12th International Conference on DEVELOPMENT AND APPLICATION SYSTEMS, Suceava, Romania, May 15-17, 2014

Matlab based Platform for the Evaluation of

Modulation Techniques used in VLC
Steven De Lausnay, Lieven De Strycker, Jean-Pierre Bart Nauwelaers
Goemaere and Nobby Stevens Faculty of Engineering Science, TELEMIC, ESAT
Faculty of Engineering Science, DraMCo Research Group KU Leuven
KU Leuven B- 3000 Leuven, Belgium
B-9000 Gent, Belgium
steven.delausnay@kuleuven.be

Abstract— In this paper, we describe an experimental buildings because of the lighting functionality and the light has
evaluation platform, which enables us to obtain the performance a high directivity compared to Radio Frequency (RF)
of baseband modulation techniques for usage in Visible Light technologies. These two advantages make VLC an interesting
Communication (VLC) applications. For the demonstration of candidate for location-based services. With these services, the
the evaluation platform, we have chosen for Manchester information on your smartphone would depend on your
encoding. This choice is exemplary, the described system is able position within the building. This can be in the form of e.g.,
to deal with a large set of analog and digital modulation indoor guiding or location based information (you get extra
techniques like Non-Return to Zero (NRZ), Return to Zero (RZ), information about the piece of art you are standing in front of
Pulse Position Modulation (PPM), Binary Frequency Shift
at that moment).
Keying, Dual Tone Multi-Frequency (DTMF), Phase Shift
Keying (PSK), … For the Manchester encoding, it is In order to consider VLC as an accurate localization
experimentally demonstrated that the statistical distribution of system, the following conditions must be met:
the received signal after matched filter operation, is Gaussian in
nature. This allows us to make an estimation of the Bit Error • Positioning should be more accurate compared to the
Rate (BER) using Q-functions for Gaussian distributions. Our existing RF wireless indoor localization systems.
results demonstrate that the estimated BER strongly depends on
the sampling rate at the receiver end. The sampling rate will • The potential degradation of lighting function as a
therefor determine whether or not the receiver can receive Non- consequence of the overlay communication system
Line Of Side (NLOS) signals at normal light intensities. should be held to a minimum, which corresponds to a
unnoticeable difference for the observer.
Keywords— Baseband Modulation; Bit Error Rate; Evaluation
Platform; Gaussian Noise distribution; Visible Light
• In order to get a rapid implementation, the
Communication infrastructural impact should be very low.
For a localization system, every LED broadcasts its own
I. INTRODUCTION unique identification (ID) and the user, equipped with a
smartphone and photodiode, receives these IDs. Based on these
Since a couple of years, the Light Emitting Diode (LED)
IDs and a previously downloaded map of the indoor
has become a major player in the market of indoor and outdoor
environment (containing the IDs of the LEDS and their
lighting applications. LEDs as a lighting technology offer a
position), the location can be calculated.
number of important advantages with regard to other lighting
technologies, such as increased efficiency, high brightness, For the transmission of the IDs, we want to explore a VLC
color selection without significantly compromising the system using Code Division Multiple Access (CDMA). As a
efficiency, etc. [1]. Another distinctive characteristic of LEDs first step, we have to evaluate the single source to single
with regard to more classical lighting technologies is the larger receiver behavior of different modulation techniques. With the
bandwidth. This has opened the door for Visible Light proposed evaluation system, a wide range off modulation
Communication systems (VLC), in which the original function techniques can easily be evaluated. The goal of this paper is to
of lighting is combined with the creation of a wireless, optical describe this set-up and to demonstrate its functionality
communication system. An important boundary condition is of towards the BER evaluation for Manchester encoding. Specific
course that the lighting functionality is not jeopardized by the emphasis is put on the discrete sampling property of the
data communication system. Due to the large bandwidth of acquisition chart.
LEDs, this condition can be met.
The paper is organized as follows. In section II, the general
Because of the interesting properties of VLC, we see this description of the platform is given, with an emphasis on the
technology as a new solution for localization in large indoor specific hardware components that were used. In section III,
environments. The infrastructure for VLC is already in we focus on one modulation technique, namely Manchester

