Measurement 46 (2013) 37833795

Contents lists available at SciVerse ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Power quality performance of energy-efcient low-wattage LED

lamps
Sohel Uddin , Hussain Shareef, Azah Mohamed
Department of Electrical, Electronic and Systems Engineering, Universiti Kebangsaan Malaysia, Bangi, 43600 Selangor, Malaysia

a r t i c l e i n f o a b s t r a c t

Article history: Light-emitting Diode (LED) lamps are more versatile and energy efcient compared with
Received 4 January 2013 conventional light sources. This paper investigates two main problems related to power
Received in revised form 13 February 2013 quality, namely, the harmonic generation from LED lamps and the effect of voltage sags
Accepted 12 July 2013
on LED lamps. Laboratory tests on various LED lamps are conducted and the electrical char-
Available online 22 July 2013
acteristics of LED lamps are tapped under different conditions. Frequency domain analysis
is then performed to investigate the generated harmonics. Then, voltage sag sensitivity
Keywords:
analysis is performed by obtaining signals from a photo sensor, the different parts of a lamp
LED lamps
THD
ballast, and input voltage. The analysis includes the effects of varying sag depth and voltage
Voltage sag sag duration. The ndings are compared with the SEMI F47 standard curve, which predicts
Voltage tolerance curve the effects of voltage sags on lamp performance. Experimental results show that the LED
Power quality lamps produce a considerable amount of current harmonics. Moreover, different types of
LED lamps were found to produce different levels of current harmonic depending on the
ballast conguration. This harmonic distortion can be reduced by combining various types
of LED lamps. Voltage sag tests show that all the tested lamps are sensitive to sag depth and
sag duration.
2013 Elsevier Ltd. All rights reserved.

1. Introduction linearity to a system. As nonlinear loads, LED lamps pro-

Lighting accounts for approximately 20% of the electric-
ity consumption worldwide. To promote energy conserva-
tion, many governments have introduced directives to ban
energy-inefcient incandescent light bulbs and replace
them with other technologies, such as LEDs [1,2]. LEDs
work under a completely different principle from that of
compact uorescent lamps (CFLs). In LED lighting technol-
ogy, when an electron is recombines with a hole, release
energy in the form of light known as photon. Many high
brightness LEDs are required to illuminate a large area be-
cause all the light is produced at the pn junction [3].
In general lighting applications, a compact AC/DC con-
verter should be used in one lighting xture to supply DC
current to high brightness LED chips, which introduce non-
Corresponding author.
E-mail addresses: sohel_091@yahoo.com (S. Uddin), shareef@eng.ukm.
my (H. Shareef), azah@eng.ukm.my (A. Mohamed).
duce highly distorted currents [3]. Although the required
active power of a LED lamp is low, a huge number of con-
sumer using LED lamps and CFLs in one location may give
rise to signicant power quality (PQ) problems [4]. Voltage
sag is an important issue among various PQ disturbances.
During sag, a sudden 1090% reduction in nominal voltage
occurs; this reduction lasts from 10 ms to 60 s [5]. LED
lamps may be extinguished or generate ickering, causing
out of order due to sag. Therefore, a good understanding of
the LED bulbs harmonic-producing characteristics and
sensitivity to voltage sags is necessary.
Numerous studies were conducted on LEDs as an
energy-efcient lamp, but most researchers focused on
internal ballast circuit design and on enhancing LED
performance [68]. Several researchers concentrated about
distribution of light [9,10]. Only few studies focused on the
harmonic emissions of LED lamps [3,11]. As indicated in
[3], the widespread use of LEDs and CFLs could increase

