space intervals, were used so that sufficient data was available for curve fitting and that all trials for each participant could be finished in a reasonable time. Whether the difference between the reference and test illuminance was positive or negative was randomly assigned for each trial. For each participant, each difference level was examined ten times under each reference illuminance condition. All trials for one reference illuminance were conducted together in one session. Participants adapted to the reference illuminance for ten minutes before each session.
Participants were instructed to press the left or right button to report which of the two booths appeared brighter on the horizontal (bottom) plane. Therefore, each subject completed 350 trials and their proportion of the correct judgments were recorded.
Their judgments were collected and plotted as a function of the illuminance differences. In a two-alternative-forced-choice experiment, 50% correct responses represent the chance level of performance. Therefore, in this experiment, 75% correct responses represented the difference detection threshold. Thus, the illuminance difference at which 75% judgments were correct was the just noticeable difference (JND) for that reference illuminance condition.
According to the well-known Weber’s Law, the relationship between the JNDs and brightness (illuminance in this case) should be theoretically linear [14]. The Weber’s Law fraction, which is the quotient of JNDs and the reference illuminance, varies under different circumstances [7]. This experiment measured this fraction for the experimental conditions, in which users compared the illuminance of two horizontal working planes. The fraction obtained in the experiment indicates users’ ability to detect illuminance differences.
3.3 Acceptability experiment (adjustment method)
In the acceptability experiment, the reference booth was illuminated with one of the five reference illuminances and test booth started with a pseudo-random illuminance. The participants’ task was to match the illuminance of test booth with the illuminance of the reference booth, using a button box. Participants’ acceptance of illuminance differences between two booths might be impacted by their ability to detect differences, psychological variables, and the design of the control system. For instance, participants may compromise and accept a greater difference when the control system is tedious to adjust. It was hypothesized that users would accept an illuminance difference that is suprathreshold (greater than difference detection threshold) when using a control system to match one brightness to another. Thus, at the end of each matching task, the illuminance difference between the reference and the test booth was recorded as the matching error, which indicated users’ acceptability of illuminance differences when they using LCSs themselves.
Theoretically, the resolution of a LCS may impact user’s matching error. If users are sensitive enough to the illuminance difference between two booths, then the higher resolution system would allow them to more precisely match the test to the reference illuminance. This means that smaller matching errors should be obtained from the experiment when a higher control resolution is used. However, in real life, when resolution is extremely high, users may grow frustrated with extent of choice they have or number of actions required to reach their desired illuminance and give up adjusting the LCS. To test this speculation, seven levels of resolution, which were determined by the product of a multiplication factor and the illuminance difference detection threshold (JND), were examined during the experiment. Although a reasonable LCS resolution was assumed to be greater than a JND, one of the multiplication factors was set to be 0.8, to investigate the usability of LCSs with extremely high resolution. In that case, participants were expected to be unable to detect the illuminance differences when pressing a button once. The other multiplication factors were 1.2, 1.6, 2.0, 2.4, 2.8, and 3.2. Each observer was required to complete four illuminance-matching trials for each combination of multiplication factor and reference illuminance. The multiplication factors and reference illuminances were randomly assigned in each trial.
Participants pressed specified buttons to initiate and end a trial, so that the task completion time could be recorded. The numbers of button presses, used to increase or decrease the illuminance, were recorded as well. The starting illuminance of the test booth was pseudo-randomly assigned for each trial. Approximately 50% of trials were assigned with pseudo-random illuminances lower than the reference illuminance; the others were assigned with illuminances higher. The number of button presses was normalized by the absolute difference between the starting test illuminance and reference illuminance, as shown in Eq. (1). Speed, computed as the inverse of the average time per button press, was normalized as well, as shown in Eq. (2).
where- N is the normalized number of button presses
- Nall is the total number of button presses in one trial
- ETs is starting illuminance of the test booth
- ER is the illuminance of the reference booth
- S is the normalized speed of button presses. It is the inverse of the average time consumed between two button presses
- Nall is the total number of button presses in one trial
- ETs is starting illuminance of the test booth
- ER is the illuminance of the reference booth
- CT is the completion time for each trial
After completing all experimental trials for one LCS resolution, each participant was instructed to complete an after-scenario questionnaire (ASQ), as shown in Table 1.
