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

net/publication/241637721

Signal Shaping and Modulation for Optical Wireless Communication

Article  in  Journal of Lightwave Technology · May 2012

DOI: 10.1109/JLT.2012.2188376

CITATIONS READS
87 1,714

3 authors:

Svilen Dimitrov Sinan Sinanovic

German Aerospace Center (DLR) Glasgow Caledonian University
35 PUBLICATIONS   1,341 CITATIONS    92 PUBLICATIONS   4,803 CITATIONS   

SEE PROFILE SEE PROFILE

Harald Haas
The University of Strathclyde
611 PUBLICATIONS   34,667 CITATIONS   

SEE PROFILE

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

Visible light communications View project

Terabit Bidirectional Multi-user Optical Wireless System (TOWS) for 6G LiFi View project

All content following this page was uploaded by Svilen Dimitrov on 17 April 2015.

The user has requested enhancement of the downloaded file.

 1

Signal Shaping and Modulation for Optical Wireless

Communication
Svilen Dimitrov, Student Member, IEEE, Sinan Sinanovic, Member, IEEE, and Harald Haas, Member, IEEE

Abstract—In this paper, a signal shaping framework for optical schemes such as M -PPM and M -PAM [3, 4]. At high data
wireless communication (OWC) is proposed. The framework is rates, where the 3-dB bandwidth of the pulse exceeds the
tailored to the single-carrier pulse modulation techniques, such coherence 3-dB bandwidth of the optical wireless channel,
as multi-level pulse position modulation (M -PPM) and multi-
level pulse amplitude modulation (M -PAM), and to multi-carrier the RMS delay spread of the channel impulse response ex-
transmission realized through multi-level quadrature amplitude ceeds the pulse duration. Therefore, such techniques suffer
modulation (M -QAM) with orthogonal frequency division mul- from severe inter-symbol interference (ISI), limiting their
tiplexing (OFDM). Optical OFDM (O-OFDM) transmission is throughput. In order to compensate for the channel effect,
generally accomplished via direct-current-biased optical OFDM
(DCO-OFDM) or asymmetrically clipped optical OFDM (ACO- the optimum receiver employs maximum likelihood sequence
OFDM). Through scaling and DC-biasing the transmitted sig- detection (MLSD) [4]. Here, the MLSD algorithm chooses
nal is optimally conditioned in accord with the optical power the sequence of symbols that maximizes the likelihood of the
constraints of the transmitter front-end, i.e. minimum, average received symbols with the knowledge of the channel taps.
and maximum radiated optical power. The OWC systems are
Even though the Viterbi algorithm can be used for MLSD
compared in a novel fashion in terms of electrical signal-to-
noise ratio (SNR) requirement and spectral efficiency as the to reduce the computational effort, the complexity of MLSD
signal bandwidth exceeds the coherence bandwidth of the optical still grows exponentially with the number of channel taps.
wireless channel. In order to counter the channel effect at Therefore, in practical system implementations, suboptimum
high data rates, computationally feasible equalization techniques equalization techniques with feasible complexity are used.
such as linear feed-forward equalization (FFE) and non-linear
decision-feedback equalization (DFE) are employed for single- These include the linear FFE or the non-linear DFE with
carrier transmission, while multi-carrier transmission combines zero forcing (ZF) or minimum mean squared error (MMSE)
bit and power loading with single-tap equalization. It is shown criteria [4]. The superior bit-error ratio (BER) performance
that DCO-OFDM has the highest spectral efficiency for a given at a lower SNR requirement of DFE comes at a significantly
electrical SNR at high data rates when the additional direct
current (DC) bias power is neglected. When the DC bias power
increased complexity as compared to FFE [4].
is counted towards the signal power, DCO-OFDM outperforms Multi-carrier modulation has inherent robustness to ISI,
PAM with FFE, and it approaches the performance of the more because the symbol duration is significantly longer than the
computationally intensive PAM with DFE. RMS channel delay spread. As a result, M -QAM O-OFDM
Index Terms—Wireless communication, optical devices, signal promises to deliver very high data rates [2]. Because of the
processing, pulse modulation, OFDM. common use of a cyclic prefix (CP), the channel frequency
response can be considered as flat fading over the subcar-
I. I NTRODUCTION rier bandwidth [5, 6]. Thus, single-tap linear FFE with low
complexity paired with bit and power loading can be used
WC has proven to be a promising candidate for medium
O range high-speed data transmission with a potential to
deliver several hundreds of Mbps data rate [1, 2]. In addition
to minimize the channel effect [7, 8]. In the literature, two
possible O-OFDM system realizations can be found: DCO-
OFDM [9] and ACO-OFDM [10]. ACO-OFDM shows a
to being a complementary non-interfering solution alongside greater optical power efficiency at the expense of a 50%
radio frequency (RF) technology, OWC has the advantage of reduction in spectral efficiency as compared to DCO-OFDM.
license-free operation over a significantly wider spectrum. Imperfections of the optical front-ends due to the use of off-
The data transmission in OWC is achieved through intensity the-shelf components result in a limited linear dynamic range
modulation and direct detection (IM/DD). Suitable candidates of radiated optical power [11]. Therefore, the transmitted sig-
for data modulation are the single-carrier pulse modulation nal is constrained between levels of minimum and maximum
Manuscript received August 16, 2011; revised November 18, 2011 and optical power. In addition, the average optical power level is
January 26, 2012. constrained by the eye safety regulations [12] and/or the design
Svilen Dimitrov, Sinan Sinanovic and Harald Haas are with the University requirements. In order to condition the signal in accord with
of Edinburgh, Institute for Digital Communications, Joint Research Insti-
tute for Signal and Image Processing, Edinburgh EH9 3JL, UK (e-mail: these constraints, signal scaling in the digital signal processor
{s.dimitrov,s.sinanovic,h.haas}@ed.ac.uk). (DSP) and DC-biasing in the analog circuitry is required.
This paper has been accepted for publication in the IEEE/OSA Journal of Since the M -PPM and M -PAM signals have a probability
Lightwave Technology (IEEE/OSA JLT), vol. 30, no. 9, pp. 1319 − 1328,
May 2012, DOI: 10.1109/JLT.2012.2188376. density function (PDF) with a finite support, they can fit the
⃝2012
c IEEE. Personal use of this material is permitted. Permission from constraints without signal clipping. However, scaling and DC-
IEEE must be obtained for all other uses, in any current or future media, biasing of the Gaussian time domain signals in ACO-OFDM
including reprinting/republishing this material for advertising or promotional
purposes, creating new collective works, for resale or redistribution to servers and DCO-OFDM result in a non-linear signal distortion which
or lists, or reuse of any copyrighted component of this work in other works. is precisely analyzed in [13]. In this paper, the analysis is
2