0263-2241/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.measurement.2013.07.022
3784 S. Uddin et al. / Measurement 46 (2013) 37833795
the voltage distortion in distribution networks depending
on the characteristics of the network. However, the arith-
metic sum of individual harmonics may lead incorrect esti-
mations of current harmonic distortion levels. Therefore,
measuring diversity factors to determine the effect of a
large number of LED lamps on PQ is necessary. The ratio
of the integral vector of current harmonics to the arithme-
tic sum of current harmonics is called diversity factor [12].
A number of past studies on voltage sag sensitivity of
other nonlinear lighting systems have also been carried
out, but no research on the voltage sag performance of
LED lamps can be found [1315]. Taekhyun et al. [16]
conclude that LED lamps are responsible for ickering. This
behavior is similar to the response of CFLs to high-order
inter-harmonics. Therefore, LED lamps with internal
electronic ballast are expected to not only be sensitive to
ickers, but also to transient under voltage events in elec-
tric distribution systems.
This paper presents a detailed harmonic and voltage sag
analysis of LED lamps. Measurement tests using various
available LED lamps are conducted. A total of 35 high
brightness LED lamps with different ratings from seven dif-
ferent manufacturers (brands) are observed. The experi-
ment is executed to monitor the current, voltage and
light-output waveforms of the LED bulbs and to analyze
the lamps in terms of power rating and brand. The test
results are compared with the IEC 61000-3-2 harmonic
standard and SEMI F47 voltage sag standard.
2. Basic operation of LED lamps and relevant standards
The principle behind the operation of LED bulbs is
described in this section. Relevant standards, including
harmonic emission limits for LEDs as dened by IEC
61000-3-2 and voltage acceptability limits dened by SEMI
F 47, are discussed.
2.1. Operating principle of LED lamps
LEDs require a constant current source from a low-DC
voltage source obtained from the AC mains. Use of a con-
verter for voltage and current regulation which is usable
to LED cheap is therefore necessary. The boost, yback, val-
lyl, resonant, buck converters are renowned power
sources for LEDs [7,17]. Fig. 1 shows a classical block dia-
gram of the LED lamp ballast circuit, which includes an
AC line input voltage (typically 220240 VAC, 50/60 Hz),
an electronic interference lter for obstructing unwanted
circuit-produced switching signals, a rectier with a
smoothing capacitor, a pulse width modulation-controlled
AC Mains EMI Filter Bridge Rectifier
Fig. 1. Basic block diagram of LED lamp internal ballast circuit.
constant current source converter for DCDC conversion,
and an array of high brightness LEDs. The directives that
govern the injection of harmonics are not particularly
stringent because rated load powers are low [14]. Thus,
power factor control circuits may or may not be found in
low-wattage LED lamp ballasts. However, introducing an
input passive lter, valley-ll circuits, or an integrated
circuit-controlled active ltering conguration reduces
generated harmonics and improves the power factor.
Under normal operation, the capacitor voltage decays
over a half-cycle as [14]
I0 T 50
DV 1
2C dc
where DV is the decay of voltage across the capacitor, I0 is
the load current, T50 is a duration of 50 cycles, and Cdc is the
DC link capacitance.
The charge of a capacitor is approximately equal to the
supply peak voltage plus DV/2, which is standard for a
constant DCDC converter. During the N cycle voltage
sag, the smoothing capacitor discharges until 2NDV/2.
Depending on the value of minimum voltage prepared by
the ballast conguration, the constant current source con-
verter delivers adequate amounts of voltage and current to
the lamp up to the capacitor voltage to obtain the intended
minimum converter operating value. The time spent to
reach the voltage is called hold-up time (Th), which can
be expressed as follows [14]:
C dc V 2norm  V 2min
Th 2
2P
where Vnorm is the peak nominal voltage, Vmin is the peak
minimum voltage prepared by the ballast conguration,
and P is active power.
2.2. Harmonic injection limits for LED lamps
Like any other appliance, LED lamps must also obey
with various instructions feasible to the product. Limits
for harmonics are classied into the different categories
according to load type. All lighting loads are under class
C [18,19]. Harmonic emission limits for class C are divided
into two subcategories based on active power up to 25 W
and above. The lamps with an active input power less than
or equal to 25 W should satisfy at least one of the two
criteria. In the rst criterion, the third harmonic current
should not exceed 86% of the fundamental frequency,
and the fth harmonic current should not exceed 61%.
This criterion indicates that the current total harmonic
distortion (THDI) should be approximately 105%. The other
Smooth DC Converter LED Array
Constant ...
Current Source
 S. Uddin et al. / Measurement 46 (2013) 37833795 3785

Table 1 Table 2
IEC 61000-3-2 standard with class C category (P 6 25 W). Technical data for the test lamp samples.

Harmonics (n) Class C (% of fundamental) Trade Nominal Equivalent to Power Luminous Life
name power P incandescent factor ux (lm) span
3 30  PF
(W) P (W) (years)
5 10
7 7 Philips 4 25 0.66 250 25
9 5 Philips 5 40 0.67 350 25
11 6 n 6 39 3 Philips 7 40 0.7 350 25
Osram 4 25 0.51 250 25
Osram 6 30 0.52 365 25
criterion for each harmonic order is given in Table 1, where Osram 8 40 0.82 450 25
only odd-order harmonics are shown because the even-or- Osram 10 50 0.89 950 25
der harmonics are almost zero because of sine-wave Toshiba 5.5 30 0.59 290 20
symmetry. Bright 5 40 0.48 230 6
Cash 7 40 0.51 350 20
Evenzo 3 15 0.5 150 18
2.3. Voltage acceptability limits for LED lamps
3.1. Harmonic measurement procedure
Equipment sensitivity to voltage sag can be considered
and presented in the form of voltage acceptability curves, As listed in Table 2, twelve samples of high brightness
which are also known as voltage tolerance or sensitivity LED lamps with different power ratings from various man-
curves. The design of the voltage acceptability curves ufacturers are tested to analyze the characteristics of LED
clearly relates to whether the distributed power can be uti- lamps. All the selected samples are suitable for 220
lized or not. Thus, the ultimate criterion of power accept- 240 V mains. The lamps have a power consumption rating
ability relates to the operating status of the industrial of 310 W.
process and equipment. This criterion depends on the nat- An experimental setup (Fig. 3a) is assembled to achieve
ure of the load such as the conventional incandescent lamp exact data of current harmonic component of LED lamps.
follows a loose criterion while particular sensitive uores- The setup comprises four components: a Fluke 434 power
cent lamp with electronic ballast has hard restrictive crite- quality analyzer, Fluke i30s current clamp, LED lamp(s) un-
ria. The most recent voltage acceptability standard (Fig. 2) der test, and a personal computer for analyzing the signals.
is called SEMI F 47 which is developed by Semiconductor For stabilization, each lamp is kept on for 10 min before
Equipment and Materials International in the year of measurements are taken. Each lamp is also tested four
2000 [20]. The specication states that semiconductor pro- times during different periods of the day to eliminate any
cessing, metrology, and automated test equipment should error. The captured current waveforms are analyzed using
be designed and built in conformance with the voltage sag Fluke 434 and MATLAB software.
ride-through capability requirement as per the dened
curve. Therefore, the performance of LED lamps may be
evaluated against the SEMI F 47 standard. 3.2. Voltage sag measurement procedure

