You are on page 1of 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/308834122

A study on lighting uniformity for LED smart lighting system

Conference Paper · November 2015


DOI: 10.1109/SSLCHINA.2015.7360705

CITATIONS READS
5 1,320

3 authors, including:

Yu Hu Ming Ronnier Luo


Zhejiang University Zhejiang University
8 PUBLICATIONS   25 CITATIONS    833 PUBLICATIONS   10,779 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Accurate method for computing correlated color temperature View project

Validity of Chromatic Adaptation View project

All content following this page was uploaded by Yu Hu on 16 November 2016.

The user has requested enhancement of the downloaded file.


A Study on Lighting Uniformity for LED Smart Lighting System
Yu Hu1, Ming Ronnier Luo1, 2*, Yang Yang1
1
State Key Laboratory of Modern Optical Instrumentation, Zhejiang University, Hangzhou, CHINA
2
School of Design, University of Leeds, Leeds, UK
*m.r.luo@zju.edu.cn

Abstract uniformity metrics applied in this study are all based on


statistical analysis of lighting distribution. The simplest method
A psychophysical experiment was conducted to study
lighting uniformity across a desk top and the wall as well. The is the min-max ratio, which also used in Slater’s work:
𝐸𝑚𝑖𝑛
data were used to evaluate three uniformity metrics. Four 𝑈1 = ,
attributes were scaled including uniformity, comfort, 𝐸𝑚𝑎𝑥
brightness and contrast perceptions using a categorical where Emax and Emin are the minimum and maximum
judgment method. Twenty-one subjects were asked to rate the illuminance of N points over the illuminated target.
attributes under different degree of lighting uniformity. Another one is the minimum-average ratio:
The results showed that using the ratio between minimum 𝐸𝑚𝑖𝑛 𝐸𝑚𝑖𝑛
𝑈2 = = , Eq. 2
and average to evaluate the uniformity performed the best, and ̅
𝐸 𝑁 ∑𝑁 𝑖=1 𝐸𝑖
a ratio of between 0.80 and 0.90 were consistently to provide where 𝐸̅ is the average illuminance of the N point over the
accepted comfort across the desk top. Also, the contrast of illuminated target.
lighting is significantly negatively correlated with uniformity. The last uniformity approach avoids errors due to local
In the same illuminance level, more uniform lights will make variations. It uses the standard deviation of illuminance σ𝐸 :
subjects feel more comfortable. And in the same uniformity
𝐸̅
level, brighter lights will also make subjects feel more 𝑈3 = , Eq. 3
comfortable. Different desk tops and different illuminance 𝐸̅ + σ𝐸
levels will not significant influence the uniformity perception. These three metrics are popularly applied to assess
Keyword: Smart lighting, lighting Uniformity, LED uniformity of simple illuminance distribution.
This paper used a Smart Lighting System as the test-bed and
1. Introduction a psychophysical experiment was conducted to assess different
With the advancing of light-emitting diode (LED) lightings, uniformities on a desk top and on a wall. The results were
people's living conditions can be largely improved. Light is no analyzed to evaluate the performance of the uniformity metrics
longer simply an illumination system, but a smart system and found that the acceptable uniformity used as a constraint in
capable of creating a large amount of functions. Study on the the Smart Lighting System.
lighting uniformity is one of the important issues of smart
lighting, especially for the daylight-linked lighting controls. 2. Experimental
LEDs not only provide energy savings because of high Setting
efficiency, but also offer what was inconceivable with
traditional sources: controllability of their spectral power
distribution (SPD), spatial distribution, colour properties and
temporal modulation [1]. Due to the complex controllability of
LEDs, a smart control method is necessary and convenient for
users. Miki et al. [2] defined the “Smart Lighting System” as a
system where multiple lighting fixtures and sensors are
connected and they cooperate, to form a network. For energy
saving, Daylight-linked control is one of the most important
control strategies [4]. As we all known, the intensity and the
Correlated Colour Temperature (CCT) of daylight are changing Figure 1 View of experiment office
all the time, so lighting uniformity becomes a major
consideration of light quality while using daylight-linked The experiment was conducted in an office-like room. This
technology. is a windowless room with a 2.7m (length)× 1.2m (width) ×
Slater and Boyce investigated the acceptable illuminance 0.76m (height) desk and a chair, furnished to simulate a single-
uniformity across a desk top using fluorescent lamps [5] [6]. person office room. Figure 1 shows the experimental
They found that illuminance uniformity (minimum/maximum) environment. The desk was covered by two tablecloths, neutral
across a desk of at least 0.7 is likely to be satisfactory for all black and neutral white respectively.
tasks. Moreover, Moreno reviewed several uniformity metrics
based on statistical analysis of light distribution and proposed
a metric based on human vision for non-uniform light
distribution [7].
The aim of this work was to investigate the performance of
several uniformity metrics and to find out acceptable
illuminance uniformity using LED light sources. The three