Table 1. The After-Scenario Questionnaire (ASQ).
3.4 Proposed step-dimming curves
Based on the results from the detectability and acceptability experiments, a series of step-dimming curves were developed for LCSs of various resolutions. Simulated illuminance was initially set to the maximum (drive current of 1.0 A for the LED light engines considered) in MatLab and illuminance was gradually reduced, in steps dictated by the control resolution. According to the Weber’s Law, the perception of a just noticeable difference in illuminance is proportional to the mean illuminance. Therefore, the change in illuminance between two dimming steps was the resolution multiplied by the illuminance. These calculations were repeated until the current reached 1% of the maximum (0.01 A). Normalized illuminance, which is the percentage of the maximum illuminance, was plotted for each step. Power consumption as a function of dimmer setting was measured during these iterative calculations.
The resultant dimming curve was compared to common continuous dimming curves used in commercially available lighting products. MatLab was used to generate the proposed step-dimming curve first and then to compare the proposed curve to continuous dimming curves. The two most popular dimming curves, a logarithmic dimming curve and a Square Law curve, were compared with the proposed function. As manufacturers do not provide equations for their continuous logarithmic curves, in this project, the continuous logarithmic curve was mathematically expressed by the equation corresponding to the base function of the proposed step-dimming curve. As this type of function is commonly referred as a “logarithmic dimming curve” in the lighting industry, that term is used here. The Square Law curve is reported to be generally rated favorably by end users [15, 16]. Both the logarithmic and Square Law curves are commonly used in products from lighting control manufacturers, such as Lutron and Dynalite [4, 17].
3.5 Observers
Twelve observers, under 50 years of age, took part in this study. Seven are men and five are women. During the experiments, they wore glasses or contact lenses, if they usually do.
4. Results
4.1 Detectability and acceptability
The JNDs for each of the five illuminance conditions were plotted in Fig. 3 to obtain simple linear regression line. It can be seen that illuminance JNDs increased proportionally as the illuminance increased, which agrees with Weber’s Law. The goodness of fit (r2) of this linear model is 0.96, while the regression coefficient was computed to be 0.074, which is the Weber’s Law fraction for the experimental conditions. This result suggests that observers can barely detect an illuminance difference when the difference is 7.4% of the given illuminance. In other words, observers’ detection threshold for illuminance differences was 7.4%.
Although this result revealed that observers were able to detect small illuminance differences, when they used LCSs to manually control the lights, the illuminance differences that they found acceptable were much greater than their detection thresholds. In the acceptability experiment, matching errors were recorded to determine the acceptability of illuminance differences and evaluate the effectiveness of each LCS resolution. Mean matching errors and standard errors of the mean for various resolution conditions were computed and they are plotted as the blue circles and error bars in Fig. 4. Matching errors were consistently greater than 17%, regardless the control resolution. All error bars in Fig. 4 are overlapped, suggesting that resolution has no significant effect on the effectiveness of the LCSs. Thus, matching errors under all resolution conditions across all participants (1440 total trials) were analyzed to determine the acceptable illuminance difference threshold. The mean matching error was 18.5%, with the 95% confidence interval spanning 17.8% to 19.1%.
Fig. 4 Matching errors and users’ confidence in matching accuracy as a function of LCS resolution. Mean matching errors and standard error of the mean (SEM) are shown as the blue circles and error bars (left y-axis). Users’ median subjective ratings of their confidence with their matching accuracy are shown by red triangles (right y-axis).
Although users felt more confident with their matching accuracy when the control resolution was very low (23.6%), the results did not show a significant difference in their performance, as shown in Fig. 4.
The normalized numbers of button presses per trial, obtained from Eq. (1), and the mean speed per button press, obtained from Eq. (2), were calculated to assess efficacy with which the LCS enable the users to reach their desired illuminance. The normalized number of button presses as a function of the resolution are shown in Fig. 5. Theoretically, if the resolution has no effect on users’ behavior, these numbers of button presses with various resolution systems should be the same or very similar. It can be seen in Fig. 5 that, when the resolution is extremely low (23.6% or 20.7%), the normalized number of button presses is significantly greater than with other resolutions. Those extra presses suggest that participants did not accept the relatively large illuminance differences induced, and tried to adjust the illuminance by repeatedly pressing the upward and downward buttons. This was observed by the experimenter and also reported by some participants.