employed in the formulation of the optimum signal scaling and in the flat fading channel. When the signal bandwidth exceeds
DC-biasing to minimize the required electrical SNR per bit for the channel coherence bandwidth, DCO-OFDM outperforms
a target BER. In general, in visible light communication (VLC) PAM with FFE, and it approaches the spectral efficiency of the
systems, the DC bias power is employed for illumination more computationally intensive PAM with DFE, while ACO-
as a primary functionality. Therefore, it can be excluded OFDM outperforms PPM with FFE and DFE.
from the calculation of the electrical signal power invested The rest of the paper is organized as follows. Section II
in the complementary data communication. In infrared (IR) presents the system model and the signal shaping framework
communication systems, the DC bias power is constrained by for M -PPM, M -PAM and M -QAM O-OFDM. Single-carrier
the eye safety regulations [12], and it is generally included in and multi-carrier transmission are compared in terms of elec-
the calculation of the electrical SNR. trical SNR requirement and spectral efficiency in Section III.
On-off keying (OOK), essentially 2-PAM, and M -PPM Finally, Section IV concludes the paper.
have been compared in terms of electrical and optical power
requirement in a dispersive channel with equalization in [3]. II. S YSTEM MODEL AND SIGNAL SHAPING
An increasing power requirement is demonstrated with the
increase of the RMS channel delay spread or, equivalently, data The conventional discrete model for a noisy communication
rate. In a later study [14], M -PPM, M -PAM and multi-carrier link is employed in this study:
M -QAM transmission, similar to M -QAM DCO-OFDM, have
y =h∗x+n, (1)
been compared assuming a flat fading channel in terms of
optical power requirement and spectral efficiency. However, where y represents the received replica of the transmitted
an unlimited non-negative dynamic range of the transmitter is signal, x, which is convolved with the channel impulse re-
considered which is hardly achievable in practice. Here, the sponse, h, and it is distorted by additive white Gaussian noise
non-negative M -QAM signal is scaled down to accommodate (AWGN), n, at the receiver. In OWC, n has a zero-mean real-
the large peak-to-average-power ratio (PAPR), resulting in valued Gaussian distribution. After optical-to-electrical (O/E)
an increased optical power requirement. Recently, a similar conversion, it has an electrical power spectral density (PSD)
comparison has been reported in [15] for the multi-carrier of N0 /2 in M -PPM and M -PAM. In optical OFDM with
transmission schemes ACO-OFDM and DCO-OFDM with M -QAM, the PSD of n amounts to N0 because of the two
a tolerable clipping distortion. To the best of the authors’ dimensional constellation [16]. Here, ∗ stands for discrete
knowledge, there is no comprehensive framework in literature linear convolution. Without loss of generality, the system
which enables the comparison of single-carrier and multi- analysis is presented in terms of discrete signal vectors. Here,
carrier transmission schemes in terms of spectral efficiency x contains Zx samples, h has Zh samples, and as a result,
and electrical SNR requirement in a dispersive realistic optical n and y have Zx + Zh − 1 samples [16]. The discrete signal
wireless channel. In addition, a study on signal shaping for a vectors are obtained by sampling of the equivalent continuous-
practical dynamic range of the transmitter front-end, where the time signals. The sampling rates over a time period of T
DC bias power is excluded or included in the calculation of differ in the considered systems, and the details are presented
the SNR is still considered an open issue. below. Through scaling and DC-biasing, x can be conditioned
In this paper, a signal shaping framework is proposed for within the optical power constraints of the transmitter front-
M -PPM, M -PAM and M -QAM O-OFDM which through end. The non-linear transfer characteristic of the LED can be
scaling and DC-biasing conditions the signals to fit within the compensated by pre-distortion [17]. A linear dynamic range
optical power constraints of the transmitter front-end. For the of the transmitter is obtainable, however, only between levels
Gaussian O-OFDM signals in particular, the signal shaping is of minimum and maximum radiated optical power, PTx,min
optimum, i.e. the required electrical SNR is minimized. The and PTx,max [13]. Furthermore, the eye safety regulations [12]
systems are compared in a novel fashion in terms of electrical and/or the design requirements constrain the level of radiated
SNR requirement and spectral efficiency in the dispersive average optical power to PTx,avg . The signal scaling and DC-
optical wireless channel, excluding or including the DC bias biasing are discussed in detail for OWC schemes below.
power in the calculation of the electrical SNR. When the It has been shown in [3] that line-of-sight (LOS) and
additional DC bias power is neglected, DCO-OFDM and PAM non-line-of-sight (NLOS) optical wireless channels can be
show the greatest spectral efficiency for a flat fading channel accurately modeled by the impulse response function h(t) =
in the SNR region above 6.8 dB. However, since O-OFDM gh(opt) f (t), where f (t) = U (t)6a6 /(t + a)7 . Here, gh(opt)
with bit and power loading suffers a lower SNR penalty than stands for the optical path gain coefficient, U (·) is the unit
PAM with DFE as the signal bandwidth exceeds the coherence step function,
√ and a is related to the RMS delay spread, D, by
bandwidth of the dispersive optical wireless channel, DCO- a = D 11/13. The 3-dB coherence bandwidth of the channel
OFDM demonstrates a superior spectral efficiency. When can be expressed as Bc = 1/(5D) [18]. RMS delay spreads
the DC bias power is counted towards the electrical signal between 1.3 ns and 12 ns are reported for LOS links, whereas
power, DCO-OFDM and ACO-OFDM suffer a greater SNR RMS delay spreads between 7 and 13 ns are reported for
penalty due to the DC bias as compared to PAM and PPM, NLOS links [3]. The channel taps in the vector h are obtained
respectively. However, the presented optimum signal shaping by sampling of the channel impulse response at the sampling
framework enables O-OFDM to greatly reduce this penalty and frequency of the received signal, y. The optical path gain can
√
minimize the gap to single-carrier transmission within 2 dB be expressed as gh(opt) = IPD SPD ρPD GTIA /(E [x] rload ),
 3