1
Eq. 1
Table 1 gives the uniformity data on the desk measured by 2 ×
3 spectral irradiance colorimeters.

a) b)

Figure 3 An example for luminance distribution on the wall


and uniformity measurement

And the same lighting conditions were also used to evaluate


uniformity on the wall in front of the desk. The luminance
distributions were captured by the ProMetric which was 2.5m
Figure 2 Illuminance distribution across the center of desk
away from the wall, at the same position of observers. Figure
3(a) shows the light distribution on the wall. Figure 3(b) shows
The room was lit by 10 LED panels, which uniformly the 3 ×3 point-sampling method, applied to get luminance of 9
arranged on the ceiling (see Figure 1). Each panel had 11 points over the wall. Table 2 shows uniformity results on the
different types of LED, giving a good coverage of the visible wall.
spectrum, so any desired SPDs could be produced. They were
remotely controlled by Matlab using Zigbee technology. Table 2 Uniformity of 14 lighting conditions on the wall
Because they worked individually, the illuminance distribution
Average
across the desk could give different levels of uniformity.
luminance U1 U2 U3
Moreover, Maltab controlled 2 × 3 spectral irradiance
(cd/m2)
colorimeters (SPIC200B) via WiFi, served as horizontal
uniformity measurement devices. The vertical uniformity was 1 53.7 0.526 0.652 0.838
measured by an imaging colorimeter and photometers 2 37.6 0.376 0.579 0.741
(ProMetric).
3 60.7 0.692 0.790 0.895
Table 1 Uniformity of 14 lighting conditions on the desk 4 49.5 0.473 0.645 0.796
Average 5 67.1 0.724 0.821 0.901
U1 U2 U3
illuminance(lx) 6 58.4 0.543 0.702 0.829
1 487.78 0.328 0.503 0.686 7 75.6 0.717 0.832 0.888
2 482.07 0.314 0.488 0.681
8 26.3 0.560 0.679 0.850
3 623.55 0.577 0.724 0.811
9 19.4 0.390 0.590 0.747
4 619.80 0.555 0.705 0.801
5 694.82 0.755 0.857 0.893 10 30.2 0.662 0.776 0.890
6 694.87 0.734 0.842 0.884 11 24.8 0.457 0.629 0.788
7 795.98 0.979 0.991 0.992 12 33.2 0.722 0.842 0.895
8 248.32 0.328 0.498 0.691 13 30.2 0.570 0.725 0.836
9 243.02 0.311 0.487 0.681 14 39.6 0.721 0.827 0.890
10 310.90 0.578 0.726 0.814
11 310.55 0.551 0.708 0.801 Psychophysical experiment
12 354.45 0.757 0.850 0.897 An Experiment was carried out to investigate the light
13 346.97 0.722 0.839 0.881 uniformity on both the desk top and the wall. For each light
setting, the subject was asked to scale all the attributes. Each
14 399.05 0.970 0.990 0.988
attribute was consisted of six-categorical point scale. The main
rating scale quantified the participant’s opinions about the
Light distributions in the experiment had 4 uniformity levels uniformity on the desk or on the wall (1 = very non-uniform, 2
at 2 illuminance levels with the maximum of 800 lx and 400lx, = non-uniform, 3 = small non-uniform but not acceptable, 4 =
from 2 directions (the brightest part is the left or the right). In small non-uniform but acceptable, 5 = just notice non-uniform,
total, there were 14 lighting conditions as illustrated in Figure 6 = very uniform). The other three categorical scales assessed
2. In addition, there were two desk top conditions, i.e. covered the comfort of the lighting on the desk or on the wall (1 =
by a white or a black tablecloth. The CCT of all lighting uncomfortable, 6 = comfortable); the brightness of the lighting
conditions was fixed at 5000K. Four lighting conditions were (1 = dark, 6 = bright); and the contrast of the desk or the wall
repeatedly assessed for studying the intra-observer variation. (1 = small contrast, 6 = large contrast).

2
Observers Table 4 Pearson’s r values between the measured data and
Twenty-one subjects (12 females and 9 males) with normal the uniformity rating data
colour vision were asked to assess the various attributes of Uniformity rating U1 U2 U3
lighting. The Ishihara vision test was used to test colour
blindness. Their ages ranged from 21 to 34, with an average of desk 0.988 0.992 0.990
23.3. As a result, 4536 assessments were made, i.e. 3 targets × wall 0.898 0.940 0.831
(14 + 4) lighting conditions × 21 observers × 4 questions.
Figure 5 shows the relation between ratio of minimum and
3. Results and Discussion average values and uniformity ratings. It can be seen that the
Observer variation excellent agreement between the U2 predictions and visual
Coefficient of variation (CV) was applied to investigate results. Also, the slopes of the two best fitted lines are different
inter-observer variation. It shows the extent of variability in between the desktop and the wall, i.e. the former uniformity is
relation to the mean of the population. For the best reliability, more discriminable on uniformity than the latter.
CV value should be zero.
𝐶𝑉 = 𝜎/𝑋̅ ×100% Eq. 4
where 𝜎 is the standard deviation and 𝑋̅ is the average value. a)
The inter-observer variation, which is also known as
observer accuracy, was calculated by all observer’s results in
the same condition. Table 3 shows inter-observer variation for
14 lighting conditions.