Fig. 5 The effect of control resolution on the normalized number of button presses. Error bars show standard error of the mean (SEM).
The mean speed per button press was calculated as the inverse of the average time consumed between two button presses. As shown in Fig. 6, mean speed per button press decreases as the resolution increases. When the LCS resolution is very high (values lower than 14.8%), the normalized speed per button press is significantly slower than with LCSs of lower resolutions. During the experiment, all participants reported that the extremely high resolution (5.9%) condition made them confused and that they spent more time checking whether the LCS was working properly.
Fig. 6 Mean normalized speed per button press and subjective ratings. The left y-axis shows the mean speed and the right y-axis shows the scale of the subjective ratings. The “+” markers show the median ratings of users’ confidence with matching accuracy. The “X” markers show the median ratings of users’ satisfaction with their matching speed. The solid circles show the median ratings of users’ overall satisfaction. Error bars show standard error of the mean (SEM).
Participants’ subjective ratings are also shown in Fig. 6, to compare with the psychophysical data. As shown in Fig. 6, the decreased speed with higher resolution LCSs is consistent with decreased user satisfaction with the speed, as well as overall satisfaction. It may suggest that satisfaction relates to the speed with which users can reach their desired illuminance, rather than the accuracy with which they can achieve their desired illuminance.
To conclude, users are able to detect illuminance differences of 7.4% of the initial illuminance and they accept illuminance differences of 17.9% to 19.1%. When matching errors, normalized numbers of button presses, mean speed per button press and users’ subjective ratings are collectively considered, the results suggest that the most efficient and user-friendly manual lighting control systems are those with resolutions ranging from 14.8% and 17.7%.
4.2 Dimming curves
A series of step-dimming curves were developed with resolutions ranging from 7% (within detectability limits) to 17% (within acceptability limits). The total number of dimming steps varied as the resolutions varied. For example, there were 65 steps to dim from 100% to 1% with a resolution of 7%, but only 26 steps when the resolution was 17%. To compare the different dimming curves on the same scale, the number of dimming steps was normalized to 100 and power consumption was normalized as a percentage of the maximum power. In Fig. 7-9, the gray bars show the difference in normalized power between the two dimming curves at each setting. The actual amount of energy conserved would depend on the frequency of each setting’s use, which would depend on the application, user behavior, etc. The average was obtained by simply assuming that each setting is equally likely to be used. This method probably underestimates the average energy savings for real applications, since higher settings may be used more frequently than lower settings.
Fig. 7 The proposed step-dimming curve with a resolution of 17%, compared to a continuous logarithmic dimming curve. The left y-axis and blue and red symbols show normalized power consumption as a function of dimmer setting. The squares and blue line show a continuous logarithmic dimming curve. The circles and red line show the proposed step-dimming curve. The right y-axis and gray bars show power conservation as a function of dimmer setting.
Fig. 8 The proposed step-dimming curve with a resolution of 7%, compared to a continuous logarithmic dimming curve. The left y-axis and blue and red symbols show normalized power consumption as a function of dimmer setting. The squares and blue line show a continuous logarithmic dimming curve. The circles and red line show the proposed step-dimming curve. The right y-axis and gray bars show power conservation as a function of dimmer setting.
Fig. 9 The proposed step-dimming curve with a resolution of 17%, compared to a continuous Square Law dimming curve. The left y-axis and blue and red symbols show normalized power consumption as a function of dimmer setting. The squares and blue line show a continuous Square Law dimming curve. The circles and red line show the proposed step-dimming curve. The right y-axis and gray bars show power conservation as a function of dimmer setting.
When the proposed step-dimming curve with a resolution of 17% (a conservative value for inducing an acceptable difference in illuminance) is used, energy consumption can be reduced by an average of 3.9%, compared with continuous logarithmic dimming. The energy reduction was as high as 17% for the maximum illuminance setting. As shown in Fig. 7, when the proposed method is used, the user’s original setting is adjusted to a close, energy-conserving level. Although users might notice the illuminance difference, the experimental results indicate that they would still accept it.
When the resolution is 7%, which is lower than users’ illuminance difference detection threshold, users would not be able to detect an illuminance difference produced by the LCS. In this case, the proposed step-dimming method can save an average of 1.5% energy, relative to continuous logarithmic dimming, as shown in Fig. 8. Although the energy savings are fairly small when with this LCS resolution is compared to a continuous logarithmic dimming method, the change in illumination would be undetectable by users.
The proposed dimming curves are designed based on a basic logarithmic dimming curve, which is believed to be sensible, since it can be deduced from the Weber’s law. Thus, the energy conservation compared to logarithmic dimming curve is less than the energy savings when compared to other dimming curves. When this method is compared to a Square Law dimming curve, the amount of conserved power increases substantially. Figure 9 shows power consumption and power conservation when the proposed method is compared to a Square Law dimming curve. Table 2 summarizes comparisons with both dimming curves.
Table 2. Energy Saved when the Proposed Step-Dimming Curve is Used, Relative to Continuous Dimming Curves
5. Discussion
From results of psychophysical experiments, which indicate that illuminance differences of about 7.4% are barely detectable, and differences from 17.8% to 19.1% are acceptable, implementation of the step-dimming curve proposed here should generally have a resolution from 7.4% to 19.1%. The study of the usability of LCSs with various resolutions shows that a resolutions from 14.8% and 17.7% are most suitable for manual LCSs.
With a 17% difference in illuminance, users will be capable of detecting the difference between their desired illuminance and the illuminance that they are able to achieve, but will still find it acceptable. This could be useful for lighting applications that are not primarily driven by the performance of critical visual tasks, such as hotels, restaurants, and some residential spaces. If lighting control systems were to use the proposed step-dimming curve with a resolution between 14.8% and 17.7%, energy consumption would be reduced, LED luminous efficacy would be increased, and the user experience of the LCSs would be improved. When using the proposed curve with a resolution of 7%, users would not be able to detect illuminance differences. However, the high resolution could potentially lead to user dissatisfaction with manual control systems. Until further research investigates user acceptance of the proposed method with very high resolutions, it might be considered primarily for automatic control systems, such as daylight harvesting systems.
In the acceptability experiment, participants performed matching tasks with two side-by-side booths—simultaneous stimuli were compared. However, in real lighting applications, users typically compare the illuminance before and after their adjustment of the control system, and the two illuminances are not simultaneously compared. Furthermore, in many situations, users compare the achieved illuminances with a desired illuminance that exists only in their imagination. Therefore, the data from the simultaneous comparison in this project presents the “worst-case scenario” and the difference detection thresholds and acceptance thresholds obtained are likely smaller than they would be in a real world applications.
Akashi reported that observers’ memory for illuminance fades and sensitivity to illuminance differences decreases more after 100 seconds than after three seconds [1]. Therefore, continuous dimming curves could be used when occupants actively control the lights. Then, after a setting is made, the illuminance could be adjusted to the closest level in the proposed curve after a delay, such as 100 seconds.
Furthermore, it is noteworthy that energy consumption can be reduced with dimming, not only by reducing unnecessary illuminance, but also by driving LEDs at higher efficacies. The efficacy droop measured and used in the calculations reported here were lower in magnitude than many commercialized products. The LED light engines in this project were directly attached to large, high-quality heat sinks, and were located in a well-ventilated space. Commercialized products in enclosed luminaires and mounted in tight spaces will be more significantly impacted by thermal droop [18]. Thus, with the proposed dimming method, higher energy savings would be expected in real-world applications than the laboratory results reported here.
These experiments were conducted under laboratory conditions, which involved two illuminated booths in a windowless dark room. The results may differ in the real lighting applications since the experimental environment differed from typical illuminated environment. This research aims to provide a generalized understanding to guide the design of control systems and initiate rigorous consideration of control resolutions. The proposed dimming curves are derived from experimental results, but still theoretical in nature. Should manufacturers adopt them for use in their products, adjustments will likely need to be made to suit the design and intended function of the lighting systems. Field studies could be conducted in the future to examine the applicability of the findings of this project to real architectural environments.