where IPD denotes the average irradiance of the PD, SPD

is the photosensitive area of the PD, ρPD is the responsivity
of the PD, GTIA is the gain of the transimpedance amplifier
(TIA), E [x] is the average transmitted optical power, where
E [·] stands for the expectation operator, and rload is the load
resistance over which the received current is measured [19,
20]. In addition, the optical path gain can be related to the
electrical path gain, gh(elec) , as follows:
∫ Fig. 1: Block diagram of single-carrier transmission in OWC
1 B/2
2
gh(elec) = gh(opt) |F (f )|2 df , (2) using pulse modulation.
B −B/2
where F (f ) is the Fourier transform of f (t), and B is the than 64, also ensures that the time domain signal follows a
double-sided signal bandwidth. close to Gaussian distribution [21]. This assumption greatly
In M -PPM and M -PAM, the RMS delay spread of the simplifies the derivations throughout the paper. In addition,
channel becomes comparable to or larger than the pulse the CP transforms the linear convolution with the channel into
duration at high data rates which causes a severe ISI. Equiv- a cyclic convolution, facilitating a single-tap linear FFE and
alently, the signal bandwidth exceeds the channel coherence eliminating the need for a non-linear DFE. Even though the
bandwidth. In general, similar effects are also caused by the channel can be considered as flat fading over the individual
low-pass frequency response of the front-end components, subcarriers, the non-flat channel frequency response over the
such as LEDs, PDs and amplifiers. As a result, the BER entire OFDM frame still leads to an SNR penalty for the
performance is degraded, and the systems effectively incur an average frame BER. Here, the single-tap equalizer is generally
SNR penalty. In practical system implementations, multi-tap paired with bit and power loading [7, 8], in order to minimize
linear FFE and non-linear DFE with ZF or MMSE criteria are this SNR penalty. Here, the gain factor of the equalizer, GEQ ,
deployed to reduce the SNR penalty. Since an equalizer with is obtained via a Monte Carlo simulation.
an MMSE criterion requires a higher computational effort, and
it only reduces the SNR penalty by approximately 0.5 dB as
compared to the ZF criterion, ZF is generally employed. The A. M -PPM
gain factor, GEQ , of a linear ZF FFE is given as follows [4]: The block diagram of single-carrier transmission with pulse
( ∫ )−1 modulation is presented in Fig. 1. In M -PPM, log2 (M )
1 B/2 1 equiprobable input bits form a time domain symbol. It is a
GEQ = df , (3)
B −B/2 W (f ) sequence
√ of M chips, where one chip has a current level
of M Ps(elec) , and the other M − 1 chips are set to zero.
where W (f ) is expressed as follows: Here, Ps(elec) is the average electrical power of the M -PPM
∞
∑ symbol, and it is related to the average electrical energy per bit,
W (f ) = |V (f − nB)H(f − nB)|2 . (4) Eb(elec) , as follows: Eb(elec) = M Ps(elec) /(log2 (M )B). The
n=−∞ M -PPM symbol with a double-sided bandwidth of B = M/T
Here, V (f ) and H(f ) are the Fourier transforms of the has a duration of T , and it is grouped in the train of L symbols,
impulse responses of the pulse shaping filter at the transmitter, sl , where l, l = 0, 1, ..., L − 1, is the symbol index. Thus, the
v(t), and the optical wireless channel, h(t), respectively. The spectral efficiency of M -PPM is log2 (M )/M bits/s/Hz [3, 4].
gain factor of a non-linear ZF DFE is given as follows [4]: The train of symbols, sl , is scaled by a factor, α, in order to fit
( ∫ ) within the front-end optical power constraints. Next, the signal
1 B/2 ( ) is passed through a digital-to-analog (D/A) converter. Here, a
GEQ = exp ln W (f ) df . (5)
B −B/2 pulse shaping filter with a real-valued impulse response of
v(t) is applied to transform the train of digital chips into a
The gain factor GEQ represents the theoretical lower bound
train of continuous-time pulses. In M -PPM, because of the
for the electrical SNR penalty which the BER performance
fact that the information carrying pulse has an optical power
incurs at high data rates. This lower bound is achieved when
level greater than PTx,min , zero bias is required. As a result,
an infinite number of channel taps are considered in the FFE
the transmitted signal vector, x, has a length of Zx = LM ,
and DFE which is hardly achievable in practice because of the
and it can be expressed as follows:
significantly increased computational complexity.
In multi-carrier systems such as OFDM-based OWC, the xl = αsl , (6)
RMS delay spread is significantly shorter than the symbol ( ) √
duration, and therefore the equalization process is considerably where α = min PTx,max , M PTx,avg / M Ps(elec) . (7)
simplified to single-tap equalization [6]. The ISI and the inter- The transmitter front-end constrains PTx,min and PTx,max ,
carrier interference (ICI) are completely eliminated by the whereas PTx,avg is independently imposed by the eye-safety
use of a large number of subcarriers and a CP which has a regulations and/or the design requirements. In general, con-
negligible effect on the electrical SNR requirement and spec- straining the average optical power level to E [xl ] ≤ PTx,avg
tral efficiency [5]. A large number of subcarriers, e.g. greater results in a suboptimal BER performance of the OWC systems.
4

The best BER performance is obtained when this constraint is B. M -PAM

relaxed, i.e. when E [xl ] is allowed to assume any level in the
dynamic range between PTx,min and PTx,max .
In order to relate the average optical symbol power, Ps(opt) , The block diagram of M -PAM is presented in Fig. 1.
to the electrical symbol power, Ps(elec) , the signal is subjected Here, log2 (M ) equiprobable input bits form a time domain
to O/E conversion defined as follows: symbol with a double-sided bandwidth of B = 1/T and a
[ ] duration
√ of T√ . The symbols are assigned to current levels of
E x2l 2 p 3Ps(elec) / (M − 1)(M + 1), p = ±1, ±3, ..., ±M − 1,
Ps(elec) = 2 Ps(opt) , (8)
( E [x ] )2 and these are grouped in the train of L symbols, sl . Here,
2
l
[ ]
where E [xl ] = min PTx,max /M, PTx,avg and E x2l = Eb(elec) = Ps(elec) /(log2 (M )B). The resulting spectral effi-
( )2 ciency of M -PAM is log2 (M ) bits/s/Hz [3, 4]. The train of
min PTx,max , M PTx,avg /M in M -PPM. Since the time symbols is scaled and passed through the D/A converter. Since
domain signal in M -PPM has a PDF with a finite support, sl is bipolar, it requires a DC bias, βDC , to fit within the front-
it can be fitted within PTx,min and PTx,max without clipping. end optical power constraints. The transmitted signal vector,
Thus, the following holds ( for its average optical ) signal power: x, has a length of Zx = L, and it can be expressed as follows:
E [xl ] = Ps(opt) = min PTx,max /M, PTx,avg ≤ PTx,avg .
xl = αsl + βDC , (11)
In this paper, the BER performance of the OWC systems is where
compared for equal average electrical signal power, Ps(elec) , √
M +1 ( )
and equal bandwidth, B. In addition, the BER is assessed α= min PTx,max −βDC , βDC −PTx,min .
as a function of the electrical SNR per bit, i.e. the average 3(M − 1)Ps(elec)
electrical bit energy normalized to the power spectral density (12)
of the AWGN, γb(elec) = Eb(elec) /N0 . In order to obtain the E/O conversion in M -PAM from (8),
In M -PPM, the received signal, yl , is passed through a the second moment of x can be expressed as follows:
matched filter, and at the analog-to-digital (A/D) converter it is [ ] M +1 ( )2
sampled at a frequency of M/T or higher [3, 4]. The BER sys- E x2l = min PTx,max − βDC , βDC − PTx,min
3(M − 1)
tem performance in the optical wireless channel with AWGN 2
+βDC .
and equalization has been discussed in [3]. The received (13)
M -PPM symbol can be treated as an on-off-keying (OOK) Because of the fact that the M -PAM time domain signal has
sequence, and the information bits can be decoded by means a PDF with a finite support, it can be fitted within PTx,min
of a hard-decision decoder. For this approach, an analytical and PTx,max without clipping. Thus, the following holds for
union bound of the BER as a function of the electrical SNR per its average optical power: E [xl ] = Ps(opt) = βDC ≤ PTx,avg .
bit is presented and verified through simulation. Alternatively, The received signal, yl , is passed through a matched filter,
the BER performance can be enhanced by means of soft- and at the A/D converter it is sampled at a frequency of 1/T or
decision decoding based on the position of the chip with the higher [3, 4]. After equalization of the channel effect, a hard-
maximum level within the received M -PPM symbol. However, decision decoder can be employed to obtain the received bits.
the analytical BER performance of this decoder is not derived. As a result, the effective electrical SNR per bit in M -PAM,
A union bound for the symbol error rate (SER) in soft-decision Γb(elec) , can be expressed as follows:
decoding can be obtained as a summation of the probabilities
of chips ci , i = 2, 3, ..., M , being greater than an intended chip Γb(elec) = γb(elec) GEQ GDC . (14)
c1 within the M -PPM symbol. Since ci are equally probable,
Here, GEQ is given in (3) and (5). The gain factor GDC
a union bound for the SER can be expressed as follows:
denotes the attenuation of the useful electrical signal power of
SER ≤ (M − 1)P (c2 |c1 ) x due to the DC component, and it is given as follows [13]:
∫ ∞ ( √ ) [ ]
s 2 1 E (xl − βDC )2
= (M − 1) Q √ √ GDC = . (15)
N0 πN0 E [x2l ]
 ( −∞ √ )2  (9)
 s − G EQ log 2 (M )Eb(elec) 
The exact closed form expression for the BER performance
× exp −  ds , of M -PAM in AWGN has been presented in [22] as a summa-
N0 tion of M terms. A tight approximation for BER bellow 10−2
can be obtained when only considering the error contributed
where Q(·) is the complementary cumulative distribution by the closest symbols in the constellation as follows [4]:
function (CCDF) of a standard normal distribution with zero
Ns ( √ )
mean and unity variance. The BER can be obtained as follows: BER = Q ds log2 (M )Γb(elec) . (16)
GGC log2 (M )
M log2 (M )
BER = SER . (10) In M -PAM, an intended symbol has an average number of
2(M − 1)
Ns = 2(M −1)/M neighboring symbols. The gain introduced
by Gray coding of the bits on the symbols is denoted by
GGC = 1. The distance between an intended√symbol and the
closest interfering symbol is given by ds = 6/(M 2 − 1) .
 5