Table 3 Inter-observer variation in term of CV values


target min max mean
white tablecloth 19.6 34.1 26.6
black tablecloth 19.3 36.9 28.1
b)
wall 18.9 39.0 27.5

And Pearson product-moment correlation coefficient


(Pearson’s r) was also applied to evaluate the intra-observer
variation. A higher r value shows a better agreement. For a
perfect one, r should be one. The intra-observer variation,
which is also known as observer repeatability, was calculated
between each observer’s two-time evaluations for four repeated
lighting conditions. Figure 4 shows intra-observer variation
Figure 5 Plots of the U2 predictions against uniformity
results.
ratings on the a) desk and b) the wall.

Figure 6(a) shows that the perceived contrast had a clear


negative correlation with the perceived uniformity, i.e. a higher
contrast leading to a lower uniformity. Figure 6(b) shows that
the perceived comfort and uniformity had a positive
correlation, i.e. more uniform light distribution will be more
comfortable. It also shows that a higher luminance level will be
more comfortable.

Figure 4 intra-observer variation results in term of Pearson’s r


values

Uniformity Evaluation
Three uniformity metrics were tested by comparing the
measured data with the uniformity rating data. Pearson’ r
values was applied to indicate the agreement between those two
datasets. Table 4 shows the Pearson’s r value results. From the
table, one can find that all the 3 uniformity metrics showed a
high correlation with perceived uniformity. The ratio of
minimum-average value (U2) agreed best with the visual Figure 6 Plots of the visual uniformity ratings against a)
results, giving the best performance. contrast and b) uniformity ratings.

3
Finally, it can be found that a category of just comfortable (4)
corresponds to 4 and 5 of the perceived uniformity values for
the high and low illuminance levels, respectively. According to
Figure 5(a), they correspond to 0.80 and 0.90 U2 uniformity
values predicted by the U2 formula, which are the acceptable
thresholds.
4. Conclusions
A psychophysical experiment was carried out to
accumulate the visual data on non-uniform light distribution
and to evaluate the performance of three uniformity metrics. It
is learned that the contrast of lighting was significantly
correlated with uniformity in a strong negative relationship.
Also, at the same illuminance level, more uniform lights will
make subjects feel more comfortable. And for the same
uniformity level, brighter lights will also make subjects feel
more comfortable. The results showed that all the 3 uniformity
metrics had a high correlation with the perceived uniformity.
The ratio of minimum-average value (U2) agreed best with the
visual results, and gave the best performance on evaluating
uniformity on the desk and on the wall. It is also found that
min-average uniformity ratios between 0.80-0.90 correspond to
just acceptable comfort.

Acknowledgement
The authors would like to thank Thousand Light
(Changzhou) Ltd. to supply the LED lighting system for the
experiment.

References
1. S.E. Fred, and J.K. Kim. "Solid-state light sources getting
smart." Science 308.5726 (2005): 1274-1278.
2. M. Mitsunori, T. Hiroyasu, and K. Imazato. "Proposal for
an intelligent lighting system, and verification of control
method effectiveness."Cybernetics and Intelligent Systems,
2004 IEEE Conference on. Vol. 1. IEEE, 2004.
3. M. Luigi. "A smart lighting control to save
energy." Intelligent Data Acquisition and Advanced
Computing Systems (IDAACS), 2011 IEEE 6th
International Conference on. Vol. 1. IEEE, 2011.
4. M.A. UI Haq, M.Y. Hassan, H. Abdullah, H.A. Rahman,
M.P. Abdullah, F. Hussin, and D.M. Said. "A review on
lighting control technologies in commercial buildings, their
performance and affecting factors." Renewable and
Sustainable Energy Reviews 33 (2014): 268-279.
5. A.I. Slater, M.J. Perry, and D.J. Carter. "Illuminance
differences between desks: Limits of
acceptability." Lighting Research and Technology25.2
(1993): 91-103.
6. A.I. Slater, and P.R. Boyce. "Illuminance uniformity on
desks: Where is the limit?." Lighting Research and
Technology 22.4 (1990): 165-174.
7. M. Ivan. "Illumination uniformity assessment based on
human vision." Optics letters 35.23 (2010): 4030-4032.
Precede

4
View publication stats

You might also like