Adaptive Performance Improvement of OFDMRadio over Fiber Systems

Bharath Umasankar*, A. D. S. Jayalath and Xavier FernandoDept. Electrical and Computer Engineering

Ryerson University,TorontoEmail: bumasank,fernando@ee.ryerson.ca

Abstract Nonlinear distortion due to large peak to averagepower ratio is a major concern with OFDM systems. In thispaper, we present an efficient adaptive modulation techniqueto mitigate nonlinear distortion effects of OFDM radio signalstransmitted over optical fiber (Radio-over-Fiber: ROF). First, wemodulate all subcarriers at primary (high) level. Then we identifythe subcarriers with high distortion and appropriately reduce themodulation level on those subcarriers to secondary levels. Thisis done at the transmitter side for each OFDM symbol. Thisprocedure is repeated until the nonlinear distortion is below apredetermined threshold. This technique is shown to improve theBER performance considerably while the reduction in data ratefor a system with 64 subcarriers and 16 QAM as primary and4 QAM as secondary modulation levels is around 4%.

Index Terms OFDM,Radio-over-fiber(ROF), nonlinear dis-tortion,adaptive modulation.

I. INTRODUCTION

OFDM will be the most widely used technique for broad-band communications for years to come. OFDM showsexcellent performance in wireless (WiMAX, 4G) as well as inwired (DSL, broadband over power line) environments due toits robustness to frequency dependent impairments. In wirelessenvironment, OFDM divides the frequency selective fadingchannel into number of flat fading channels and conquers. Inwired environments like DSL, it provides robustness againstreflections and high frequency attenuation.

There are many occasions in which OFDM modulated RFsignals need to be transmitted over optical fiber, referred toas in OFDM-ROF systems [1]. An OFDM-ROF system maybe followed by wireless channel (in fiber-wireless systems) orcopper wire (in some DSL or fiber-coaxial systems) [2].

Although OFDM provides excellent benefits, the large peakto average power ratio (PAPR) is an important issue in OFDMsystems that demands large linear range for the transmissionchannel. The radio-over-fiber links typically suffer from lim-ited linear dynamic range due to the nonlinear distortion of theoptical transmitter, either the laser diode or the Mach-Zehndermodulator (MZM).

The distortion caused by MZM/laser nonlinearity, increasedaverage power requirement in intensity modulated opticalsystems ,loss of orthogonality of subcarriers resulting ininter-carrier interference(ICI), increased hardware complexityand the need for more processing power are some issuesof OFDM. The nonlinear distortion due to Laser/opticalchannel could be a major limiting factor in high speed

wired/wireless OFDM systems employing thousands of sub-carriers (eg:1024,2048,4096,8192). There has been a numberof papers in recent times which concentrate on nonlineardistortion and peak to average power ratio (PAPR) reductionof OFDM systems [3], [4], [5], [6], [7].

In the paper [7] the authors illustrate that if informationabout the nonlinearity of the channel or the HPA/Laser isavailable at the transmitter side, then a better performance canbe achieved by directly minimizing the distortion as a resultof the nonlinearity rather than minimizing the OFDM signalsPAPR. This approach is especially useful if the nonlinearity isrelatively stable and can be estimated with sufficient accuracyas in the case of MZM/laser nonlinearity in radio-over-fibersystems. In this work we present a simple adaptive modulationtechnique for nonlinear noise mitigation in optical OFDMsystems, assuming that information about the nonlinearity isavailable at the transmitter side.

II. SYSTEM MODELThe baseband OFDM signal can be represented by,

x(t) = 1N

N1

k=0

Xk exp(j2kt

T), 0 t T (1)

where, N is the total number of subcarriers, T is the OFDMsymbol duration and X(k) is mapped M QAM symbol asshown in Fig. 1. Since N is generally very large, accordingto the central limit theory, x(t) approximates Gaussian distri-bution, and its levels are infinite [7]. In high bit rate opticalsystems the Mach-Zehnder modulator is commonly used. Itis shown in [6] that the baseband equivalent of this ROFsection corresponds to a memoryless nonlinearity that can becharacterized by a Bessel function of the first kind.

f() = 2J1(

v). (2)

where, = |x(t)|, and hence represents a pure AM-AM com-pression. v is a device parameter representing the smallestdrive level for which the modulator produces phase shiftbetween its branches.

III. ADAPTIVE MODULATION TECHNIQUEIn this section we describe a simple algorithm to mitigate

the nonlinear distortion assuming the nonlinearity can beestimated as described in [7]. For each OFDM symbol, its

1-4244-1449-0/07/$25.00 2007 IEEE 611

Data bits Adaptive Modulator

Adaptive ModulationAlgorithm

IFFTS/P

FFTEstimatednonlinearity

Laser/Fibernonlinearity

Calculatedistortion

Nonlinear distortion estimation and subcarrier selection

adaptive_subcarrier_infodistortion

x(t)

xdistorted(t)

X(k)

Fig. 1. Block schematic of the adaptive modulation system

corresponding time domain signal x(t) is sent through the es-timated nonlinearity and an FFT is performed at the receivedsignal as shown in Fig.1. Then the distortion on the symbolsof each subcarrier is calculated. Then the subcarriers whichexperience high distortion are estimated and the modulationlevel on those subcarriers is reduced. Then the combinedsignal is again transmitted over the nonlinear device and theprocedure is repeated.

When the distortion on all the subcarriers is below a certainthreshold level, the OFDM symbol is transmitted over theactual ROF link along with side information containing themodulation info of each subcarrier. This approach differsfrom the criteria developed in [7] where the goal is to mini-mize the difference between the original OFDM time domainsignal x(t) and the estimated distorted time domain signalxdistorted(t) after nonlinearity. However, in our approach weattempt to minimize the estimated distortion on each subcarrier(by taking an FFT after the estimated nonlinearity at thetransmitter side as in Fig.1).

This adaptive modulation approach is made possible sincethe serial bits to be transmitted are converted to parallel beforeOFDM modulation-that is we have the freedom to choosethe modulation level of each subcarrier without having tochange the modulation level of other subcarriers from the poolof available parallel bits. Another observation is that sincethe modulation levels on the subcarriers are independent, achange in one subcarrier would not considerably affect thetime domain signal x(t) according to the IFFT equation in(1) -this means the distortion effects after nonlinearity alsowould be relatively same except for the subcarrier positionwhose modulation is changed. Therefore by identifying thesubcarriers (frequency positions) that are more likely to havedistortion and reducing their modulation level we can improvethe BER performance for the symbols on those subcarriers.

With our algorithm, x(t) is passed through the estimatednonlinearity and an FFT is taken and the distortion onall subcarriers is estimated. The subcarrier indices whosedistortion level exceeds a predetermined threshold (dthr) aredetermined and sorted according to descending order of dis-tortion. Out of these high distortion subcarrier indexes, only

the first selection length number of indexes are chosen asthe selected subcarrier indexes and added to the solution setas described below. The modulation level of the entries in thesolution set are reduced to a lower order. We need not reducethe modulation level on all the subcarriers whose distortionexceed the threshold. This is because when the modulationlevel of the highest distortion subcarriers are reduced it bringsdown the distortion level of other subcarriers as well. If theimprovement on distortion of the other subcarriers still fallsbelow the threshold, this procedure is repeated and moresubcarriers are added to the solution set in the same manner.

When all the subcarriers have acceptable distortion levels ( /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth 8 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /FlateEncode /AutoFilterGrayImages false /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 2.33333 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /PDFX1a:2001 ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /DocumentCMYK /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice