1734 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 11 ... · 1734 JOURNAL OF LIGHTWAVE TECHNOLOGY,...

1734 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 11, JUNE 1, 2013

Optical Wireless Transmitter Employing DiscretePower Level Stepping

Thilo Fath, Christoph Heller, and Harald Haas, Member, IEEE

Abstract—A major shortcoming of light-emitting diodes (LEDs)is their highly non-linear optical-power-versus-current character-istic. This non-linearity largely restricts the dynamic range andthe transmission power of common optical wireless transmitters.This restriction degrades the performance of optical wirelesscommunication (OWC) systems. In this paper, a novel transmitterconcept for OWC is proposed which employs discrete powerlevel stepping. The transmitter consists of several on-off-switch-able emitter groups. These groups are individually controlledand emit fixed specific optical intensities in parallel. As opticalintensities constructively add up, the total emitted intensity isgenerated by the sum of the emitted intensities of all activatedemitter groups. Therefore, the proposed transmitter solutioncan generate several discrete optical intensity levels which canbe used for optical wireless signal transmission. The transmitterdesign allows the utilization of the full dynamic range of LEDsor laser diodes by avoiding non-linearity issues. Moreover, costsand complexity of the optical front-end are significantly reducedas neither a digital-to-analogue converter (DAC) nor high-speedcurrent controllers are required. This simple design also providesimproved power efficiency. Transmission experiments prove thefunctionality of the implemented optical transmitter. It is shownthat the practical performance of the transmitter closely matchesthe expected performance determined by computer simulations.Moreover, the implemented optical transmitter is compared to anelectrical transmission which provides ideal linearity character-istics, and therefore corresponds to an ideal conventional opticaltransceiver.

Index Terms—Intensity modulation, non-linearity, optical wire-less communications, power efficiency, transmitter design.

I. INTRODUCTION

C URRENT and future wireless data applications, such ashigh-speed internet access or high-quality audio/video

streaming, require wireless data links that offer large through-puts. Commonly, wireless links are based on signal transmissionin the radio frequency (RF) bands. The ever increasing demandfor higher data rates and the enlarging number of devicesmake the RF spectrum a precious commodity [1]. Besides, RF

Manuscript received January 06, 2013; revised March 21, 2013; acceptedMarch 25, 2013. Date of publication April 12, 2013; date of current versionApril 26, 2013.T. Fath is with EADS Deutschland GmbH, Innovation Works, 81663 Mu-

nich, Germany. He is also with the University of Edinburgh, Institute for DigitalCommunications, Edinburgh EH9 3JL, U.K. (e-mail: [email protected]).C. Heller is with EADS Deutschland GmbH, Innovation Works, 81663 Mu-

nich, Germany (e-mail: [email protected]).H. Haas is with the University of Edinburgh, Institute for Digital Communi-

cations, Edinburgh EH9 3JL, U.K. (e-mail: [email protected]).Color versions of one or more of the figures in this paper are available online

at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/JLT.2013.2257984

devices cannot be used in environments which have stringentelectromagnetic compatibility (EMC) restrictions like hospi-tals, fabrication plants or aeroplanes. As RF communicationstransmits data by modulating electromagnetic fields in the rangeof several MHz and GHz, it requires complex and elaboratedhigh-frequency circuit designs. Moreover, these designs are notpower-efficient as only a small amount of the consumed poweris actually radiated.Given this situation, ongoing research activities are focusing

on alternative means and media for wireless data transmission[2]. Optical wireless communications (OWC) has the potentialto become a viable complement to RF signal transmission anda remedy for the shortage of the RF spectrum [1], [3]. Com-pared to RF transmission, OWC has better EMC and is not sub-ject to frequency regulations. OWC is based on modulating theoptical power (intensity) emitted by light sources. For instance,the light bulbs (in fact high power light-emitting diodes (LEDs))installed at the ceiling of a room can simultaneously provide il-lumination and data transmission.Commonly, OWC employs pulsed digital modulation tech-

niques to convey information [4]. For instance, the widely usedon-off-keying (OOK) technique relies on purely switching thelight source on and off. However, since OOK is a binary modu-lation scheme, it provides low spectral efficiency. As off-the-shelf LEDs achieve only a limited modulation bandwidth ofabout 30–50 MHz in the case of infra-red (IR) light and evenless for visible light, OOK can only provide limited data rates.Consequently, more advanced modulation schemes are requiredto enable higher data rates. Higher order modulation schemessuch as -level pulse amplitude modulation (PAM) or -levelpulse position modulation (PPM) provide larger throughputs.However, all these pulsed modulation techniques suffer frommultipath effects which mostly occur in diffuse and scatteredtransmission scenarios [5], [6]. Multipath effects cause inter-symbol interference (ISI), and therefore decrease the systemperformance to a large extent by constraining both bandwidthand data rate.Orthogonal frequency division multiplexing (OFDM) is

widely used in modern RF standards as well as in OWC [7].This is due to the fact that OFDM is a bandwidth efficient trans-mission technique which can cope with frequency-selectivefading [8]–[10]. In contrast to pulsed modulation techniques,OFDM provides high data rates even in severe multipathscenarios. In order to mitigate multipath effects and to avoidISI, OFDM uses a guard interval, the so-called cyclic prefix,which is placed between the transmitted OFDM symbols.Moreover, the actual OFDM symbol duration is typically muchlonger than the delay spread caused by multipath propagation.OFDM conveys digital data on multiple orthogonal sub-carrierfrequencies. Hence, a wideband channel is subdivided into