Eight different LED lamp samples from Table 2 are

3. Experimentation
This section illustrates the design of the experimental
setup and the testing procedures for determining the har-
monic and voltage sag performance of LED lamps.
tested to analyze the characteristics of LED lamps during
voltage sag. The testing method is follows the general out-
line of IEC Standard 61000-4-11.
The experimental setup contains four constituent: an
industrial power corruptor (IPC) for sag generator with

100
90
80
Nominal Voltage (%)

70
60
50
40
30
20
10
0
0 0.05 0.1 0.2 0.5 1.0
Duration of Voltage Sag (Second)

Fig. 2. SEMI F 47 standard curve.

3786 S. Uddin et al. / Measurement 46 (2013) 37833795

(a)

(b) Illuminance Photo Meter

Measurement

measurement
IPC for Sag Generation

channels
IPC low
voltage
Light Chamber

LED Lamp
without
.
ballast
LED chip
voltage Power Supply
Measurement to IPC
Rectifier
Converter

Constant
DC link Current
voltage Source

Measurement
DC Link EMI Filter
Ballast

Fig. 3. Experimental setup: (a) harmonic measurements and (b) voltage sag measurements.

an internal data acquisition system, illuminance meter
(Konica Minolta T-10), lamps and a laptop computer for
analyzing the signals. IPC has. The IPC generates and inter-
rupts voltages up to 480 V and 50 A currents. Light output
is measured by the photo detector of Konica Minolta T-10
inside lighting chamber by adjusting the distance between
the photo detector and the lamp to specied lux value of
the lamp in the data sheet. The experiment is conducted
to determine the variations in light illumination of LED
lamps throughout voltage sags. Given this goal, the light
meter in the experimental system should able accurately
entrap the variations in light illumination of corresponding
lamps.
The experimental setup shown in Fig. 3b is built to
determine voltage sag interference and measure the light
illuminance from the tested lamps. This point source tech-
nique evaluates the light generated from lamp by using a
photo sensor. The photo-current from sensor is transfer
into voltage. The detector can detect ickers in the human
visible range of 035 Hz [15]. The measurement of DC link
voltage and LED array voltage are performed by taking
signal from smoothing DC capacitor terminal and of the
outpour of the converter circuit of ballast circuit respec-
tively as shown in Fig. 3b. Moreover there is no data
acquisition system inside photometer so the converted
voltage is sent to the Industrial power corruptor for post
processing and analysis.
4. Results and discussion
This section provides, the measurements of the
harmonics and the sag test conducted on the LED lamps
are analyzed and discussed.
4.1. Harmonic analysis results
Lamp current waveforms were analyzed using the
Fourier theorem. This method provides the frequency
spectrum of the lamp currents. The spectrum is
 S. Uddin et al. / Measurement 46 (2013) 37833795 3787

Harmonic (%)
Current (A) 0.3 100
0.2 No Filter 80 THD=170-175 %

Current
0.1 60
0
-0.1 40
-0.2 20
0
0 5 10 15 20 1 3 5 7 9 11 13 15 17 19
Time (ms) Harmonic Order (n)

Harmonic (%)
0.2 100
Current (A)

Passive Filter 80 THD=105-110 %

0.1

Current
60
0
40
-0.1 20
-0.2 0
0 5 10 15 20 1 3 5 7 9 11 13 15 17 19
Time (ms) Harmonic Order (n)

Harmonic (%)
0.2 100
Current (A)

Valley Fill 80 THD=63-70 %

0.1

Current
60
0
40
-0.1 20
-0.2 0
0 5 10 15 20 1 3 5 7 9 11 13 15 17 19
Time (ms) Harmonic Order (n)

Harmonic (%)
0.1 100
Current (A)

Active Filter 80 THD=30-35 %

0 Current 60
40
20
-0.1 0
0 5 10 15 20 1 3 5 7 9 11 13 15 17 19
Time (ms) Harmonic Order (n)

Fig. 4. Different types of current waveforms obtained from several test LED lamps.

