Design and Analysis of a CO-OFDM

Transmitter With Limited Modulator

Extinction Ratio
Volume 6, Number 3, June 2014

Lingchen Huang
Changjian Guo
Sailing He, Fellow, IEEE

DOI: 10.1109/JPHOT.2014.2317676
1943-0655 Ó 2014 IEEE
IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

Design and Analysis of a CO-OFDM

Transmitter With Limited Modulator
Extinction Ratio
Lingchen Huang, 1;2 Changjian Guo, 1 and Sailing He,1;2;3 Fellow, IEEE

1
ZJU-SCNU Joint Research Center of Photonics, South China Academy of Advanced Optoelectronics,
South China Normal University, Guangzhou 510631, China
2
Centre for Optical and Electromagnetic Research, State Key Laboratory for Modern Optical
Instrumentation, Zhejiang University, Hangzhou 310058, China
3
Division of Electromagnetic Engineering, School of Electrical Engineering, Royal Institute of
Technology, 100 44 Stockholm, Sweden

DOI: 10.1109/JPHOT.2014.2317676
1943-0655 Ó 2014 IEEE. Translations and content mining are permitted for academic research only.
Personal use is also permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Manuscript received February 12, 2014; revised March 26, 2014; accepted April 5, 2014. Date of pub-
lication April 15, 2014; date of current version May 2, 2014. This work was supported in part by the
Guangdong Innovative Research Team Program under Grant 201001D0104799318, by the Ministry of
Science and Technology of China through the 863 Project under Grant 2012AA011001, by the National
Natural Science Foundation of China under Grant 61307090, and by the China Postdoctoral Science
Foundation under Grant 2013M531865. Corresponding author: C. Guo (e-mail: changjian.guo@coer-
scnu.org).

Abstract: In this paper, we consider the design and analysis of a coherent optical OFDM
(CO-OFDM) transmitter using an optical in-phase/quadrature (I/Q) modulator with limited
extinction ratio (ER) under four modulation formats. The transmitter-side digital signal
processing (DSP), including hard clipping and predistortion, is introduced to combat the
penalty induced by the limited ER of the modulator and the quantization noise of the digital-
to-analog converter (DAC). The DAC resolution needed for transmitters using modulators
with various ER values are examined. The peak-to-peak driving voltage (Vpp) and the
clipping ratio (CR) of the transmitted signals are also optimized with respect to minimizing
the Q-factor penalty for the modulator with different ER values. The simulation results show
that, unlike previous work using modulators with an infinite ER, the optimum Vpp and CR
values vary with different ERs and modulation levels.

Index Terms: Coherent optical OFDM (CO-OFDM), extinction ratio (ER), digital signal
processing (DSP).

1. Introduction
Coherent optical OFDM (CO-OFDM) has attracted lots of interest due to its high spectral efficiency
(SE) and robustness to fiber dispersion [1], and is considered as a promising candidate for long haul
optical fiber transmission systems. However, multicarrier signals including OFDM have high peak-
to-average power ratio (PAPR) [2]. Signals with a high PAPR are sensitive to fiber nonlinearity
distortions. In CO-OFDM systems, an optical in-phase/quadrature (I/Q) modulator is usually
employed to upconvert the baseband OFDM signal to optical domain [3]. The optical I/Q modulator,
which has a nonlinear and peak-limited transfer function, will also compromise the signal quality. It’s
therefore beneficial to combat these impairments through transmitter side digital signal processing
(DSP). Recently, optical OFDM systems employing adaptive modulation schemes has also gained

Vol. 6, No. 3, June 2014 7200207

IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

Fig. 1. CO-OFDM transmitting scheme using an optical I/Q modulator.

much attention [4], [5]. Hence, it is also desirable to study the optimal conditions to operate the
optical transmitter for various modulation formats.
Several contributions have been made to investigate the transmitter design in both multicarrier
[3], [6]–[8] and single-carrier systems [9]–[11]. The nonlinear characteristic of optical I/Q modulator
and the optimal biasing point is investigated in [3]. The quantization effect of DACs and transmitter
signal processing including hard clipping and predistortion is studied in [6] and [7]. The Q-factor
penalty was shown to remain unchanged at small modulation index (small Vpp) scenarios (from 0 to 1)
when infinite ER is applied [6]. However, previous researches merely considered QPSK and optical
I/Q modulators with infinite ER [3], [6], [7]. None of these works considered optical I/Q modulator
with limited ER for CO-OFDM systems.
In this work, we consider the transmitter design for CO-OFDM systems under limited ER and
DAC resolution values. We analyze the impact of hard clipping, predistortion, and peak-to-peak
voltage (Vpp) of the modulator driving signals on the system performance. Four modulation formats,
i.e., 4-QAM, 16-QAM, 32-QAM, and 64-QAM, are investigated. The results show that the system
performance could be optimized by carefully choosing appropriate values of the clipping ratio (CR)
and the Vpp of the modulator driving signals. We’ll show that unlike previous work using modulators
with infinite ER [6], the optimum Vpp and CR vary with different ERs and modulation formats.