0733-8724/$31.00 © 2013 IEEE

FATH et al.: OPTICAL WIRELESS TRANSMITTER EMPLOYING DISCRETE POWER LEVEL STEPPING 1735

several narrowband sub-channels. These orthogonal sub-car-rier channels are used to transmit independent data streamsin parallel in the frequency domain. The frequency divisionmultiplexed channels are summed up and transformed into thetime domain by using an inverse fast Fourier transformation(IFFT). The utilization of narrowband sub-channels enables alow-complex channel equalization: each sub-channel can beregarded as a non-frequency-selective channel which is indi-vidually equalised in the frequency domain using a single-tapzero forcing (ZF) equaliser. These properties make OFDM anideal candidate for diffuse and scattered OWC transmissionscenarios [3], [7], [11], [12]. Moreover, due to the use of fastFourier transformation (FFT) and IFFT, OFDM enables simpleand efficient signal processing implementations. For instance,compared to -PAM, OFDM requires approximately threetimes less computational complexity [13].However, a common drawback of OFDM is the fact that it

requires sophisticated transceiver designs (RF or optical) whichmust have good linearity characteristics and large dynamicranges. Non-linear distortion effects, e.g. caused by amplifiers,largely decrease the system performance [14]–[17]. Comparedto RF communications, OWC does not require high-frequencycircuit designs because the signals are transmitted in the base-band representation and directly modulate the intensity of theemitting LEDs. However, LEDs are highly non-linear due totheir optical-power-versus-current characteristic. This non-lin-earity largely restricts the dynamic range and the transmissionpower of common optical transmitters as well as the overallsystem performance [18]–[21]. Moreover, the non-linearity ofLEDs requires complex pre-distortion and intricate equaliza-tion techniques [22], [23]. Consequently, conventional opticaltransmitter front-ends cannot provide the required linearity in apower- and cost-efficient way.In this paper, an OWC transmitter concept is proposed which

addresses the shortcomings of conventional optical trans-mitters. In particular the concept addresses: non-linearities,restricted dynamic range and complex system designs. Theproposed transmitter employs discrete power level steppingwhich enables a simple and power-efficient front-end design.Sophisticated pre-distortion or intricate equalization techniquesto mitigate transmitter induced non-linearities are not required.As a result, a low-complex single-tap equalization is sufficient.The concept can be used to implement transmitters for visibleor IR light transmission as well as for line-of-sight (LOS) anddiffuse propagation scenarios.The remainder of this paper is organized as follows:

In Section II, the considered system model is introduced.Section III reviews conventional transmitter front-end designsand discusses implementation issues. In Section IV, the effectsof quantization on the bit error ratio (BER) performance areshown. Based on this analysis, appropriate bit resolutions aredetermined which provide sufficient precision. In Section V,the proposed transmitter concept is introduced which employsdiscrete power level stepping. Moreover, the focus is set onseveral realization aspects and the implemented transmitter isdescribed. In Section VI, the results of several transmissionexperiments are presented. The transmitter performance isevaluated in terms of error vector magnitude (EVM) and BER.Measurements prove the functionality of the implementedtransmitter. Finally, Section VII concludes the paper.

Fig. 1. Optical wireless transmission system.

II. SYSTEM MODEL

Fig. 1 illustrates an OWC system which consists of an opticaltransmitter (TX) and receiver (RX). The optical transmittercomprises a modulator that transforms the incoming datastream into a signal that is to be transmitted by the light source.The output of the modulator is a digital signal. For higherorder modulation techniques, such as -PAM or OFDM,the modulator output is a bit vector representing an analoguesignal value. The TX front-end transforms the bit vector into ananalogue optical waveform by adequately controlling the lightsource. As a result, the front-end performs intensity modulation(IM) by changing the intensity of the emitted light accordingto the signal (data) to be sent. At the receiver side, a lightdetector retransforms the impinging optical waveform into anelectrical signal. Commonly, this component is a photo-diodewhich transforms optical power into current by direct detec-tion (DD). In this work, a silicon positive intrinsic negative(PIN) photo-diode [24] is employed. Additionally, the RXfront-end contains an anti-aliasing filter, i.e. a low-pass filter torestrict the signal bandwidth in order to satisfy the samplingtheorem, and an amplifier. After the received signal has beenfiltered and amplified, it is converted into a digital signal by ananalog-to-digital converter (ADC). The demodulator processesthe digital signal and reconstructs the transmitted data. Theremainder of this paper focuses on the TX front-end and thegeneration of the optical waveform.Typically, quadrature amplitude modulation (QAM) is con-

sidered as digital modulation scheme for OFDM transmission.However, as IM/DD based OWC requires real-valued non-neg-ative time domain signals, QAM cannot be directly applied tooptical OFDM. Therefore, the input vector to the transmitterIFFT is constrained to have Hermitian symmetry. The inputvector provides Hermitian symmetry if its elements fulfil

, where represents com-plex conjugation and is the IFFT size. The complex-valueddata symbols are given by , with . Thezero sub-carrier is not modulated as it results in a direct-current(DC) bias. Hermitian symmetry of the input vector creates areal-valued time domain waveform by cancelling the imaginarycomponents of the IFFT output. This real-valued signal can besent by incoherent light sources.However, the time domain signal is still bipolar. In order

to provide unipolar signals which can modulate the intensityof the optical carrier, further processing is required. To thisend, the multi-carrier techniques direct-current-biased op-tical OFDM (DCO-OFDM), asymmetrically clipped opticalOFDM (ACO-OFDM) and pulse-amplitude-modulated dis-crete multitone modulation (PAM-DMT) have been proposed.DCO-OFDM adds a DC offset to the waveform to be sentin order to provide non-negative time domain signals [12].