symbolized by the fundamental sinusoidal component and
a series of higher order harmonic components at frequen-
cies that are integer multiples of the fundamental
frequency. The square roots of the sum of the amplitudes
of the harmonics represent the THD.
4.1.1. Harmonic characteristics of individual lamps
Harmonic from single lamp tests are performed to
explore the type of harmonic lters used in the LED lamp
ballast circuit and harmonic emissions from all tested
lamps.
Fig. 4 shows the typical current waveforms obtained
from the different tested lamps. As can be seen in Fig. 4,
the current waveforms are not sinusoidal, indicating that
they inject harmonics into the power system. Furthermore,
this result indicates that the ballast in dissimilar LED lamps
uses different ltering methods to reduce harmonic
generation. For example, most of the tested Philips LED
lamps utilize valley-lled circuit, the Toshiba 5.5 W lamp
contains a passive lter, Osram 8 W embeds an active
lter, and some other tested lamps do not implement
any ltering technique.
The relative harmonic currents of all the tested lamps
are presented in Table 3. Some of the tested LED lamps
generate high levels of harmonics that are unacceptable
to the IEC 61000-3-2 standard described above. As can be
seen in Table 3, the Philips lamps exhibited THDI values
between 67% and 72%, whereas the Osram LED lamps with
power rating less than 6 W produced high levels of har-
monics ranging from 173% to 175% because they do not
use any lter in their ballast circuit. On the other hand,
the 8 and 10 W Osram LED lamps produced the lowest
THDI ranging from 30% to 35%. All other tested lamps
exhibited THDI values greater than 100% (Table 3).

Table 3
Harmonic contents of single LED lamps.

Tested lamp Type of lter Harmonic (%)

Fund 3rd 5th 7th 9th THDI
Philips 4 W Valley ll 100 34.61 6.28 22.27 18.94 67.05
Philips 5 W Valley ll 100 36.92 7.16 19.81 19.39 69.83
Philips 7 W Valley ll 100 32.34 11.6 23.4 19.03 71.23
Osram 4 W No lter 100 89.24 70.65 51.19 38.57 173.9
Osram 6 W No lter 100 91.96 77.05 58.51 41.01 174.3
Osram 8 W Active lter 100 22.25 15.04 2.34 9.73 30.94
Osram 10 W Active lter 100 32.12 4.68 2.67 6.45 34.78
Evenzo 3 W No lter 100 90.7 77.34 58.89 45.32 164.4
Bright 5 W No lter 100 86.65 75.92 61.75 48.87 167.2
Cash 7 W No lter 100 91.23 74.81 56.41 41.06 168.2
Toshiba 5.5 W Passive lter 100 73.4 45.4 35.62 31.38 106.3
3788 S. Uddin et al. / Measurement 46 (2013) 37833795

Table 4 lamps). Therefore, the diversity factor decreases as the

Diversity factors of several LED lamps.
Combination Lamps
1 Philips 4 W, Osram 4 W, Bright 5 W
2 Philips 7 W, Osram 4.5 W, Osram 6 W,
Bright 5 W, Evenzo 3 W
3 Philips 7 W, Osram 4.5 W, Bright 5 W,
Cash 7 W, Evenzo 3 W
4 Philips 7 W, Osram 4.5 W, Osram 6 W,
Bright 5 W, Toshiba 5.5 W, Cash 7 W,
Evenzo 3 W
4.1.2. Harmonic characteristics of multiple lamps
These tests are intended to analyze the effects of the
harmonic characteristics of various groups of LED lamps
by calculating their diversity factors. A small diversity
factor indicates the occurrence of a considerable cancella-
tion because of the superposition of individual current
wave shapes. The diversity factors of different combina-
tions are presented in Table 4. As can be seen in Table 4,
Combination-1 had a diversity factor of 0.83, whereas
Combination-4 exhibited a much lower diversity factor of
0.72, indicating that Combination-1 (with 3 lamps)
produced more harmonics than Combination-4 (with 7
number of lamps increase.
Diversity Figs. 5 and 6 show the diversity factors of individual
factors harmonic orders in Combination-1 and Combination-4,
0.83 respectively. Only odd-order harmonics are presented
0.77 because the even-order harmonics are almost zero.
The diversity factors of all the harmonic orders for Combi-
0.78
nation-1 exceeded 0.5 (Fig. 5), whereas those of the
0.72 individual harmonic orders for Combination-4 were signif-
icantly lower and fell below 0.5 starting from the 19th
order harmonic. Therefore, Combination-1 is more prob-
lematic than Combination-4.
Figs. 7 and 8 illustrate the phase angles of different har-
monic current components for Combination-1 and Combi-
nation-4 respectively. The magnitude of absolute current is
plotted with respective angle for some harmonics such as
3rd, 5th, 7th and 9th. From Fig. 7, it can be seen that ith
harmonic component of individual lamps of the Combina-
tion-1 fall in the same quadrant of the polar plot. This
graphically represents that there is very little harmonic
cancellation. Meanwhile the 3rd, 5th, 7th and 9th
harmonic components in lamps of Combination-4 is found
to scatter in all four as seen in Fig. 8. Therefore, harmonic
cancellation is obvious in Combination-4.