978-1-4799-5094-2/14/$31.00 ©2014 IEEE 57

 A. Transmitter
At the transmitter side, random bits are generated at a bit
rate Rb. The bits are modulated with a baseband modulation
technique, like Non-Return to Zero (NRZ), Return to Zero
(RZ), Manchester Coding, Pulse Position Modulation (PPM),
Binary Frequency Shift Keying (BFSK), etc. These are all
baseband modulation techniques with low complexity, so they
can be easily implemented. The modulated sequence is up-
sampled by N where N is the ratio of the bit period (Tb) to the
sample period (TS). The up-sampled sequence is fed to the
pulse shaping, which produces a discreet time pulse train at the
sample rate TS (Equation 1). The discrete time pulse train is
directly fed to the LED driver through an analog or digital
output port of the DAQ which results in an analog signal s(t)
(Equation 2) [2]. The a(k) coefficients depends on the chosen
modulation technique. Further details can be found in [2].

s ( nTS ) = ∑ a ( k ) p ( nTS − kTb ) (1)

k

s (t ) = ∑ a ( k ) p (t − kTb ) (2)
Fig. 1 Practical set-up room
k
encoding. The sampling rate is left as a parameter in a non line
of sight configuration. Based on our observation, it is clear that In order to accelerate the implementation of the technology,
the BER strongly depends on the sampling rate at the receiver we use LEDs, which are currently used for lighting
end. At the end, a summarized overview of the main topics applications. High power LEDs currently used for lighting
covered in this paper can be found in the conclusions applications are composed of a blue LED with a phosphor
section IV. layer. Previous research has show that the phosphor layer will
limit the optical bandwidth of the LED and thus the maximum
II. DESCRIPTION OF THE PLATFORM data rate for communication [3]–[5]. For the experimental set-
up described in this paper, the Bridgelux BXRA-30E0740-A-
Fig. 1 gives an overview of the point-to-point VLC system 00 is used. This is a neutral white LED with a Lambertian C-
under consideration. The system consists of three building plane and a total Luminous flux of 820 lm. For the tests, the
blocks, namely the transmitter (LED) which is mounted at the LED is mounted onto the ceiling at a height of 3.5 m (Fig. 1).
ceiling, the receiver (photodiode) placed on a desk and the
room (channel) with a window, to introduce additional noise. As driving circuit for the LED, there are two possibilities
namely a LED driver that is currently used in lighting
An overview of the different blocks of the evaluation applications or a Trance Conductance Amplifier (TCA). When
platform is given in Fig. 2. For quick evaluation and no a LED driver is used, the digital Pulse Width Modulation
synchronization problems, the transmitter and receiver are both (PWM) input can be used to inject the data into the LED
implemented in Matlab. The connection to the hardware is current. To perform the measurements in this paper, a circuit
made with the National Instruments USB-6211 Data was build with the LT3756 from Linear Technologies. This
Acquisition box (DAQ). driver can control LEDs up to 30 W electrical power. It was

Fig. 2 The practical evaluation system

58
 synchronization issues. Remark that the delay due to the
channel is in the order of several nanoseconds for indoor
environments. This can be neglected for data rates up to several
Mbps [9]. The sampled version of the received signal (r(nTs))
is filtered by a matched filter with a impulse response which is
the time reversed version of the pulse shape filter. The output
of the matched filter is given by Equation 3 [2].

x ( nTS ) = ∑ r ( mTS ) p (( m − n ) TS ) (3)