The DC offset can be added by adjusting the bias point of thetransmitter LED. The DC bias has to be chosen appropriatelyto provide non-negativity of the time domain signals, whilekeeping upper and lower clipping effects to a minimum [25].In ACO-OFDM transmission, only the odd-numbered sub-car-riers are modulated, whereas the even-numbered sub-carriersare set to zero. This composition of the input vector to thetransmitter IFFT generates a real-valued bipolar time domainsignal. Armstrong and Lowery have shown that a hard-clip-ping can be applied to the entire negative signal amplitudeswithout affecting the data conveyed in the OFDM signal [26].This clipping converts the bipolar signal into an unipolarwaveform as all negative amplitude values are set to zero.However, as only the odd-numbered sub-carriers convey datasymbols, ACO-OFDM provides half the spectral efficiency ofDCO-OFDM. In contrast to DCO-OFDM and ACO-OFDMwhich employ complex-valued QAM symbols, PAM-DMTuses real-valued PAM symbols to modulate the sub-carriers.As proposed by Lee et al., the real parts of the input vectorare set to zero and only the imaginary parts are pulse am-

plitude modulated [27]. Similar to ACO-OFDM, the negativeamplitude values of the time domain signal can be hard-clippedwithout affecting the data symbols. Consequently, only thepositive amplitude values are used to modulate the intensityof the optical carrier. As shown in [13], [27], PAM-DMTand ACO-OFDM provide similar performance for all spectralefficiencies.Without loss of generality, DCO-OFDM is considered

as transmission technique in the following. However, thetransmitter concept proposed in this paper is not restricted toDCO-OFDM but can also be used with any other multi-carriertransmission scheme, such as ACO-OFDM, PAM-DMT orDMT employing adaptive bit-loading and power-loading [28],for instance. Moreover, single-carrier transmission schemeslike OOK, Manchester coding [29], PAM, -ary ampli-tude-shift-keying (ASK) [30] or PPM can be used as well. Thecomparison of different modulation techniques for OWC isbeyond the scope of this paper. The reader is kindly referred toe.g. [13], [31], [32] for such comparison.

III. CONVENTIONAL TRANSMITTER FRONT-END DESIGNS

A transmitter front-end for OWC transforms the electricalsignal generated by the modulator into an optical signal whichhas sufficient power to enable its detection at the receiver. Theblock diagram of such a transmitter front-end is displayed inFig. 2. The modulator output of common transmission tech-niques, such as OFDM, is a bit vector of size . This bit vectoris a binary digital representation of the signal to be transmitted.Conventional optical transmitter front-ends transform this bitvector into an analogue voltage signal by using a digital-to-ana-logue converter (DAC). The generated voltage signal is con-verted into a current signal using a transconductance amplifier(TCA). The TCA finally drives the light source. Commonly, theoptical light source is a LED, either in the visible or IR lightrange. Only LEDs are fast enough to transform a modulatedcurrent signal into an optical signal of several MHz bandwidth.Therefore, sufficiently low rise and fall times of the LEDs arerequired. Moreover, arrays of several LEDs are commonly usedto increase the emitted optical power. To this end, the LEDsare connected in series and are modulated by the same current

Fig. 2. Conventional transmitter front-end.

signal. Laser diodes are an alternative light source. However,laser diodes are not considered as illumination devices becauseof their high costs and issues regarding eye safety.The main drawback of LEDs is their optical-power-versus-

current characteristic which is highly non-linear. Particularlyin combination with OFDM transmission, non-linearities area critical issue. Non-linearities directly result in a degradedsignal-to-interference-plus-noise ratio (SINR), and conse-quently in a degraded system performance [14], [15], [17].Therefore, OWC requires elaborated pre-distortion techniquesat the transmitter or complex equalization techniques at thereceiver to compensate for the non-linear characteristic ofLEDs. Without these techniques, LEDs can only be modulatedin a narrow, quasi-linear operational area. Consequently, theoperating point of LEDs has to be carefully chosen. However,this limited operational area largely restricts the dynamic rangeand the transmission power of OWC systems. Moreover, inorder to enable high data rate transmission, the whole trans-mitter chain has to operate at several tens to hundreds ofMHz. For instance in [33], a visible light transmitter system isreported which employs a DAC operating at a sampling rateof 275 Msps to provide a data rate of about 100 Mbit/s. Inorder to provide the required bandwidth, broadband amplifiercircuits have to be used as well. However, these circuit designsresult in a high parasitic electrical power consumption, whichis not transformed into optical power and is therefore wasted.Additionally, these circuit designs have to provide a good lin-earity characteristic. Therefore, conventional optical wirelesstransmitters are subject to complex and elaborated front-enddesigns.

IV. QUANTIZATION

In the following, the effect of transmitter induced quantiza-tion on the BER performance is analysed. Fig. 3 illustrates acontinuous OFDM symbol which is not quantized (dashed line)and a quantized OFDM symbol with a resolution of 5 bits (solidline). Due to the considered resolution of 5 bits, the quantizedsymbol can take only 32 discrete amplitude levels, whereas thecontinuous OFDM symbol has a (nearly) unlimited precision(depending on the precision of the computer system, which iscommonly 32 or 64 bits). As DCO-OFDM is considered withinthis paper, the illustrated symbols have a DC offset to providenon-negative amplitude values.Fig. 4 shows the BER performance of 16- and 64-QAM for

different bit resolutions. Note that any effects caused by channelestimation, synchronization etc. are not considered in these sim-ulations to highlight the effect of bit resolution on the BER per-formance. The error ratios are obtained by error counting takingthe BER testing considerations given in [34, Ch. 4] into account(how many symbols have to be simulated to get accurate BER


Fig. 3. Continuous OFDM symbol and quantized OFDM symbol.

Fig. 4. Effect of quantization on BER performance. (a) 16-QAM; (b) 64-QAM.

estimates). Table I shows the considered OFDM system param-eters. A FFT/IFFT size of and a cyclic prefix lengthof samples is applied. Due to Hermitian symmetry, the

TABLE ISYSTEM PARAMETERS USED FOR DCO-OFDM TRANSMISSION

amount of data sub-carriers is . The signal tonoise ratio (SNR) is defined as , where denotesthe received signal energy and is the power spectral den-sity of the noise. The noise is assumed to be the sum of am-bient shot light noise and thermal noise as commonly consid-ered in OWC [4]. Moreover, it is assumed to be independentof the transmitted symbols, and consequently it is assumed tobe additive white Gaussian noise (AWGN). The BER curves inFig. 4 show that a low resolution of only 2 or 3 bits leads toan error floor due to insufficient precision and low SINR. For16-QAM a resolution of 5 bits is sufficient as it provides only aminor SNR performance degradation (less than 1 dB) comparedto 16-QAM without quantization. For 64-QAM a resolution ofat least 7 bits is required to provide sufficient precision. Conse-quently, the simulation results show that -level QAM trans-mission can operate with a reduced bit resolution without majorperformance degradation.