1.0

0.8
Mean Diversity Factor

0.6

0.4

0.2

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Harmonic Order (n)

Fig. 5. Diversity factors of different harmonic orders for Combination-1.

1.0

0.8
Mean Diversity Factor

0.6

0.4

0.2

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Harmonic Order (n)

Fig. 6. Diversity factors of different harmonic orders for Combination-4.

 S. Uddin et al. / Measurement 46 (2013) 37833795 3789

3rd Harmonic 5th Harmonic

90 90
120 60 120 60

150 30 150 30

180 0 180 0

210 330 210 330

240 300 240 300

270 270 Combined
Philips 4 W
7th Harmonic 9th Harmonic Osram 4 W
90 90 Bright 5 W
120 60 120 60

150 30 150 30

180 0 180 0

210 330 210 330

240 300 240 300

270 270

Fig. 7. Phase angle of all lamps for Combination-1.

3rd Harmonic 5th Harmonic

90 90
120 60 120 60

150 30 150 30

180 0 180 0

Combined
210 330 210 330
Philips 7 W
240 300 240 300 Osram 4.5 W
270 270 Osram 6 W
7th Harmonic 9th Harmonic Bright 5 W
90 90 Toshiba 5.5 W
120 60 120 60 Cash 7 W
Evenzo 3 W
150 30 150 30

180 0 180 0

210 330 210 330

240 300 240 300

270 270

Fig. 8. Phase angle of all lamps for Combination-4.

The effects of the LED lamps of the same brand and
wattage are connected in groups, as observed via another
test. The results of the groups of Philips 5 W and Osram
4 W LED lamps are presented in Tables 5 and 6, respec-
tively. The addition of similar LED lamps in the power
network did not vary the harmonic level because the wave
shapes generated by the added similar lamps were unaf-
fected, except for the magnitude of the current drawn from
the system. A similar observation was made when lamps
with different power wattages, the same manufacturer,
3790 S. Uddin et al. / Measurement 46 (2013) 37833795

Table 5 tions. The turn-off condition occurred after three cycles

Harmonic contents of a group of Philips 5 W LED lamps.
No. of lamps Power P (W) Harmonic (%)
Fund
1 5 100
2 10 100
3 15 100
Table 6
Harmonic contents of a group of Osram 4 W LED lamps.
No. of lamps Power P (W) Harmonic (%)
Fund
1 4 100
2 8 100
3 12 100
and the same type of ballast conguration were consid-
ered. In these cases, the diversity factor was unaffected
and the diversity factor was close to 1.
4.2. Voltage sag analysis results
This section analyzes some of the experimental results.
All analyses are performed by obtaining signals from the
photo sensor, different parts of the lamp ballast, and the
supply voltage. A generalized voltage tolerance curve for
LED lamps was created based on the ndings and it was
compared with the SEMI F 47 and the Information Technol-
ogy Industry Council (ITIC) curve.
The signals obtained from the photo sensor and source
voltage were analyzed to determine the sensitivity of the
LED lamps to voltage sags. Fig. 9 shows the signal obtained
from the photo sensor and source voltage for the Cash 7 W
lamp, that is, the impact of changing the sag depths be-
tween 32.5% and 27.5% of remaining voltage for four cycles
of the light illuminance variations of the lamp. The lamp
switched off at the 30% remaining voltage on four cycles.
The disturbance only started at 32.5% remaining voltage,
but the lamp did not experience zero-luminance condi-
of sag initiation and returned to normal operation seven
cycles after the sag recovery point.
3rd 5th 7th THDI Fig. 10 shows the effects of increasing sag duration with
constant sag depth. Fig. 10a shows several waveforms of
35.3 11 18.5 64.1
35.7 11.2 17 64.5
light output variation and DC LED array voltage across
35 12.1 18.7 62.3 the Cash LED lamp. This phenomenon occurs with the
sag duration at 32.5% constant sag depth and varies be-
tween three and ten cycles just before the turn off condi-
tion. The minimum light illuminance fell to 90% and
reverted to normal brightness when the supply voltage
stabilized. The DC voltage across the LED arrays remained
constant at 22.5 V. In this condition, the lamp is considered
3rd 5th 7th THDI
to function normally. A different phenomenon was ob-
91.4 74.2 54.9 174.1 served in the impact of varying the duration at 30% sag
92.1 77.8 59.9 174.6
89.1 70.8 51 173.2
depth displayed in Fig. 10b. The lamp always completely
turned off from the fourth cycle but lighted up with full
brightness for the rst three cycles. The voltage drop across
the LED arrays fell sharply during malfunctions, indicating
the absence of sufcient voltage across LED junctions to re-
vert instantaneously to full brightness.
The second tested lamp was a Philips 7 W LED lamp.
Similar to the Cash lamp, the Philips lamp exhibited
turn-off conditions caused by voltage sag distraction.
Fig. 11 shows the variations in light illuminance level of
the Philips lamp under voltage sag that lasted for two cy-
cles. As can be seen in Fig. 11, the Philips LED lamp was
much more resistant to sag depth than the Cash LED lamp.
The zero-illuminance disturbance conditions of the Philips
lamp close to 35 ms started at 25% remaining voltage.
Zero-illuminance also occurred at 30% remaining voltage
for three cycles. Therefore, the Philips lamp is more sensi-
tive than the Cash LED lamp in terms of sag duration
(Fig. 12). Similar to the Cash 7 W lamp, analysis of the
LED array voltage indicates visible malfunction conditions.
The voltage level across the LED junction reached 11 V for
the Philips lamp because the lamp contained fewer LEDs
than the Cash lamp.
The 5 W lamps are considered in the next step. The rst
lamp tested was the Bright 5 W lamp. The knee points in
the Bright lamp appeared at 12.5% remaining voltage start-