m
The matched filter output is down-sampled to one
sample/bit at
n = kT / T
b S which produces a sequence x(kTb).
The decision block will estimated the received bits and
compare them with the send bits to calculate the BER.
For the performed test in Section III, the receiver is placed
on a desk at a distance of 2.7 m from the LED. Furthermore,
we have rotated the receiver over 90 degrees, in order to have a
non line of sight (NLOS) configuration. As can be seen on
Fig. 3, there is at least one reflection required for the
transmitted data to reach the photodiode.
Fig. 3: Practical set-up room NLOS
C. Room
possible to get bit rates up to 2 Mbps before the functionality of The room is a normal working office with gray plaster
the driver was corrupted. The driver is limited for the number walls and a large window over the entire length of one side of
of modulation types possible namely only digital modulation the office. The room will have an impact on the channel
types with full On-Off Keying (OOK). impulse response and will allow noise to intrude the signal for
an optical communication link. Because the data is traveling at
When analog modulation techniques or digital modulation
light speed, the delay spread will be in the order of
techniques with more flexibility in modulation depth want to
nanoseconds. For localization purposes, a high bit rate is not
be evaluated, a TCA can be used. In [6] they proposed a block
necessary, so the influence of Inter Symbol Interference (ISI)
diagram of a TCA transmitter for VLC. This block diagram
can be neglected [9].
leaves a lot of possibilities for the implementation design,
where we build several transmitters. The final design has an Two important noise sources in a wireless optical link are
electrical bandwidth of about 20 MHz and can handle digital shot noise and terminal noise. Terminal noise is introduced by
and analog signals [7]. The TCA can drive the LED with a the pre-amplifier electronics and is thereby signal independent
current swing from 0 mA up to 1000 mA which is enough to Gaussian noise [9]–[11]. Terminal noise will be the dominant
drive LEDs currently used in lighting applications which have noise source when there is less ambient light (sun light, other
typical currents from 350 mA or 700 mA. light sources, etc.). When a well-developed pre-amplifier is
used, the limiting noise factor is shot noise. This noise source is
B. Receiver introduced by ambient light as well as the received signal
The receiver part starts at the photo diode (Fig. 2) where the power and can be modeled as Additive White Gaussian Noise
incoming light is transformed to a current. The generated (AWGN) [9]–[11]. As shown in Fig. 1 the room has a window
current will depend on the responsivity curve of the that will allow sunlight to fall in onto the receiver, which will
photodiode. Because the LED has a high blue component in the introduce shot noise.
spectrum [7], the responsivity curve of the photo diode should
be high in the wavelength region of 400 nm. This specification III. RESULTS
was crucial for the selection of the appropriate detector. After
some research, a detector from Thorlabs was found to be the A. Modulation techniques
most suited to fulfill this requirement. The PDA10A is a fixed The modulation technique we have chosen is the
amplified detector with a Si-photo diode which is sensitive in Manchester encoding. Remark that our set-up is not limited to
the visible light region and has a active area of 0.8 mm2. The Manchester encoding; other binary line codes can be easily
current from the photodiode is converted to a voltage with a implemented in the Matlab environment as well analog
Trans-Impedance Amplifier (TIA). The PDA10A has a build in modulation techniques. Manchester encoding is a split-phase
TIA with a gain of 10 kV/A [8]. The output voltage of the TIA code, which is well suited for our purpose. In case of non-
is amplified a last time with a voltage amplifier, resulting in the return to zero codes, a long sequence of zero’s could introduce
analog signal r(t). The signal r(t) is the transmitted signal s(t) flickering of the light intensity, which violates the
plus noise from the surrounding light. The received signal r(t) compatibility with the normal lighting function. Remark that
is sampled using the DAQ at the same rate of the transmitted the bandwidth of the human eye is around 60 Hz. The zero
signal with synchronized clocks. This way, there are no

59
 Fig. 7 Distribution x(kTb) for N = 50
Fig. 5 The average value µ as function of N

Fig. 6 Distribution x(kTb) for N = 200

crossing bandwidth of Manchester encoding is twice the Fig. 4 The standard deviation σ as function of N
inverse of the bit time Tb, which has a positive impact on
flickering, even at lower data rates.
Other advantages of Manchester coding compared to other approximation as the number of sample points increases. In
baseband modulation techniques, are the easy clock recovery order to evaluate the impact of the sampling rate, we have
(bit synchronization) and the received signal can be AC- chosen a high sample rate, namely 200 kSps, which results in
coupled at the receiver. In a VLC system the transmitted signal 200 (= N) samples every bit time Tb. Once we have this high
can not be negative (s(t) > 0) and so also the received signal at sample ratio, we can easily reduce N by only selecting a
the photo diode. reduced set of these samples.

In the specific experiment we performed, 106 bits were sent On Fig. 4 and Fig. 5, one can observe the distribution of the
at a data rate of 1 kbps by use of Manchester encoding. The measured x(kTb) samples for N = 50 and N = 200 when the
zero crossing bandwidth is 2 KHz and the duration of the set-up is used of Fig. 3. An observation one can make is that
experiment was 1000 seconds, or almost 17 minutes. the scaled histograms can be well approximated by a Gaussian
distribution. We have plotted the magnitude of the average
B. Influence of the sample rate value µ and the standard deviation σ as a function of N. The
result can be found on Fig. 6 and Fig. 7. There is no distinction
One parameter that is left undetermined right now is the between these trends for a 0 or 1-bit sent. To determine
ratio N, defined as the sample rate (1/TS) to the bit rate (1/Tb). whether the histograms of the received data have a Gaussian
According to the zero crossing bandwidth and the Nyquist distribution, a KolmogorovSmirnov test (KS-test) was
sampling theorem, a value for N of 4 should be sufficient. The performed. This test confirmed that the received data has a
critical component in the practical evaluation system of Fig. 2 Gaussian distribution, so proves that the noise introduced in a
is the discrete calculation of the matched filter output. The VLC is AWGN.
matched filtering operation, which is basically the calculation
of an integral in the continuous time domain, is approximated Based on the histogram plots, it is clear that lower values of
by a finite sum. It is clear that this sum is a better N will lead to an increased BER. This trend is clearly