Fig. 2: Block diagram of multi-carrier transmission in OWC using OFDM.

C. M -QAM O-OFDM log2 (M )GT GB /2 bits/s/Hz, where GT = N/(N +NCP ) is the

utilization factor for the information carrying time. The train
The block diagram for multi-carrier O-OFDM transmission of OFDM symbols with CPs, sl,k , follows a close to Gaussian
is presented in Fig. 2. The two O-OFDM realizations known as distribution for IFFT/FFT sizes greater than 64 [21]. In order
DCO-OFDM [9] and ACO-OFDM [10] are studied. In general, to fit the signal within the optical power constraints of the
N subcarriers form the l-th OFDM frame, fl,m , corresponding transmitter, the train of OFDM symbols is scaled and clipped
to the l-th OFDM symbol, where m, m = 0, 1, ..., N − 1, is at normalized bottom and top clipping levels of λbottom and
the subcarrier index. Each subcarrier occupies a bandwidth λtop relative to a standard
( normal distribution
) [13]. In DCO-
of 1/N T in a total OFDM frame double-sided bandwidth OFDM, λbottom = PTx,min (( − βDC /σ, whereas ) in ACO-
)
of B = 1/T . The two O-OFDM systems utilize a different OFDM, λbottom = max PTx,min − βDC /σ, 0 . Here, σ is
portion of the available bandwidth, and the bandwidth uti- the target standard deviation of the( non-clipped time) domain
lization factor is denoted by GB , where GB = (N − 2)/N signal. In both schemes, λtop = PTx,max − βDC /σ. The
in DCO-OFDM and GB = 0.5 in ACO-OFDM. In order to clipping levels in DCO-OFDM can be negative and/or positive,
ensure a real-valued time domain signal, both schemes have whereas in ACO-OFDM, these are strictly non-negative. For
the Hermitian symmetry imposed on the OFDM frame, and reasons of plausibility, λtop > λbottom . Next, the train of
the subcarriers with indices m = {0, N/2} are set to zero. symbols with CPs is subjected to a parallel-to-serial (P/S)
In DCO-OFDM, (N − 2)/2 subcarriers in the first half of conversion, and it is passed through the D/A converter. Here,
the frame carry the information. In ACO-OFDM, only the a pulse shaping filter is applied to obtain the continuous-time
odd subcarriers are enabled, while every even subcarrier is signal. As a next step in the signal shaping framework to fit the
set to zero. Both schemes can utilize bit and power loading front-end optical power constraints, the signal is DC biased by
of the frequency domain subcarriers, in order to optimally βDC . Therefore, the transmitted signal vector, x, with a length
adapt the signal to the channel conditions. For a desired bit of Zx = L(N + NCP ) can be expressed as follows:
rate, the Levin-Campello algorithm [7, 8] can be applied, in
order to maximize the received power margin, or equiva- xl,k = CLIP [αsl,k ] + βDC , (17)
lently, in order to minimize the required electrical SNR. The
where v
optimum solution achieved by the algorithm yields the bm u N −1
bits which modulate the complex-valued information carrying α = σu
u N −1 . (18)
u∑
frequency domain subcarrier from fl,m in an M -QAM fashion. t |f |2
l,m
In addition, the algorithm provides subcarrier power scaling m=0
factors, wm , which ensure an equal maximized received power
margin for every active subcarrier. Without loss of generality, Before the scaling clock, the average electrical power of the
only integer average bit rates, i.e. E [bm ] = log2 (M ), are GB N QAM symbols on the enabled subcarriers amounts to
considered in this study. In both systems, the unitary inverse Ps(elec) = 1. In order to maintain the signal variance of
fast Fourier transform (IFFT) and fast Fourier transform (FFT) σ 2 , the power of the enabled subcarriers is scaled through
are utilized as multiplexing and demultiplexing techniques α to Ps(elec) /GB , where Ps(elec) = σ 2 . Thus, the aver-
at the transmitter and the receiver, respectively [6]. The l-th age bit energy can be expressed as follows: Eb(elec) =
OFDM symbol in the train of L symbols, sl , is obtained by σ 2 /(log2 (M )GT GB B). The non-linear clipping distortion
the IFFT of the l-th OFDM frame in the train of L frames, represented by the CLIP [·] operator can be translated by
fl . Next, NCP samples from the end of each OFDM symbol means of the Bussgang theorem [23] and the central limit
are appended at the beginning of the symbol, creating the CP theorem [24] into a gain factor, K, representing the attenuation
extension, in order to remove the ISI and ICI [6, 16]. Here, of the information carrying subcarriers plus a zero-mean
2
the time domain sample index within the l-th OFDM symbol complex Gaussian noise component with a variance of σclip .
with CP, sl,k , is denoted by k, k = 0, 1, ..., N + NCP − 1. As In DCO-OFDM and ACO-OFDM, K is given as follows [13]:
a result, the time domain OFDM symbol with CP occupies a K = Q(λbottom ) − Q(λtop ) . (19)
double-sided bandwidth of B = 1/T , and it has a duration
of (N + NCP )T . Because of the Hermitian symmetry, the
resulting spectral efficiency of M -QAM O-OFDM amounts to
6

The variance of the clipping noise in DCO-OFDM and Ns GGC ds

ACO-OFDM, respectively, can be expressed as follows [13]: BPSK 1 1 2
( ( √
2
σclip = Ps(elec) K − K 2 − ϕ(λbottom ) − ϕ(λtop ) M =2 2i
4− √
4 3
)2 1
+(1 − Q(λbottom ))λbottom + Q(λtop )λtop i = 1, 2, ... M M −1
(20) √
+(1 − Q(λbottom ))λ2bottom + Q(λtop 2
) )λtop
4 2
M =8 3 √
+ϕ(λbottom )λbottom − ϕ(λtop )λtop , 5 3+ 3
( ( ) 13 6 1
2
σclip = Ps(elec) K λ2bottom + 1 − 2K 2 M = 32 √
4 7 10
( ) √
−λbottom ϕ(λbottom ) − ϕ(λtop ) M = 22i+1 4− √ 6 6M 96
( ) (21) √
−ϕ(λtop ) λtop − λbottom i = 3, 4, ... 2M 6M + 3 2M + 2 31M − 32
( )2 )
+Q(λtop ) λtop − λbottom , TABLE I: Parameters in (16) for M -QAM.

where ϕ(·) stands for the PDF of a standard normal distribu-

tion. In addition to the distortion of the information carrying Given: γb(elec) , M , PTx,min , PTx,max and PTx,avg
subcarriers, time domain signal clipping modifies the average
optical power of the transmitted signal as follows: Find: argmin BER(σ, βDC ) ≥ 0
( σ≥0
E [xl ] = σ λtop Q(λtop ) − λbottom Q(λbottom ) βDC ≥ 0
) (22)
+ϕ(λbottom ) − ϕ(λtop ) + Pbottom ≤ PTx,avg . Constraints: E [xl ] ≤ PTx,avg
In DCO-OFDM,( Pbottom = PTx,min ) , while in ACO-OFDM, λtop > λbottom in DCO-OFDM
Pbottom = max PTx,min , βDC because of the default zero- λtop > λbottom ≥ 0 in ACO-OFDM
level clipping. Thus, for a given set of front-end optical power
constraints, one can obtain the signal scaling factor, α, for a TABLE II: Minimization of BER over σ and βDC for given
target signal variance, σ 2 , and the required DC bias, βDC , target γb(elec) , M , PTx,min , PTx,max and PTx,avg .
from (18) and (22). The optimum choice of these design
parameters is elaborated below.
and ACO-OFDM, respectively, as follows [13]:
Since the signal is clipped, the resulting average optical
power of the signal, E [xl ], differs from the undistorted op- σ2
GDC = 2 , (26)
tical power of the OFDM symbol, Ps(opt) . In DCO-OFDM, σ2
+ βDC
Ps(opt) √
√ = βDC , whereas in ACO-OFDM, Ps(opt) = (βDC + 2πσ 2
σ/ 2π). The O/E conversion is obtained in DCO-OFDM and GDC =√ √ 2
. (27)
ACO-OFDM, respectively, as follows: 2πσ 2 + 4σβDC + 2 2πβDC
σ 2 + βDC
2 The exact closed form expression for the BER performance
2
Ps(elec) = 2 Ps(opt) , (23) of square and cross M -QAM constellations in AWGN has
βDC
√ been presented as a summation of M terms in [22] and [25],
2
2πβDC + 2σβDC 2π + πσ 2 2 respectively. However, the same tight approximation from (16)
Ps(elec) = √ Ps(opt) . (24)
2 + 2σβ
2πβDC DC 2π + σ
2 can be applied, and the respective parameters are given in
TABLE I [4, 26]. Thus, the BER on the intended subcarrier,
The received signal, yl,k , is passed through a matched filter,
BERm , can be obtained by inserting (25) into (16), consider-
and at the A/D converter it is sampled at a frequency of 1/T or
ing the parameters from TABLE I for the number of loaded
higher [6, 16]. Next, the CP extension of every OFDM symbol
bits. As a result, the link BER can be obtained as the average
is removed, and after serial-to-parallel (S/P) conversion the
of the BER of all enabled subcarriers: BER = E [BERm ].
signal is passed through an FFT block back to the frequency
The choice of the biasing parameters, such as the signal
domain. A single-tap equalizer and a hard-decision decoder
variance, σ 2 , and the DC bias, βDC , which minimize the link
are employed to obtain the received bits. Thus, the effective
BER for a target γb(elec) can be formulated as an optimization
electrical SNR per bit on an enabled subcarrier in O-OFDM,
problem. Additional input parameters for the optimization
Γmb(elec) , is given for linear ZF FFE as follows: are the front-end optical power constraints, PTx,min , PTx,max
wm K 2 and PTx,avg , and the desired average bit rate, equivalent to
Γm
b(elec) = −1 , (25) a QAM modulation order, M . This optimization problem is
2 GB γb(elec)
bm GB σclip
+ summarized in TABLE II, and its solution can be used to
Ps(elec) GT GDC |H(f )|2 iteratively solve the dual problem, i.e. the minimization of the
where H(f ) is the channel frequency response on the intended γb(elec) for a target BER.
subcarrier. The factor GDC can be expressed in DCO-OFDM The optimization problem from TABLE II has a trivial
solution when the DC bias power is not included in the
 7

calculation of the effective electrical SNR per bit, Γb(elec) , i.e.