100 500

80 Variation of Light
32.5% remaining voltage 250
Light Output (%)

30% remaining voltage

Voltage Sag (V)

60 27.5% remaining voltage

0
Supply Voltage
40 32.5% remaining voltage
30% remaining voltage
27.5% remaining voltage -250
20

0 -500
-50 -25 0 25 50 75 100 125 150 175 200 225 250
Time (ms)

Fig. 9. Impact of sag depth on the light variation of Cash 7 W LED lamp after four cycles.
 S. Uddin et al. / Measurement 46 (2013) 37833795 3791

(a) 100 25

80 Variation of Light 20

LED Array Voltage (V)

10 cycle

Light Output (%)

8 cycle
60 4 cycle 15
3 cycle

40 LED Array Voltage 10

10 cycle
8 cycle
20 4 cycle 5
3 cycle

0 0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375
Time (ms)

(b) 100 25

Variation of Light
80 20
10 cycle

LED Array Voltage (V)

8 cycle
4 cycle
Light Output (%)

3 cycle
60 15

LED Array Voltage

40 10
10 cycle
8 cycle
4 cycle
3 cycle
20 5

0 0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375
Time (ms)

Fig. 10. Impact of sag duration on the light variation of Cash 7 W LED lamp: (a) 32.5% remaining voltage and (b) 30% remaining voltage.

100 400

80 Variation of Light
27.5% remaining voltage 200
25% remaining voltage
Light Output (%)

Voltage Sag (V)

60

40
Supply Voltage
27.5% remaining voltage
-200
20 25% remaining voltage

0 -400
-50 -25 0 25 50 75 100 125 150
Time (ms)

Fig. 11. Impact of sag depth on light variation of the Philips 7 W LED lamp for two cycles.

ing at the sixth cycle. Fig. 13 shows the effect of sag dura- approximately 60% but reverted to normal at the instance
tion at 15% and 12.5% remaining voltage for sag durations of sag recovery point at 15% remaining voltage, as shown
between ve and ten cycles. The light output dropped to in Fig. 13a. All these events exhibited similar characteris-
3792 S. Uddin et al. / Measurement 46 (2013) 37833795

100 15

80 12

LED Array Voltage (V)

Variation of Light

Light Output (%)

60 10 cycle 9
5 cycle
3 cycle
40 6
LED Array Voltage
10 cycle
20 5 cycle 3
3 cycle

0 0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275
Time (ms)

Fig. 12. Impact of sag duration on light variation and LED junction voltage of the Philips 7 W LED lamp at 30% remaining voltage.

(a) 100

80
10 cycle
Light Output (%)

8 cycle
60 6 cycle
5 cycle

40

20

0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275
Time (ms)

(b) 100

80
10 cycle
8 cycle
Light Output (%)

6 cycle
60
5 cycle

40

20

0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275
Time (ms)

Fig. 13. Impact of sag duration on light variation of Bright 5 W lamp: (a) 15% remaining voltage and (b) 12.5% remaining voltage.

tics, except for the ve-cycle event, wherein the light out- The second 5 W tested lamp was the Philips 5 W lamp.
put did not decrease. Fig. 13b shows the light output at Similar to the light output of other LED lamps, the Philips
12.5% remaining voltage, wherein a malfunction condition lamp reached the zero-illumination level starting at the
was observed at the sixth cycle. second cycle at 22.5% remaining voltage. The knee point
 S. Uddin et al. / Measurement 46 (2013) 37833795 3793

(a) 350
Cash 7 W
300 Philips 7 W
Bright 5 W

Dc Bus Voltage (V)

250 Philips 5 W

200

150

100

50

0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275 300
Time (ms)

(b) 24
21
Cash 7 W
LED Array Voltage (V)

18 Philips 7 W
Bright 5 W
15 Philips 5 W
12

0
-50 -25 0 25 50 75 100 125 150 175 200 225 250 275 300
Time (ms)