V. PROPOSED TRANSMITTER CONCEPT

The proposed transmitter concept addresses the shortcomingsof conventional optical transmitter front-ends like complex cir-cuit designs, non-linearities and low power efficiency due tothe required high-speed DACs and high-bandwidth/high-cur-rent TCAs. The architecture of the proposed optical transmitterfront-end is shown in Fig. 5. The basic idea of this concept isto omit both the DAC and the analogue current control circuitwhich powers the LEDs. Instead of this analogue signal shaping,an array of LEDs is used which applies discrete power scaling.The LEDs are arranged in several groups which can be switchedon and off individually. Each LED group emits a specific op-tical intensity resulting in a discrete power level stepping. Likefor conventional transmitter front-ends, the waveform to be sentis represented by a digital bit vector whose elements are binaryvalues, i.e. ones and zeros. These binary values determine whichLED groups are switched on or off. Consequently, the basebandmodulator switches a specific combination of LED groups onor off according to the signal to be sent. Each signal sample isrepresented by a specific optical power level which is generatedby a particular combination of activated LED groups.As shown in Fig. 5, the transmitter circuit consists of

groups of LEDs, to . Each group is set to emit a con-stant optical power level (intensity) using a series resistor. The value of determines the electrical current which

powers the LED group . Consequently, determines the op-tical power level which is emitted by . Assuming

1Hermitian symmetry is considered.


Fig. 5. Proposed optical wireless transmitter front-end architecture.

is the maximum power level that the LEDs can emit, the cur-rent driving the first group of LEDs is set accordingly toresult in an emitted power level of . The remaining LEDgroups are driven with scaled currents to result in: ,

, and . TheLED group is modulated (switched on/off) by the most sig-nificant bit of the digital bit vector generated by the basebandmodulator. Group is modulated by the second most signif-icant bit, down to group which is controlled by the leastsignificant bit of the vector. Thus, each bit of the signal vectorto be sent is assigned to a specific emitter group radiating adefined optical intensity. As the discrete power levels emittedby the single LED groups constructively add up at the receiver,the proposed transmitter concept performs a digital-to-analogueconversion in the optical domain. Consequently, the circuit canbe considered as a direct digital input to analogue optical outputconverter with a resolution of bits. The bit resolutioncan be chosen with regard to the required precision as analysedin Section IV. If a resolution of 5 bits is considered, the quan-tized signal waveform can take 32 discrete intensity levels. Forinstance, the intensity level “12” is represented by the binarysequence “0 1 1 0 0”. If this intensity level is to be emitted, theLED groups and are activated, whereas the groups ,and are switched off. If DCO-OFDM is considered, the

LED group represents the DC bias offset which has an op-tical power level of . For , the OFDM time domainsignal samples have approximately Gaussian distribution [12].In order to increase the maximum transmission power and

range, it is additionally proposed to combine two or more LEDgroups to one common group. For instance, two groups can bejointly switched on and off to commonly build group . Ifboth groups emit the maximum power level , the overalltransmission power is doubled. This results in: ,

, and . Thisimplementation constitutes a compromise between signal preci-sion and transmission power, respectively SNR. The combina-tion of two LED groups provides a reduced resolution of onlybits if the same number of groups is used as above. Alter-

natively, the amount of LEDs per emitter group can also bevaried in order to create different power levels .The power levels of the individual LED groups can be ad-

justed using, e.g., an optical power meter. This adjustment guar-antees that the optical signal generated by the superposition ofall groups has best linearity. Hence, the appropriate values of

Fig. 6. Proposed LED arrangements for optical wireless transmitter front-endemploying 6 LED groups with 6 LEDs per group.

Fig. 7. Developed optical wireless transmitter front-end.

the series resistors , respectively the currents , can be deter-mined using an optical power meter.Moreover, it is beneficial toalign the LEDs in an arrangement which generates a light beamwith most homogeneous emitted optical power. Two possibleLED arrangements are illustrated in Fig. 6. The figures showa hexagonal (left hand side) and a diagonal (right hand side)arrangement for an implementation with 6 LED groups and 6LEDs per group. As shown, the single LEDs belonging to a spe-cific group, and thus emitting a specific optical power level ,are distributed across the LED array. For the hexagonal arrange-ment, the groups form inner and outer circles. For the diagonalarrangement, the single LEDs are diagonally arranged withinthe emitter square. It is also advisable to use LEDs with widebeam angles to ensure a broad coverage and a homogeneousillumination of the emitted light beam. The switching stages(see Fig. 5) that digitally control the different LED groups

are realised by fast solid state switching devices, such as tran-sistors or gate drivers for metal-oxide-semiconductor field-ef-fect transistors (MOSFETs). The switching speed of the stagesaffects the achievable bandwidth of the optical transmitter.

Only by fast switching speeds, a bandwidth in the range of5–50 MHz or higher can be achieved. However, the bandwidthalso depends on the typical rise and fall time of the LEDs. Sincesimple low-cost transistors are used as switching components,the proposed design enables the implementation of low-com-plex optical transmitters. A DAC and linear power amplifiersare not required. The OFDM signal processing can be done byan off-the-shelf field programmable gate array (FPGA) for in-stance, which enables flexible transceiver designs.Fig. 7 shows the developed optical wireless transmitter front-

end. The transmitter front-end has the proposed hexagonal LEDarrangement. It consists of 6 LED groups. Two of these groups


are combined to be jointly switched on and off in order to in-crease the transmission power and the SNR at the receiver. Con-sequently, the implemented transmitter provides a resolution of5 bits. The emitting diodes are off-the-shelf SFH 4502 IR LEDs[35] which have a wavelength of 950 nm. These diodes have arise and fall time of about 10 ns and a typical radiant power of40 mW. The semiangle of the LEDs is . The overall valueof raw materials of the transmitter is less than $25. As neitherheat sinks nor fans are required for cooling, the implementedtransmitter has a good power efficiency. The used receiver em-ploys a SD 445-14-21-305 PIN photo-diode [24] with a detectorarea of 1 . The response time of the photo-diode is about13 ns and the responsivity at 950 nm is about 0.65 A/W. Theactual receiver design is not within the scope of this paper.