60
 For an indoor localization system, this will be an imported
factor, which will determine if the positioning algorithm could
take into account reflections. When there are a lot of LOS
signals, it will be better to decrease the sample rate, so
reflections don’t influence the positioning algorithm. When
there are no LOS signals, the sample rate could be increased to
still receive an ID from LEDs so a rough estimation of the
position could be made. By changing the sample rate, a hybrid
system could me made where a rough position estimation could
be made when there are no LOS signal and an accurate position
could be calculated when different LOS signals are received.

ACKNOWLEDGMENT
The authors would like to thank the Agency for Innovation
by Science and Technology in Flanders (IWT).

Fig. 8 The estimated BER as function of N REFERENCES

[1] Company McKinsey, “Lighting the way: Perspectives on the global
lighting market,” McKinsey & Company, August,
http://img.ledsmagazine.com/pdf/LightingtheWay.pdf, 2011.
demonstrated on Fig. 8. The test was also repeated with the set-
[2] M. Rice, “Digital Communications: A Discrete-Time Approach,”
up from Fig. 1. These results showed the same trend for the Pearson Education, ISBN: 978-0-13-030497, 2009.
BER as the one shown on Fig. 8 but a couple of orders lower. [3] D. C. O’Brien and L. Zeng and H. Le-Minh and G. Faulkner and J. W.
This shows that the sample rate (1/TS) will strongly determine Walewski and S. Randel, “Visible Light Communications challenges
if the receiver will be able to also receive NLOS signals in a and possibilities, Personal, Indoor and Mobile Radio Communications,”
VLC system. This can be an imported factor when the system 2008.
is used as an indoor localization system. [4] L. Zeng and D. O’Brien and H. Le-Minh and K. Lee and D. Jung and Y.
Oh, “Improvement of Date Rate by using Equalization in an Indoor
Visible Light Communication System,” IEEE, 2008.
[5] H.Le-Minhand and D. O’Brienand and G. Faulknerand and L. Zengand
IV. CONCLUSION and K.Lee and D. Jung and Y. Oh, “80 Mbits/s Visble Light
Communications Using Pre-Equalized White LED,” IEEE ECOC, 2008.
In this paper we have presented a platform that allows us to [6] R.Baumgather, A. Kornbichler, J. W. Walewski, “High-power high-
evaluate analog and digital modulation techniques. With this bandwidth linear driving circuit for VLC applications,” IEEE P802.15.7
platform, modulation techniques like NRZ, RZ, PPM, Task group Visible-Light Communication. March 2010
Manchester Coding, BFSK, DTMF, PSK, etc. can easily be [7] Steven De Lausnay, Lieven De Strycker, Jean-Pierre Goemaere, Bart
synthesized and analyzed using Matlab. The system allows us Nauwelaers and Nobby Stevens, “Design of a Visible Light
Communication Transmitter for the Evaluation of a Wide Range of
to quickly compare the different modulation techniques and to Modualtion Techniques,” IWOW 2013, October 2013
evaluate which encoding technique is most suitable to be used
[8] Thorlabs, “PDA10A,”
in an indoor localization system based on LED lighting. In this http://www.thorlabs.de/thorproduct.cfm?partnummer=PDA10A
paper, Manchester Coding is chosen because it doesn’t [9] T. Komine, “Visible Light Communication and Its Fundamental Study,”
introduce flickering at low bit rates, an easy bit synchronization PhD Thesis, 2005
is possible and the received signal can be amplified with an [10] T. Komine and M. Nakagawa, “Performance Evaluation of Visible-Light
AC- coupled amplifier. Wireless Communication System using White LED Lightings,” IEEE,
2004.
The results of the practical set-up have shown that the BER [11] J. M. Kahn and J. R. Barry, “Wireless Infrared Communications,” Pro-
of a VLC system will depend on the sample rate of the received ceedings of the IEEE, February, No. 2, Vol. 85, 1997.
signal. When the ratio of the sample rate to bit rate increases,
this decreases exponentially the BER of the system. For e.g., a
NLOS signal, the ratio sample rate over bit rate will determine
whether the BER becomes acceptable or not.

61