Fig. 14. Sensitivity of different LED lamps at zero-illumination sag depth for seven cycles on DC bus voltage.
LED Array Voltage*40 (V), DC Bus Voltage (V)

400 100

350 87.5

300 75
Light Output (%)

250 62.5
LED Array Voltage
200 DC Bus Voltage 50

150 37.5
Variation of Light
100 25

50 12.5

0 0
-50 -25 0 25 50 75 100 125 150 175 200 225
Time (ms)

Fig. 15. Relationship between LED array voltage and light output variation with DC bus voltage of Philips 5 W LED lamp at 25% remaining voltage for seven
cycles.

appeared at 25% remaining voltage at the third cycle, indi- Fig. 14 shows the relationships among the variations in
cating that the Philips lamp was more sensitive to both sag light output, LED array voltage, and DC bus voltage across
depth and sag duration than the Bright lamp under the the smoothing capacitor for several LED lamps. The varia-
same power rate. tions in zero-illuminance conditions for sag depth that
3794
100
95
90
85
80
Remaining Voltage (%)
75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50
Fig. 16. Voltage tolerance curves of various LED lamps.
leaves 30% remaining voltage for the Cash 7 W and Philips
7 W lamps, 12.5% for the Bright 5 W lamp, and 25% for the
Philips 5 W lamp at seven cycles were compared with one
another, as shown in Fig. 14a. The decay rate during zero
malfunction differed for all lamps because this rate de-
pends on the size of the DC link capacitor. The current
source converter in Fig. 1 began to malfunction when the
minimum holdup voltage was reached, and then the DC
LED array voltage began to decrease sharply, as shown in
Fig. 14b. Fig. 15 shows the relationship between LED array
voltage and light output variation with DC bus voltage of
the Philips 5 W lamp at 25% remaining voltage for seven
cycles. The DC link capacitors of 4.7, 3.4, 10, and 2.5 lF
were used in the 7 W Cash, 7 W Philips, 5 W Bright, and
5 W Philips lamps, respectively. The holding time from
Eq. (2) increases with capacitor value. Among these lamps,
the Philips lamps are the most sensitive because they use
the smallest DC link capacitors.
Fig. 16 shows the overall tolerance levels of all tested
LED lamps to voltage sags in the form of voltage tolerance
curves and well-known voltage acceptability standards
ITIC and SEMI F47. The upper portion of the standard indi-
cates the appropriate operating region, whereas the lower
portion indicates the turn-off conditions for LED lamp
operation. As can be seen in Fig. 16, the Philips 5 W and
7 W lamps were the most sensitive to sag duration. The
zero-illuminance conditions for these lamps started at
the second cycle. All other tested lamps began to malfunc-
tion for at least four sag cycles. The Philips and Cash 7 W
lamps reached the turn-off conditions at a sag depth of less
than 30% remaining voltage, making them the most sensi-
tive in terms of sag depth. The Bright 5 W lamp was the
least sensitive in terms of sag depth. Several tested LED
lamps were not able to meet the design criteria of the ITIC
and SEMI F47 standards.
5. Conclusion
This paper characterized the power quality-related is-
sues of low-wattage, energy-efcient LED lamps that are
currently being used as lighting sources in the residential,
S. Uddin et al. / Measurement 46 (2013) 37833795
ITIC
SEMI F47
Cash 7W
Philips 7 W
Bright 5 W
Philips 5 W
Time (cycle)
commercial, and industrial sectors. This study investigated
two power quality-related issues, including the harmonic
generation of LED lamps, and the impact of voltage sag.
These studies were performed by conducting the
experiments.
In the rst set of experiments, various types of LED
lamps from different manufacturers were tested to evalu-
ate their harmonic performance in terms type of ballast
used. Experimental results show that all LED lamps gener-
ated harmonics because they contain a power electronic
converter as a ballast to drive the LED arrays. The THDI val-
ues of the tested lamps ranged from 30.94% to 174.38%.
Different manufacturers of LED lamps use different ballast
technologies to reduce harmonic generation, including bal-
lasts with passive lters, active lters, and valley-lled cir-
cuits. Among these lamps, the lamps that use active lters
exhibited better performance and injected only 3035% of
THDI, whereas the ballasts that incorporated valley-lled
circuits injected 6370% of THDI. Lamps with passive lters
injected over 100% of THDI. Therefore, an LED lamp with
active lter circuit is recommended, even though it costs
higher than a lamp without ltering. The diversity factor
remains almost constant as dictated by the individual
lamps when the LED lamps with the same ballast type
are used together. The harmonic generation changes and
decreases the diversity factor when LED lamps with differ-
ent ballast types are used together.
In the second set of experiments, an experimental
investigation was conducted to determine the impact of
voltage sags on LED lamps. Voltage tolerance levels were
plotted to represent the voltage sag sensitivity levels of
several LED lamps. After investigating the light illuminance
variations of several lamps, it can be concluded that the
sensitivity of LED lamps depends on voltage sag depth,
sag duration events, and LED junction voltage, which, in
turn, depend on the internal ballast conguration of the
lamps and DC link capacitance. Some lamps are more sen-
sitive to sag duration than others depending on the design
conguration of the internal ballast circuit. Furthermore,
some LED lamps were found unable to fulll the design cri-
teria of ITIC and SEMI F47 standards.
 S. Uddin et al. / Measurement 46 (2013) 37833795 3795