when GDC = 1. From (19) it follows that that K 2 decreases
when the signal is more severely clipped. In addition, because
of the fact that the clipping noise variance is non-negative,
Γb(elec) is maximized and BER is minimized when the signal
clipping is minimized. For instance, such a clipping scenario
in DCO-OFDM is represented by λbottom = −4 and λtop = 4.
It is similar to the one used in [15], in order to minimize the
clipping distortion. The equivalent scenario for ACO-OFDM
is λbottom = 0 and λtop = 4. These setups enable modulation
orders as high as M = 1024 with a deviation from the true
minimum required γb(elec) of only 0.1 dB at BER of 10−3 .
However, the optimization problem has a non-trivial solu-
tion when the DC bias power is included in the calculation
of the effective electrical SNR per bit, Γb(elec) , i.e. when Fig. 3: Minimum BER in DCO-OFDM as a function of σ and
GDC < 1. The analytical approach to solve the minimization βDC for a fixed γb(elec) = 16 dB, 4-QAM with linear ZF FFE,
problem leads to a system of non-linear transcendental equa- h(t) = δ(t), PTx,min = 5 mW and PTx,max = 50 mW. DC
tions which does not have a closed-form solution. Therefore, bias power is included in the electrical SNR.
a numerical optimization procedure is required, and the min-
imization can be carried out through a computer simulation
for a particular choice of front-end optical power constraints.
In general, the formal proof of convexity of the objective
function from TABLE II over the constrained function domain
is equally intractable as the analytical minimization approach.
However, the convexity can be illustrated by means of a
computer simulation. A practical linear dynamic range of
a Vishay TSHG8200 LED between PTx,min = 5 mW and
PTx,max = 50 mW at room temperature is assumed at the
transmitter [11]. It can be inferred from [19] that this LED
is eye-safe, even if the average optical power level is set to
the maximum of the dynamic range. Therefore, the average
optical power constraint is relaxed, in order to obtain the
best BER system performance for the given dynamic range
of the front-end. The objective function from TABLE II is Fig. 4: Minimum BER in ACO-OFDM as a function of σ and
illustrated in Fig. 3 and Fig. 4 for DCO-OFDM and ACO- βDC for a fixed γb(elec) = 12 dB, 4-QAM with linear ZF FFE,
OFDM, respectively, in the case of a flat fading channel with h(t) = δ(t), PTx,min = 5 mW and PTx,max = 50 mW. DC
impulse response of h(t) = δ(t), where δ(·) is the Dirac bias power is included in the electrical SNR.
delta function. It is shown that the objective function for a
flat fading channel has a unique optimum convex region. In III. S INGLE - CARRIER vs. MULTI - CARRIER TRANSMISSION
OFDM systems, the dispersive channel is represented by a
superposition of orthogonal flat fading channels. Therefore, The performance of M -PPM and M -PAM vs. M -QAM
the objective average BER function can be obtained as the optical OFDM is assessed in terms of electrical SNR re-
average of the BER functions for each flat fading channel quirement, γb(elec) , to achieve a target BER of 10−3 and the
which are shown to be convex. Since the expectation oper- corresponding spectral efficiency. In the first set of results,
ator is a non-negative weighted summation, it preserves the the DC bias power is not counted towards the signal power,
convexity [27]. Therefore, the objective BER function in the and a flat fading channel without dispersion, i.e. h(t) = δ(t),
dispersive channel remains convex. is assumed. The following modulation orders are chosen:
The details of the optimum biasing parameters for the M = {2, 4, 16, 64, 256, 1024}. Here, single-carrier BPSK is
abovementioned dynamic range of the transmitter front-end identical to 2-PAM. The above mentioned eye-safe linear
and QAM modulation orders, M = {4, 16, 64, 256, 1024}, dynamic range of the transmitter LED between PTx,min = 5
with linear ZF FFE in a flat fading channel are presented mW and PTx,max = 50 mW is assumed for the comparison
in TABLE III. Considering the electrical power invested in of the OWC systems. The transmitted signal spans the entire
the DC bias, it is shown that DCO-OFDM requires a non- dynamic range of optical power, and no constraint is imposed
symmetric clipping setup with the DC bias placed below the on the radiated average optical power, in order to obtain the
middle of the dynamic range, in order to minimize the required best BER system performance for the given dynamic range.
γb(elec) for a target BER. The optimum performance of ACO- In single-carrier transmission, no signal clipping is assumed.
OFDM is obtained when the downside clipping is kept at As a result, the average optical power level is set in M -PPM
minimum by setting the DC bias close to PTx,min . to E [xl ] = 50/M mW, and in M -PAM to E [xl ] = 27.5 mW.
8

ACO-OFDM DCO-OFDM
σ [mW] βDC [mW] λbottom λtop σ [mW] βDC [mW] λbottom λtop
4-QAM 19.25 4.12 0.05 2.38 11.58 22 −1.47 2.42
16-QAM 16.4 4.78 0.01 2.76 9.04 24.07 −2.11 2.87
64-QAM 14.39 4.95 0 3.13 7.66 24.95 −2.6 3.27
256-QAM 12.92 4.99 0 3.48 6.75 25.47 −3.03 3.63
1024-QAM 11.81 5 0 3.81 6.1 25.81 −3.41 3.97

TABLE III: Optimum biasing parameters, σ and βDC , and optimum normalized clipping levels, λbottom and λtop , in ACO-
OFDM and DCO-OFDM with M -QAM and linear ZF FFE for PTx,min = 5 mw, PTx,max = 50 mW and a 10−3 BER in a
flat fading channel with impulse response h(t) = δ(t). DC bias power is included in the electrical SNR.

Fig. 6: Equalizer gain for signal bandwidth exceeding the

Fig. 5: Spectral efficiency vs. electrical SNR requirement for a channel coherence bandwidth.
10−3 BER of the OWC schemes in a flat fading channel with
impulse response h(t) = δ(t) and neglected DC bias power.
process, it directly translates into an SNR penalty, gh(elec) = 1
is assumed. The equalizer gain of multi-carrier transmission
In multi-carrier transmission, a large number of subcarriers, with bit and power loading and single-tap ZF FFE is compared
e.g. 2048, is chosen. Minimum signal clipping is assumed with the equalizer gain of single-carrier transmission with
in O-OFDM, i.e. λbottom = −4 and λtop = 4 in DCO- multi-tap ZF FFE and ZF DFE as the signal bandwidth grows
OFDM , and λbottom = 0 and λtop = 4 in ACO-OFDM. larger than the channel coherence bandwidth. The result is
In both systems, the average optical power level, E [xl ], can presented in Fig. 6. It is shown that multi-carrier transmission
be obtained from (22). The resulting spectral efficiency vs. incurs a lower SNR penalty in the equalization process.
electrical SNR requirement plot of the transmission schemes When the DC bias power is added to the signal power, the
for OWC is presented in Fig. 5. It is shown that PPM is the systems incur an SNR penalty, becasue the DC bias reduces
only system which can operate at very low SNR in the range the useful AC signal power for a fixed total signal power.
of 4.1 − 6.8 dB. For a given higher SNR, DCO-OFDM and Based on the different signal statistics, the compared systems
PAM demonstrate an equal highest spectral efficiency. incur a different SNR penalty due to the DC bias. The DC
However, in a practical non-flat channel with dispersion [3], bias gain is presented in Fig. 7 as the signal bandwidth
the signal bandwidth becomes larger than the channel coher- exceeds the channel coherence bandwidth. For the considered
ence bandwidth at high data rates. Therefore, the equalization dynamic range of the transmitter between PTx,min = 5 mw
process incurs an SNR penalty. In such a scenario, single- and PTx,max = 50 mW, the optimum signal clipping reduces
carrier transmission suffers a severe ISI. In multi-carrier the SNR penalty by up to 6.5 dB for DCO-OFDM and up
transmission, a CP is employed which completely eliminates to 1.4 dB for ACO-OFDM as compared to minimum signal
ISI and ICI, and it has a negligible impact on the spectral clipping. In addition, bit and power loading in combination
efficiency and electrical SNR requirement [5]. It transforms with optimum signal clipping allow the DC bias gain to
the channel into a flat fading channel over the subcarrier saturate above the DC bias gain in the minimum clipping case.
bandwidth, and therefore single-tap equalization with bit and Nevertheless, because of the close to Gaussian distribution of
power loading [7, 8] can be performed, in order to minimize the signals, DCO-OFDM and ACO-OFDM still incur a larger
the channel effect. In addition, since the electrical path gain SNR penalty as compared to PAM and PPM, respectively,
coefficient, gh(elec) , is merely a factor in the equalization which have distributions with finite support. Therefore, in
 9