VI. TRANSMISSION EXPERIMENTS

In this section, the results of some transmission experimentsusing the implemented optical transmitter front-end are pre-sented. Table I shows the considered OFDM system parame-ters. The transmitter front-end has a resolution of 5 bits which issufficient to perform a 16-QAM transmission with appropriateprecision as shown in Section IV. The switching speed ofthe LED groups is set to 25 Msps. The achievable bit rate ofDCO-OFDM employing -QAM can be approximated by:

(1)

Note that this calculation does not take additional symbolsinto account which are required for frame synchronization andchannel estimation. These parameters depend on the specificOFDM system implementation. Considering the system pa-rameters given in Table I, the implemented optical transmitterprovides a data rate of about 12, 24 or 48 Mbit/s if 2-, 4- or16-QAM is used. The data rate can be increased by employinghigher order M-QAM. For instance, 64-QAM provides a datarate of about 72 Mbit/s. However, the application of 64-QAMrequires an increased resolution of at least 7 bits as shownin Fig. 4(b). In order to further increase the data rate, higherswitching speeds can also be applied if the LEDs provide suffi-ciently low rise and fall times. For instance, applying 64-QAMwith a switching speed of provides a data rateof more than 144 Mbit/s.The maximum optical output power (all 6 LED groups are

switched on) emitted by the transmitter is. The overall electrical power consumption of all

LED groups is , which results in a totalpower dissipation of . Consequently,the LEDs have a power efficiency of .The total electrical power consumption of the transmitterfront-end is if all LED groups are switchedon. Therefore, the major portion of power is consumed bythe LEDs resulting in a percentage of consumed power of

. Therefore, the implemented opticaltransmitter has a low parasitic electrical power consumptionand has an overall power efficiency of .For a continous DCO-OFDM transmission, the mean electricalpower consumed is and the mean emittedoptical power is .Table II(a) shows the SNR at the receiver for different dis-

tances between the transmitter and the receiver. The transmitter

TABLE IISNR OVER DISTANCE FOR LOS AND NLOS TRANSMISSION

(A) LOS TRANSMISSION; (B) NLOS TRANSMISSION

Fig. 8. NLOS transmission setup.

and the receiver are aligned towards each other to provide adirected LOS link. The SNR is ascertained by measuring themean received signal energy of the OFDM symbols and bymeasuring the mean noise energy. The noise energy is mea-sured in the interframe gaps between the OFDM frames sincein these gaps no optical signal is emitted. Table II(b) shows theSNR for a non-line-of-sight (NLOS) transmission. Fig. 8 illus-trates the measurement setup for the NLOS transmission. Asshown, the transmitter and the receiver are straightly directedtowards a wall. The wall is a typical office wall with white paintand non-glossy surface. This setup provides a merely diffusetransmission scenario as an opaque obstacle between the trans-mitter and the receiver blocks the LOS link. The distance ofthe transmitter, respectively of the receiver towards the wall is. This means that the entire optical path length is about .

Since uncoded wireless data transmission requires an SNR ofat least 10 dB to provide reasonable BER (less than ) inAWGN channels [10], the implemented transmitter enables atransmission of up to 5.0 m in LOS scenarios and up to 3.5 min NLOS scenarios. The achievable transmission distance canbe increased if more powerful LEDs are used. Additionally,the amount of LEDs per emitter group can be enlarged to fur-ther increase the optical transmission power. The comparisonof Table II(a) and II(b) shows that for the same transmissiondistance, the SNRs of the LOS and NLOS link differ by about5.5 dB. This difference is induced by the higher transmissionloss of the NLOS link due to the wall reflections.Fig. 9 displays a received sine wave (with transmission in-

duced noise) transmitted by the optical front-end. The transmis-sion distance is 1m and a LOS link is applied. As shown, the dis-crete power level stepping generates a quantized sine wave withdiscrete amplitude values. Since a resolution of 5 bits is consid-ered, the quantized sine wave can take 32 discrete amplitude


Fig. 9. Received sine wave transmitted by optical transmitter front-end em-ploying discrete power level stepping.

levels. Employing a larger number of LED groups results in anincreased resolution, and thus provides more discrete amplitudevalues and smaller quantization intervals. Conventional trans-mitter front-ends typically employ DACs having a resolution ofabout 12 bits [33], whereas the effective number of bits (ENOB)is mostly smaller. Moreover, the received sine wave has a DCoffset of about 0.2 V due to the non-negativity constraint of theemitted optical signal because of IM/DD. Fig. 10(a) shows thespectrum of a quantized sine wave which is transmitted by theoptical front-end. For means of comparison, Fig. 10(b) showsthe spectrum of a sine wave which is generated by a conven-tional signal generator. This sine wave is electrically generatedwithout quantization effects and is transmitted via a cable. Sincea conventional signal generator is applied which provides ideallinearity characteristics, this electrical setup corresponds to anideal conventional optical transceiver as shown in Fig. 1 andFig. 2. Therefore, the implemented optical transmitter is com-pared in the following to an electrical transmission using a con-ventional signal generator. The electrical signal generator rep-resents the reference performance for the implemented opticaltransmitter. As shown in Fig. 10, the spectra of the optically andelectrically generated sine waves closely match. Besides the ac-tual frequency of the emitted sine wave, the harmonic frequen-cies can also be seen. Note that both transmission methods usea similar power level to enable a fair comparison. Moreover,it can be seen that the noise level of the electrical transmis-sion is lower. This is due to the fact that the electrical trans-mission uses a cable connection in comparison to the opticalwireless transmission which undergoes ambient shot light noise.Since both spectra closelymatch, the implemented optical trans-mitter shows good linearity characteristics. Additionally, Fig. 11shows the spectrum of a DCO-OFDM signal. An IFFT size of

is used and 16-QAM is applied. The spectrum of theoptical wireless transmission (1 m LOS) and the spectrum elec-trically generated by the signal generator (transmitted via cable)closely match. Again, the optical wireless transmission under-goes a higher noise level. Fig. 12 displays the measured fre-quency gain of the optical transceiver (transmitter and receiver).A directed 1 m LOS link between the transmitter and the re-ceiver is established. The optical transceiver has a 3 dB cut-off

Fig. 10. Spectrum of sine wave. (a) Optical wireless transmission using imple-mented transmitter front-end. (b) Electrical transmission via cable using signalgenerator.

frequency of about 18 MHz and a 6 dB cut-off frequency ofabout 25 MHz.In the following, a -QAM DCO-OFDM transmission is

considered. Each transmitted OFDM frame contains a symbolused for frame detection and synchronization. Furthermore,each frame comprises several data symbols and a dedicatedsymbol used for channel estimation. The estimation of thechannel coefficient of each OFDM sub-carrier is done byusing this pilot symbol. The estimated channel coefficientsare used for maximum-likelihood (ML) detection at the

receiver. Therefore, for the OFDM sub-carrier, the decoderdecides for the symbol which minimises the Euclideandistance between the actual received symbol and all potentialsymbols leading to

(2)

The QAM symbol transmitted on the OFDM sub-carrier isgiven by . For an ideal transmission in a mere AWGN channelwithout any fading and hardware effects, the received symbol


Fig. 11. Spectrum of DCO-OFDM signal. IFFT size is 256 and 16-QAM isapplied. (a) Optical wireless transmission using implemented transmitter front-end. (b) Electrical transmission via cable using signal generator.

Fig. 12. Frequency gain of optical transceiver.

is given by with . The AWGN is de-noted by . As a single-tap ZF channel equaliser is considered,the equalised received symbols are given by . If an ideal16-QAM DCO-OFDM transmission with 5 bit resolution and

Fig. 13. Comparison of simulated BER performance and actual DCO-OFDMtransmission experiments.

an SNR of 38 dB is simulated, the EVM is about 7.63 . TheEVM is definded as follows:

(3)

where and are the imaginary and the real part of thetransmitted symbol. The imaginary and the real part of

the equalised QAM symbol are represented byand . The number of symbols is given by . If an

actual 16-QAM DCO-OFDM transmission using the opticaltransmitter front-end is conducted, the measured SNR is about38 dB for a LOS transmission over a distance of 0.75 m. Themeasured EVM is about 8.42%. Consequently, the transmissionexperiments fairly comply with the simulation results.Fig. 13 shows the BER performance of a simulated

DCO-OFDM transmission using 2-, 4- and 16-QAM. Forthese simulations, an ideal AWGN channel is assumed asconsidered above. Moreover, the BERs of an actual opticalwireless transmission using the implemented transmitter aredisplayed. A relation between SNR and transmission distancefor the optical wireless transmission is given in Table II. Addi-tionally, the results of an electrical DCO-OFDM transmissionare shown as well. The electrical transmission is done usingthe same (de)modulating hardware (signal generator and oscil-loscope) and signal processing as for the optical transmission.However, the optical transmitter and receiver front-ends arereplaced by a (DAC) generating analogue electrical signals.Moreover, the optical wireless link is replaced by a cable whichdirectly connects the transmitter and the receiver. Since theelectrical signal generator has ideal linearity characteristics, it


can be regarded as an ideal conventional optical transceiver.As shown, the BER performance of the three scenarios isidentical for low SNR values. This is due to the fact that at lowSNRs, the noise is the most predominant impairment affectingthe link performance. However for higher SNR values, theelectrical transmission has a performance loss of about 1 dBcompared to the simulation results. This is due to practicalrealization issues and hardware effects like frequency offsetsbetween the transmitter and the receiver. These effects arenot considered in the mere AWGN simulations. Because ofthese hardware limitations, the performance of the electricaltransmission represents the practical performance bound of thetransmission experiments. This bound cannot be exceeded withthe used setup and devices. Consequently, the performance ofthe wired electrical transmission system can be considered asthe target performance for the optical wireless transmission. Asillustrated, the implemented optical wireless transmitter showsa minor performance loss of less than 1 dB compared to theelectrical transmission. This loss is due to the actual transmitterand receiver implementations. The receiver implementation isnot within the scope of this paper but it might be optimisedseparately. Moreover, the optical transmission is wireless com-pared to the electrical transmission which is wired. This resultsin an additional performance degradation due to the nature ofwireless transmission. Therefore, the proposed optical trans-mitter concept shows good performance as it closely matchesthe theoretical and practical error bounds.

VII. SUMMARY & CONCLUSIONS

An optical wireless transmitter which employs discretepower level stepping has been proposed and implemented.The transmitter consists of several LED groups which can beswitched on and off individually. The LED groups emit specificstepped optical intensities which constructively add up at thereceiver. Therefore, the optical transmitter can be used forintensity modulated signal transmission using techniques likePAM or OFDM. The proposed transmitter concept has goodlinearity characteristics and is not subject to the non-linear op-tical-power-versus-current characteristic of LEDs compared toconventional optical wireless transmitters. Complex pre-distor-tion techniques are not required. The transmitter is controlledby simple digital signal modulation, i.e. on-off-switchingof the LED groups. Thus, the implemented transmitter pro-vides a larger dynamic range and higher optical output powercompared to conventional optical front-ends which have tocarefully adjust the bias point and operating area of the LEDs.Moreover, the proposed solution enables a very simple andcost-effective transmitter design without the need for DACsor complex amplifier circuits. The digital-to-analogue con-version is done in the optical domain by the superpositionof the emitted discrete intensities. Additionally, the proposedtransmitter concept has a better power efficiency as it doesnot require high-bandwidth/high-current TCAs which wastepower. Transmission experiments prove the functionality ofthe implemented optical transmitter as its performance closelymatches the theoretical and practical error bounds. The perfor-mance of the optical transmitter is nearly similar to an electricaltransmission. The latter one provides ideal linearity characteris-tics, and therefore corresponds to an ideal conventional optical

transceiver. Nonetheless, both transmitter and receiver mightbe further optimised in future work to decrease the minor per-formance gap. Moreover, the bit resolution of the implementedtransmitter might be further increased to enable higher order-QAM transmission in order to provide larger data rates. A

shortcoming of the proposed transmitter concept is the fact thata broken LED within the transmitter array directly affects theemitted optical waveform. However, given the high life time ofLEDs of about 35000–50000 hours, this is unlikely to happen.

ACKNOWLEDGMENT

The authors acknowledge C. Blümm, J. Schalk andN. Schmitt for their support and technical advice.

REFERENCES[1] GBI Research, Visible Light Communication (VLC)—A Potential So-

lution to the Global Wireless Spectrum Shortage, Sep. 2011 [Online].Available: http://www.gbiresearch.com

[2] T. Komine and M. Nakagawa, “Fundamental analysis for visible-lightcommunication system using LED lights,” IEEE Trans. Consum. Elec-tron., vol. 50, no. 1, pp. 100–107, Feb. 2004.

[3] H. Elgala, R. Mesleh, and H. Haas, “Indoor optical wireless commu-nication: Potential and state-of-the-art,” IEEE Commun. Mag., vol. 49,no. 9, pp. 56–62, Sep. 2011.

[4] J. Kahn and J. Barry, “Wireless infrared communications,”Proc. IEEE,vol. 85, no. 2, pp. 265–298, Feb. 1997.

[5] J. Barry, J. Kahn, W. Krause, E. Lee, and D. Messerschmitt, “Simu-lation of multipath impulse response for indoor wireless optical chan-nels,” IEEE J. Sel. Areas Commun., vol. 11, no. 3, pp. 367–379, Apr.1993.

[6] F. R. Gfeller and U. Bapst, “Wireless in-house data communication viadiffuse infrared radiation,” Proc. IEEE, vol. 67, no. 11, pp. 1474–1486,Nov. 1979.

[7] O. Gonzalez, R. Perez-Jimenez, S. Rodriguez, J. Rabadan, and A.Ayala, “OFDM over indoor wireless optical channel,” Optoelectron.,IEE Proc., vol. 152, no. 4, pp. 199–204, Aug. 2005.

[8] H. Liu and G. Li, OFDM-Based BroadbandWireless Networks: Designand Optimization.. New York, NY, USA: Wiley, 2005.

[9] J. Heiskala and J. Terry, OFDM Wireless LANs: A Theoretical andPractical Guide, T. Reid and D. B. Williams, Eds. New York, NY,USA: Sams Publishing, 2002.

[10] A. Goldsmith, Wireless Communications.. Cambridge, U.K.: Cam-bridge Univ. Press, 2005.

[11] M. Afgani, H. Haas, H. Elgala, and D. Knipp, “Visible light communi-cation using OFDM,” in Proc. TRIDENTCOM, Barcelona, Spain, Mar.1–3, 2006, pp. 129–134.

[12] J. Armstrong, “OFDM for optical communications,” J. Lightw.Technol., vol. 27, no. 3, pp. 189–204, Feb. 2009.

[13] D. Barros, S. Wilson, and J. Kahn, “Comparison of orthogonal fre-quency-division multiplexing and pulse-amplitude modulation in in-door optical wireless links,” IEEE Trans. Commun., vol. 60, no. 1, pp.153–163, Jan. 2012.

[14] E. Costa,M.Midrio, and S. Pupolin, “Impact of amplifier nonlinearitieson OFDM transmission system performance,” IEEE Commun. Lett.,vol. 3, no. 2, pp. 37–39, Feb. 1999.

[15] E. Costa and S. Pupolin, “M-QAM-OFDM system performance inthe presence of a nonlinear amplifier and phase noise,” IEEE Trans.Commun., vol. 50, no. 3, pp. 462–472, Mar. 2002.

[16] D. Dardari, V. Tralli, andA. Vaccari, “A Theoretical characterization ofnonlinear distortion effects in OFDM systems,” IEEE Trans. Commun.,vol. 48, no. 10, pp. 1755–1764, Oct. 2000.

[17] W. Jun and Y. Chenyang, “The influence of analog device on OFDMsystem,” in Proc. ICCT, Beijing, China, Apr. 9–11, 2003, vol. 2, pp.1060–1062.

[18] B. Inan, S. C. J. Lee, S. Randel, I. Neokosmidis, A.M. J. Koonen, and J.W.Walewski, “Impact of LED nonlinearity on discrete multitone mod-ulation,” IEEE/OSA J. Opt. Commun. Netw., vol. 1, no. 5, pp. 439–451,Oct. 2009.

[19] I. Neokosmidis, T. Kamalakis, J. W. Walewski, B. Inan, and T. Sph-icopoulos, “Impact of nonlinear LED transfer function on discrete mul-titone modulation: Analytical approach,” J. Lightw. Technol., vol. 27,no. 22, pp. 4970–4978, Nov. 2009.


[20] H. Elgala, R. Mesleh, and H. Haas, “A study of LED nonlinearity ef-fects on optical wireless transmission using OFDM,” in Proc. WOCN,Cairo, Egypt, Apr. 28–30, 2009.

[21] I. Stefan, H. Elgala, R. Mesleh, D. O’Brien, and H. Haas, “Opticalwireless OFDM system on FPGA: Study of LED nonlinearity effects,”in Proc. IEEE VTC, Budapest, Hungary, May 15–18, 2011, pp. 1–5.

[22] H. Elgala, R. Mesleh, and H. Haas, “Non-linearity effects and predis-tortion in optical OFDM wireless transmission using LEDs,” Intersci.Int. J. Ultra Wideband Commun. Syst. (IJUWBCS), vol. 1, no. 2, pp.143–150, 2009.

[23] A. Behravan and T. Eriksson, “PAPR and other measures for OFDMsystems with nonlinearity,” in Proc 5th Int. Symp. Wireless PersonalMultimedia Commun., Honolulu, HI, USA, Oct. 27–30, 2002, vol. 1,pp. 149–153.

[24] Advanced Photonix Inc., Datasheet: Red Enhanced High Per-formance Silicon Photodiode SD 445-14-21-305, Apr. 2006[Online]. Available: http://www.advancedphotonix.com/ap_prod-ucts/pdfs/SD445-14-21-305.pdf

[25] S. Dimitrov, S. Sinanovic, and H. Haas, “Clipping noise in OFDM-based optical wireless communication systems,” IEEE Trans.Commun., vol. 60, no. 4, pp. 1072–1081, Apr. 2012.

[26] J. Armstrong and A. Lowery, “Power Efficient Optical OFDM,” Elec-tron. Lett., vol. 42, no. 6, pp. 370–372, Mar. 16, 2006.

[27] S. C. J. Lee, S. Randel, F. Breyer, andA.M. J. Koonen, “PAM-DMT forintensity-modulated and direct-detection optical communication sys-tems,” IEEE Photon. Technol. Lett., vol. 21, no. 23, pp. 1749–1751,Dec. 2009.

[28] A. M. Khalid, G. Cossu, R. Corsini, P. Choudhury, and E. Ciaramella,“1-Gb/s transmission over a phosphorescent white LED by using rate-adaptive discrete multitone modulation,” IEEE Photon. J., vol. 4, no.5, pp. 1465–1473, Oct. 2012.

[29] C. W. Chow, C. H. Yeh, Y. F. Liu, and P. Y. Huang, “Mitigation ofoptical background noise in Light-Emitting Diode (LED) optical wire-less communication systems,” IEEE Photon. J., vol. 5, no. 1, pp. 7 900307–7 900 307, Feb. 2013.

[30] C. H. Yeh, Y. F. Liu, C.W. Chow, Y. Liu, P. Y.Huang, andH. K. Tsang,“Investigation of 4-ASK modulation with digital filtering to increase20 times of direct modulation speed of white-light LED visible lightcommunication system,”OSA Opt. Exp., vol. 20, no. 15, pp. 16 218–16223, Jul. 2012.

[31] J. Armstrong and B. J. C. Schmidt, “Comparison of asymmetricallyclipped optical OFDM and DC-biased optical ofdm in AWGN,” IEEECommun. Lett., vol. 12, no. 5, pp. 343–345, May 2008.

[32] S. Dimitrov, S. Sinanovic, and H. Haas, “Signal shaping and modula-tion for optical wireless communication,” J. Lightw. Technol., vol. 30,no. 9, pp. 1319–1328, May 2012.

[33] O. Bouchet, G. Faulkner, L. Grobe, E. Gueutier, K.-D. Langer, S.Nerreter, D. O’Brien, R. Turnbull, J. Vucic, J. W. Walewski, and M.Wolf, “Deliverable D4.2b physical layer design and specification,”Proc. 7th Framework Programme Inf. Communication Technol. Feb.2011 [Online]. Available: http://www.ict-omega.eu

[34] D. Derickson and M. Müller, Digital Communications: Test and Mea-surement.. Englewood Cliffs, NJ, USA: Prentice Hall, 2008.

[35] OSRAM Opto Semiconductors GmbH, Datasheet: SFH 4501, SFH4502, SFH 4503 High Power Infrared Emitter, Jan. 2008 [Online].Available: http://catalog.osram-os.com

Thilo Fath holds a M.Sc. degree in Electrical Engineering and InformationTechnology from the University of Karlsruhe, Germany. In 2008, he joinedEADS Innovation Works Germany, where he currently works on optical wire-less communications. Since 2010, he has beenwith the Institute for Digital Com-munications (IDCOM) at the University of Edinburgh, UK, where he is workingtowards his Ph.D. degree in Electrical Engineering. His main research interestsare in the area of digital signal processing with a particular focus on MIMOtechniques and optical wireless communications.

Christoph Heller got a degree in communications engineering from the RWTHAachen University, Germany in 2006 and received his PhD in 2010. He joinedthe European Aeronautic Defence and Space Company (EADS) in 2005 andis employed in the communications research department of EADS InnovationWorks. Christoph Heller is responsible for research activities in the field ofnext-generation wireless data links for aeronautical and military applications,focusing on digital baseband processing and physical layer designs. He is in-volved in several EADS internal, national and international research projectsdealing with software-defined radio and cognitive radio.

Harald Haas (SM’98–AM’00–M’03) holds the Chair of Mobile Communica-tions in the Institute for Digital Communications (IDCOM) at the Universityof Edinburgh and he currently is the CTO of a university spin-out companyVLC Ltd. His main research interests are interference coordination in wirelessnetworks, spatial modulation and optical wireless communications. Prof. Haasholds 23 patents. He has published more than 50 journal papers including aScience Article and more than 150 conference papers. Nine of his papers areinvited papers. Prof. Haas has co-authored a book entitled “Next GenerationMobile Access Technologies: Implementing TDD” with Cambridge UniversityPress. Since 2007 Prof. Haas has been a Regular High Level Visiting Scien-tist supported by the Chinese “111 program” at Beijing University of Posts andTelecommunications (BUPT). He was an invited speaker at the TED Globalconference 2011, and his work on optical wireless communications was listedamong the “50 best inventions in 2011” in the Time Magazine. He received theEPSRC Established Career Fellowship in 2012.

1734 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 11 ... · 1734 JOURNAL OF LIGHTWAVE TECHNOLOGY,...

Documents

Transcript of 1734 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 31, NO. 11 ... · 1734 JOURNAL OF LIGHTWAVE TECHNOLOGY,...