References [11] S. Uddin, H. Shareef, A. Mohamed, M.A. Hannan, An analysis of

harmonics from LED lamps, in: Proceeding Asia-Pacic Symposium
of Electromagnetic Compatibility, Singapore, May 2124, 2012, pp.
[1] C. DiLouie, Advanced Lighting Controls: Energy Savings, Productivity,
837840.
Technology and Applications, rst ed., Georgia, United State, 2005.
[12] A. Mansoor, W.M. Grady, A.H. Chowdhury, M.J. Samotyj, An
[2] B.J. Huang, P.C. Hsu, M.S. Wu, C.W. Tang, Study of system dynamics
investigation of harmonics attenuation and diversity among
model and control of a high-power LED lighting luminaire, Energy 32
distributed single-phase power electronic loads, IEEE Trans. Power
(2007) 21872198.
Deliv. 10 (1995) 467473.
[3] V. Cuk, J.F.G. Cobben, W.L. Kling, R.B. Timens, An analysis of diversity
[13] H. Shareef, A. Mohamed, K. Mohamed, Sensitivity of compact
factors applied to harmonic emission limits for energy saving lamps,
uorescent lamps during voltage sags: an experimental
in: Proceeding 14th International Conference on Harmonics and
investigation, WSEAS Trans. Power Syst. 5 (2010) 2231.
Quality of Power, Bergamo, Italy, September 2629, 2010, pp. 16.
[14] H. Shareef, A. Mohamed, N. Marzuki, Analysis of ride through
[4] R. Dwyer, A.K. Khan, M. McGranaghan, L. Tang, R.K. McCluskey, R.
capability of low-wattage uorescent lamps during voltage sags, Int.
Sung, T. Houy, Evaluation of harmonic impacts from compact
Rev. Electr. Eng. 4 (2009) 10931101.
uorescent lights on distribution systems, IEEE Trans. Power Syst.
[15] L.P. Frater, N.R. Watson, Light icker sensitivity of high efciency
10 (1995) 17721779.
compact uorescent lamps, in: Proceeding Australasian Universities
[5] International Electrotechnical Commission (IEC): Testing and
Power Engineering Conference, Perth, Australia, December 912,
Measurement Techniques: Voltage Dips, Short Interruptions and
2007, pp. 16.
Voltage Variations Immunity Tests, Std. IEC 61000-4-11, 2004.
[16] K. Taekhyun, M. Rylander, E.J. Powers, W.M. Grady, A. Arapostathis,
[6] B. Wang, X. Ruan, K. Yao, M. Xu, A method of reducing the peak-to-
LED lamp icker caused by interharmonics, in: Proceeding
average ratio of LED current for electrolytic capacitor-less acdc
Conference on Instrumentation and Measurement Technology,
drivers, IEEE Trans. Power Electron. 25 (2010) 592601.
Victoria, Canada, May 1215, 2008, pp. 19201925.
[7] Y.C. Li, C.L. Chen, A novel single-stage high power-factor ac-to-dc
[17] J.M. Alonso, J. Vina, D.G. Vaquero, G. Martinez, R. Osorio, Analysis
LED driving circuit with leakage inductance energy recycling, IEEE
and design of the integrated double buckboost converter as a high-
Trans. Ind. Electron. 59 (2012) 793802.
power-factor driver for power-LED lamps, IEEE Trans. Ind. Electron.
[8] S. Choi, T. Kim, Symmetric current-balancing circuit for LED
59 (2012) 16891697.
backlight with dimming, IEEE Trans. Ind. Electron. 59 (2012) 1698
[18] Institute of Electrical and Electronic Engineering (IEEE): IEEE
1707.
Recommended Practices and Requirements for Harmonic Control
[9] I. Kiyak, V. Topuz, B. Oral, Modeling of dimmable high power LED
in Electrical Power Systems, Std. IEEE 519-1992, 1993.
illumination distribution using ANFIS on the isolated area, Expert
[19] International Electrotechnical Commission (IEC): Limits for
Syst. Appl. 38 (2011) 1184311848.
Harmonic Current Emissions (Equipment Input Current 6 16A per
[10] M.A. Reyes, S. Gallagher, J. Sammarco, Evaluation of visual
phase), Std. IEC 61000-3-2, 2009.
performance when using incandescent, uorescent, and LED
[20] Semiconductor Equipment and Materials International (SEMI):
machine lights in mesopic conditions, in: Annual Meeting of IEEE
Specication for Semiconductor Processing Equipment Voltage Sag
Industry Applications Society, Houston, USA, October 48, 2009, pp.
Immunity. Std. SEMI F47, 2000.
17.