Fig. 7: DC bias gain for signal bandwidth exceeding the chan-

nel coherence bandwidth. A dynamic range of the transmitter
between PTx,min = 5 mw and PTx,max = 50 mW is assumed.
Fig. 9: Spectral efficiency for signal bandwidth exceeding the
channel coherence bandwidth. The target BER is 10−3 with
an available electrical SNR per bit of 25 dB.

bandwidth. When the DC power is not counted towards

the electrical signal power, DCO-OFDM and ACO-OFDM
show a superior spectral efficiency in the dispersive optical
wireless channel as compared to PAM and PPM, respectively.
When the DC power is included in the calculation of the
electrical SNR, ACO-OFDM still outperforms PPM. DCO-
OFDM outperforms PAM with FFE, and it approaches the
performance of PAM with DFE. However, it has to be noted
that the analysis of PAM with DFE represents an upper bound
for the performance which is achieved when an infinite number
of channel taps are considered in the equalizer. In a practical
indoor optical wireless channel, where the impulse response
Fig. 8: Required electrical SNR per bit for signal bandwidth only changes slowly, the channel taps and the required bit
exceeding the channel coherence bandwidth. The target BER and power loading parameters with optimum signal shaping
is 10−3 for a dynamic range of the transmitter between can be pre-computed and stored in look-up tables in memory.
PTx,min = 5 mw and PTx,max = 50 mW. Therefore, the computational complexity at the receiver comes
from the convolution operation of the DFE equalizer in single-
order to obtain the electrical SNR requirement when the DC carrier transmission and the FFT operation in multi-carrier
bias power is counted towards the signal power in a non-flat transmission. It has been shown in [4] that the most efficient
dispersive channel, the DC bias gain and the equalizer gain DFE implementation requires one FFT and one IFFT operation
need to be subtracted from the electrical SNR requirement for N channel taps. Therefore, for a fixed FFT size, O-OFDM
from Fig. 5. The result is presented in Fig. 8. It is shown is expected to require half of the computational complexity of
that optimum signal clipping allows O-OFDM to close the single-carrier transmission with DFE.
gap to single-carrier transmission down to 2 dB in a flat
fading channel when the DC bias power is included in the IV. C ONCLUSION
calculation of the SNR requirement. However, when the signal In this paper, single-carrier transmission, e.g. M -PPM and
bandwidth exceeds the channel coherence bandwidth in a M -PAM, and multi-carrier transmission, e.g. M -QAM DCO-
dispersive channel, ACO-OFDM shows a lower electrical SNR OFDM and ACO-OFDM, are studied for OWC. A signal
requirement as compared to PPM with both FFE and DFE. shaping framework is presented which through scaling and
Equivalently, DCO-OFDM is shown to have a lower SNR DC biasing conditions the transmitted signal within the op-
requirement than PAM with FFE, and it approaches the SNR tical power constraints of the transmitter front-end. The op-
requirement of PAM with DFE. timal signal shaping enables the Gaussian O-OFDM signals
By fixing the electrical SNR requirement, the relative per- to minimize the electrical SNR requirement. The analytical
formance of the systems can be obtained in terms of spectral expressions for the BER performance of the transmission
efficiency. This is illustrated in Fig. 9 for γb(elec) = 25 schemes with equalization of the optical wireless channel
dB as the signal bandwidth exceeds the channel coherence in AWGN are obtained, excluding or including the DC bias
10