6. Conclusion
The first experiment found that horizontal illuminance differences of approximately 7.4% were able to be detected by participants in controlled laboratory conditions. Surprisingly, in those same conditions, participants found matching errors of 17.8% to 19.1% to be acceptable in the second experiment. This suggests that users are capable of detecting small illuminance differences, but have a considerably higher illuminance difference acceptance threshold when given control of the lighting system. A theoretical step dimming method was proposed to conserve energy. Field studies could be conducted to validate and refine the proposed method for use in commercial products.
References and links
1. Y. Akashi and J. Neches, “Detectability and acceptability of illuminance reduction for load shedding,” J. Illum. Eng. Soc. 33(1), 3–13 (2004). [CrossRef]
2. Y. Akashi and P. R. Boyce, “A field study of illuminance reduction,” Energy Build. 38(6), 588–599 (2006). [CrossRef]
3. T. Shikakura, H. Morikawa, and Y. Nakamura, “Research on the perception of lighting fluctuation in a luminous offices environment,” J.Illum. Engng. Inst. Jpn. 85(5), 346–351 (2001).
4. Philips, “An introduction to DALI” http://www.lighting.philips.com.au/v2/static/usr/auto-grab/An Introduction to DALI.pdf.
5. M. F. Schubert, S. Chhajed, J. K. Kim, E. F. Schubert, D. D. Koleske, M. H. Crawford, S. R. Lee, A. J. Fischer, G. Thaler, and M. A. Banas, “Effect of dislocation density on efficiency droop in GaInN/GaN light-emitting diodes,” Appl. Phys. Lett. 91(23), 231114 (2007). [CrossRef]
6. C. Weng, “Advanced thermal enhancement and management of LED packages,” Int. Commun. Heat. Mass. 36(3), 245–248 (2009). [CrossRef]
7. H. R. Blackwell, “Contrast thresholds of the human eye,” J. Opt. Soc. Am. 36(11), 624–643 (1946). [CrossRef] [PubMed]
8. H. B. Barlow, “Increment thresholds at low intensities considered as signal/noise discriminations,” J. Physiol. 136(3), 469–488 (1957). [CrossRef] [PubMed]
9. W. H. Ehrenstein and A. Ehrenstein, “Psychophysical methods” in Modern Techniques in Neuroscience Research (Springer Berlin Heidelberg, 1999), 1211–1241.
10. T. Shikakura, H. Morikawa, and Y. Nakamura, “Perception of lighting fluctuation in office lighting environment,” J. Light Vis. Environ. 27(2), 75–82 (2003).
11. ISO, “Ergonomic requirements for office work with visual display terminals (VDTs),” in Part 11: Guidance on usability ISO 9241–11:1998, ISO, (1998)
12. J. Sauro and E. Kindlund, “A method to standardize usability metrics into a single score,” in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (ACM, 2005), 401–409. [CrossRef]
13. N. Jakob and L. Jonathan, “Measuring usability: preference vs. performance,” Commun. ACM 37(4), 66–75 (1994). [CrossRef]
14. T. T. Norton, D. A. Corliss, and J. E. Bailey, The Psychophysical Measurement of Visual Function (Butterworth-Heinemann, 2002).
15. D. Marcus, K. S. Hickcox, E. Bear, J. P. Freyssinier, and N. Narendran, “ASSIST recommends: dimming: a technology-neutral definition” (LRC, 2013).
16. IESNA, The IESNA Lighting Handbook, 10 ed. (2011).
17. Lutron, “Measured light vs. perceived light” http://www.lutron.com/TechnicalDocumentLibrary/Measured_vs_Perceived.pdf
18. Y. Narukawa, J. Narita, T. Sakamoto, T. Yamada, H. Narimatsu, M. Sano, and T. Mukai, “Recent progress of high efficiency white LEDs,” Phys. Status Solidi., A Appl. Mater. Sci. 204(6), 2087–2093 (2007). [CrossRef]