power in the calculation of the electrical SNR. This enables [14] R. J. Green, H. Joshi, M. D. Higgins, and M. S. Leeson, “Recent
a novel comparison of system performance in terms of SNR Developments in Indoor Optical Wireless,” IET Communications, vol. 2,
no. 1, pp. 3–10, Jan. 2008.
requirement and spectral efficiency. When the DC bias power [15] J. Armstrong and B. J. C. Schmidt, “Comparison of Asymmetrically
is neglected, DCO-OFDM and ACO-OFDM show a superior Clipped Optical OFDM and DC-Biased Optical OFDM in AWGN,”
spectral efficiency in the dispersive optical wireless channel IEEE Commun. Lett., vol. 12, no. 5, pp. 343–345, May 2008.
[16] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.
as compared to PAM and PPM. DCO-OFDM is expected Cambridge University Press, 2005.
to deliver the highest throughput in applications, where the [17] H. Elgala, R. Mesleh, and H. Haas, “Non-linearity Effects and Predis-
additional DC bias power required to create a non-negative tortion in Optical OFDM Wireless Transmission Using LEDs,” Inder-
science International Journal of Ultra Wideband Communications and
signal can serve a complementary functionality, such as il- Systems (IJUWBCS), vol. 1, no. 2, pp. 143–150, 2009.
lumination in VLC. In IR communication, where the DC [18] T. S. Rappaport, Wireless Communications: Principles and Practice,
power is generally constrained by eye-safety regulations, and 2nd ed. Prentice Hall PTR, 2002.
[19] S. Dimitrov, R. Mesleh, H. Haas, M. Cappitelli, M. Olbert, and E. Bas-
it is included in the calculation of the electrical SNR, the sow, “On the SIR of a Cellular Infrared Optical Wireless System for an
optimum signal clipping enables O-OFDM to reduce the Aircraft,” IEEE Journal on Selected Areas in Communications (IEEE
SNR requirement gap to single-carrier transmission down to JSAC), vol. 27, no. 9, pp. 1623–1638, Dec. 2009.
[20] S. Dimitrov, H. Haas, M. Cappitelli, and M. Olbert, “On the Throughput
2 dB in the flat fading fading channel. However, when the of an OFDM-based Cellular Optical Wireless System for an Aircraft
signal bandwidth exceeds the channel coherence bandwidth, Cabin,” in Proc. of European Conference on Antennas and Propagation
(EuCAP 2011), Rome, Italy, 11–15 Apr. 2011.
DCO-OFDM shows a higher spectral efficiency than PAM
[21] D. Dardari, V. Tralli, and A. Vaccari, “A Theoretical Characterization
with FFE, and it approaches the performance of the more of Nonlinear Distortion Effects in OFDM Systems,” IEEE Transactions
computationally intensive PAM with DFE. on Communications, vol. 48, no. 10, pp. 1755–1764, Oct. 2000.
[22] J. Li, X. Zhang, Q. Gao, Y. Luo, and D. Gu, “Exact BEP Analysis
for Coherent M-arry PAM and QAM over AWGN and Rayleigh Fading
ACKNOWLEDGEMENT Channels,” in Proc. of the IEEE Vehicular Technology Conference (VTC
We gratefully acknowledge EADS UK Ltd. for the support 2008-Spring), Singapore, 11-14 May 2008, pp. 390–394.
[23] J. Bussgang, “Cross Correlation Function of Amplitude-Distorted Gaus-
of this research. Professor Haas acknowledges the Scottish sian Signals,” Research Laboratory for Electronics, Massachusetts Insti-
Funding Council support of his position within the Edinburgh tute of Technology, Cambridge, MA, Technical Report 216, Mar. 1952.
Research Partnership in Engineering and Mathematics between [24] J. Rice, Mathematical Statistics and Data Analysis, 2nd ed. Duxbury
Press, 1995.
the University of Edinburgh and Heriot Watt University. [25] P. K. Vitthaladevuni, M.-S. Alouini, and J. C. Kieffer, “Exact BER
Computation for Cross QAM Constellations,” IEEE Transactions on
R EFERENCES Wireless Communications, vol. 4, no. 6, pp. 3039–3050, Nov. 2005.
[26] J. Smith, “Odd-Bit Quadrature Amplitude-Shift Keying,” IEEE Trans-
[1] F. R. Gfeller and U. Bapst, “Wireless In-House Data Communication Via actions on Communications, vol. 23, no. 3, pp. 385–389, Mar. 1975.
Diffuse Infrared Radiation,” Proceedings of the IEEE, vol. 67, no. 11, [27] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge
pp. 1474–1486, Nov. 1979. University Press, 2004.
[2] Y. Tanaka, T. Komine, S. Haruyama, and M. Nakagawa, “Indoor
Visible Communication Utilizing Plural White LEDs as Lighting,” in
Proceedings of the 12th IEEE International Symposium on Personal, Svilen Dimitrov (S’09) received the BSc degree in electrical engineering and
Indoor and Mobile Radio Communications, vol. 2, San Diego, CA, USA, computer science in 2008, and the MSc degree in communications, systems,
Sep. 30–Oct. 3, 2001, pp. 81–85. and electronics in 2009 from Jacobs University, Bremen, Germany. He wrote
[3] J. M. Kahn and J. R. Barry, “Wireless Infrared Communications,” his BSc thesis (2007-2008) with the department of Pre-Development of Cabin
Proceedings of the IEEE, vol. 85, no. 2, pp. 265–298, Feb. 1997. Electronic Systems of Airbus Germany on a simulation model for reproduction
[4] J. G. Proakis, Digital Communications, 4th ed., ser. McGraw-Hill Series of infrared wireless path loss distribution in an aircraft cabin, using a Monte
in Electrical and Computer Engineering, S. W. Director, Ed. McGraw- Carlo Ray-tracing algorithm. During his MSc study (2008-2009), he extended
Hill Higher Education, December 2000. the work on the characterization of the optical wireless channel with the
[5] H. Elgala, R. Mesleh, and H. Haas, “Practical Considerations for Indoor department of Simulation and Graphical Technologies of EADS Innovation
Wireless Optical System Implementation using OFDM,” in Proc. of the Works Germany. Currently, he is working towards his PhD degree in electrical
IEEE 10th International Conference on Telecommunications (ConTel), engineering at the University of Edinburgh, UK. His main research interests
Zagreb, Croatia, Jun. 8–10 2009. are in the area of computer-aided system design, test and optimization with
[6] J. Armstrong, “OFDM for Optical Communications,” Journal of Light- emphasis on wireless communication systems.
wave Technology, vol. 27, no. 3, pp. 189–204, Feb. 2009.
[7] J. Campello, “Practical Bit Loading for DMT,” in Proc. of IEEE
International Conference on Communications (IEEE ICC 1999), vol. 2, Sinan Sinanovic (S’98-M’07) obtained his Ph.D. in electrical and computer
Vancouver, BC, Canada, 6–10 Jun. 1999, pp. 801–805. engineering from Rice University, Houston, Texas, in 2006. In the same year,
[8] H. E. Levin, “A Complete and Optimal Data Allocation Method for he joined Jacobs University Bremen in Germany as a post doctoral fellow. In
Practical Discrete Multitone Systems,” in Proc. of IEEE Global Telecom- 2007, he joined the University of Edinburgh in the UK where he currently
munications Conference (IEEE GLOBECOM 2001), vol. 1, San Antonio, works as a research fellow in the Institute for Digital Communications
TX , USA, 25–29 Nov. 2001, pp. 369–374. (IDCOM). While working with Halliburton Energy Services, he developed
[9] J. B. Carruthers and J. M. Kahn, “Multiple-subcarrier Modulation acoustic telemetry receiver which was patented. He also worked for Texas
for Nondirected Wireless Infrared Communication,” IEEE Journal on Instruments. He is a member of the Tau Beta Pi engineering honor society
Selected Areas in Communications, vol. 14, no. 3, pp. 538–546, Apr. and a member of Eta Kappa Nu electrical engineering honor society. He won
1996. an honorable mention at the International Math Olympiad in 1994.
[10] J. Armstrong and A. Lowery, “Power Efficient Optical OFDM,” Elec-
tronics Letters, vol. 42, no. 6, pp. 370–372, Mar. 16, 2006.
[11] Vishay Semiconductors, “Datasheet: TSHG8200 High Speed Infrared Professor Harald Haas (S’98-A’00-M’03) holds the Chair of Mobile Com-
Emitting Diode, 830 nm, GaAlAs Double Hetero,” Retrieved Jul. 26, munications in the Institute for Digital Communications (IDCOM) at the
2011 from http://www.vishay.com/docs/84755/tshg8200.pdf, Jul. 2008. University of Edinburgh, and he currently is the CTO of a university spin-out
[12] BS EN 62471:2008, Photobiological Safety of Lamps and Lamp Systems, company VLC Ltd. His main research interests in interference coordination
BSI British Standards Std., Sep. 2008. in wireless networks, spatial modulation and optical wireless communication.
[13] S. Dimitrov, S. Sinanovic, and H. Haas, “Clipping Noise in OFDM- Prof. Haas holds more than 15 patents. He has published more than 50
based Optical Wireless Communication Systems,” IEEE Transactions journal papers including a Science Article and more than 150 peer-reviewed
on Communications (IEEE TCOM), 2011, to appear. conference papers. Nine of his papers are invited papers. Prof. Haas has
 11

co-authored a book entitled ”Next Generation Mobile Access Technologies:

Implementing TDD” with Cambridge University Press. Since 2007 Prof. Haas
has been a Regular High Level Visiting Scientist supported by the Chinese
”111 program” at Beijing University of Posts and Telecommunications
(BUPT). He was an invited speaker at the TED Global conference 2011,
and his work on optical wireless communication was listed among the ”50
best inventions in 2011” in the Time Magazine.

View publication stats