2. OFDM Transmitter Architecture

Fig. 1 shows the conceptual diagram of the CO-OFDM transmitter scheme. In Fig. 1, the block
BOFDM Tx[ includes mapping the transmitted sequences into subcarriers, performing inverse
discrete Fourier transform (IDFT) and adding the cyclic prefix (CP). The real/image part of the
synthesized signal is then processed before digital-to-analog (D/A) conversion, after which the
analog signal is sent to driving an optical I/Q modulator.
The I/Q modulator employed in this paper is an x-cut LiNbO3 optical modulator consisting of an
MZM with an MZM inserted in each arm, as shown in Fig. 2 [12]. The two inner MZMs in the arm are
named child MZM and the outer MZM is named parent MZM according to the specification of I/Q
modulator [12]. The MZMs are comprised of two parallel phase modulators connected at both ends
by Y-junctions [13] or directional couplers [14]. These two cases behave just the same [14], and in
this contribution, we adopt the Y-junctions scheme. In ideal cases, a Y-junction is characterized by
a 50/50 power splitter or combiner; and the phase modulators have identical lengths and insertion
losses. Whereas in reality, both unbalanced power distribution (resulted from both unequal splitting/
combining ratios of Y-junctions and unequal insertion losses between phase modulators) and
phase difference between two phase modulators exist. Under this circumstance, and when push–
pull symmetry is used, the electrical field at the output of an MZM is given by [10]
EMZM;i ðt Þ h j2V
v ðt Þ  v ðt Þ 
i
EMZM;o ðt Þ ¼ e  þj 2 þ   e j2V j 2 (1)
1þ
where EMZM;i ðt Þ and EMZM ;o ðtÞ are the input and output electrical field, respectively. v ðt Þ is the
analog driving voltage comprised of a DC offset d and AC coupled signal sðt Þ, i.e., v ðt Þ ¼ d þ sðt Þ.

Vol. 6, No. 3, June 2014 7200207

IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

Fig. 2. I/Q modulator structure.

V is the half-wave driving voltage of the MZM.  is the scaling factor used to account for un-
balanced splitting/combining ratio between both arms of the MZM ( ¼ jE2 j=jE1 j, where E1 and E2
are the electrical fields in arm 1 and 2 of the MZM, respectively), and can also be calculated from
the definition of extinction ratio (ER)
 
1þ 2
ER ¼ : (2)
1
In the ideal case where the ER is infinite, the scaling factor  ¼ 1. By controlling the bias voltage
d ¼ V ð= þ 1Þ, Eq. (1) becomes
    
sðt Þ 1 sðt Þ
EMZM;o ðt Þ ¼ EMZM;i ðt Þ sin j  cos : (3)
2V 1þ 2V
It was shown that linear upconversion can be achieved by biasing both child MZMs of the optical
I/Q modulator at null point [3]. By controlling the bias of both child MZMs at null point and the bias of
parent MZM to bring in a =2 phase difference between the two child MZMs, the modulator output
can be written as
     
Ein ðt Þe j 4 s1 ðt Þ 1  2 s2 ðt Þ
Eo ðt Þ ¼ sin þ 3 cos
1 þ 3 2V;1 1 þ 2 2V;2
    
s2 ðt Þ 1  1 s1 ðt Þ
þ j 3 sin  cos (4)
2V;2 1 þ 1 2V;1
where Ein ðt Þ and Eo ðt Þ are the input and output electrical field of the I/Q modulator, respectively.
s1 ðt Þ and s2 ðt Þ are the driving voltage applied to the two child MZMs. V;1 , V;2 , and V;3 are the half-
wave driving voltage of the child MZM1, child MZM2, and the parent MZM, respectively. 1 , 2 and
3 are the scaling factor of the child MZM1, child MZM2 and the parent MZM, respectively. In the
following discussion, we assume that V;1 ¼ V;2 ¼ V;3 and 1 ¼ 2 .
The first term in the bracket of Eq. (3) shows that the signal of interest endures a nonlinear sine
shape transfer function. It can be compensated by using a predistortion approach [3], defined as:

Vout ¼ 2V  sin1 ðVin Þ= (5)

where Vin and Vout denote the signal before and after predistortion, respectively.
The second term in the bracket of Eq. (3) shows that a portion of the modulation power, the
maximum magnitude of which is determined by the ER value, is drifted to the image part. For
signals with large magnitudes, this term will introduce extra noises to the signals and effectively
reduce the signal-to-noise ratio (SNR), whereas for signals with small magnitudes, this term can be

Vol. 6, No. 3, June 2014 7200207

IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

approximated to a constant, and thus SNR may not be reduced explicitly. However, relatively more
optical power will be shifted to the nondata carrying optical carrier.
It should be noted that, from Eq. (4), the finite ER of the parent MZM induced by the scaling
factor 3 would only cause I/Q amplitude imbalance, which is beyond the scope of this contribution.
Ref. [15] has proposed an approach for OFDM I/Q imbalance compensation. In this work, we
consider the case when scaling factor 3 ¼ 1. The goal of this work is to extend the analysis by
considering the impact of limited ER of the child MZMs, since the state-of-the-art optical I/Q
modulator typically has a child MZM ER of around 20 dB [12].
Despite the modulation nonlinearity induced by the MZM, another major challenge of OFDM
systems is the high PAPR. It will result in four problems: 1) severe quantization noise after D/A
conversion; 2) insufficient modulation depth; 3) intermodulation distortion induced by power
amplifier; 4) severe fiber nonlinearity distortions. Several approaches to reduce PAPR have been
proposed both in wireless or optical fiber communications, such as clipping and filtering [16], coding
[17], partial transmit sequence [18], selected mapping [19], etc. The most straightforward approach
to reduce PAPR is to hard clip the time domain signals to a lower level. Despite the fact that clipping
would introduce distortion to the signals, the penalty can be neglected for moderate clipping levels
[6]. The clipping process can be expressed as [6]
sc ½n ¼ min½A; maxðA; s½nÞ (6)
where s½n and sc ½n denote the signal before and after hard clipping, respectively. The clipping
level A is determined by the clipping ratio CR ¼ A=srms , where srms is the root mean square (RMS)
value of s½n.

3. Simulation and Results

Monte Carlo simulations were carried out to investigate the transmitter design of the CO-OFDM
system. The parameters used in the simulations were: OFDM symbol period of 8 ns, 256 subcarriers,
The DAC sample rate 32 Gs/s [20], 1/8 of an IDFT block used as guard interval, and 1000 OFDM
symbols for each simulation. Among the 256 subcarriers, 128 were m-QAM modulated, 10 were
reserved for pilot tones, and the rest were zero-padded for oversampling. The laser linewidth at both
transmitter and receiver were set to 10 KHz. The baseline Q-factor value for 4-QAM is 8.53 dB
(BER ¼ 3:8E  3 for 7% overhead HD-FEC) [21]–[23], calculated at an optical signal-to-noise ratio
(OSNR) of 10.7 dB. For 16-QAM, 32-QAM, and 64-QAM, a smaller baseline Q-factor of 6.25 dB
(BER ¼ 2E  2 for 20% overhead SD-FEC) calculated at OSNR of 17.2, 21.8, and 27.5 dB,
respectively, is used to reduce the OSNR requirement [24]–[26]. Note that the Q factors throughout
this paper are calculated from BER using the complementary error function ðerfcÞ. Infinite DAC
resolution, infinite ER, and zero laser linewidth are assumed when computing the baseline Q-factor.
The system performance discussed below is based on Q-factor penalty, defined as the reduction
from the reference Q-factor (8.53 dB for 4-QAM, and 6.25 dB for the rest) when the OSNR is set to
10.7, 17.2, 21.8 and 27.5 dB for 4-QAM, 16-QAM, 32-QAM, and 64-QAM, respectively. A predis-
tortion algorithm using Eq. (5) is also applied in the simulation.
Fig. 3(a)–(d) show the contour plots of the Q-factor penalty versus the ER of the child MZM and
DAC resolution for 4-QAM, 16-QAM, 32-QAM, and 64-QAM, respectively. The peak-to-peak driving
voltage (Vpp) and the CR are optimized for each combination of DAC resolution and ER value, i.e.,
for each combination of DAC resolution (in terms of effective number of bits, ENOB) and ER value,
the Vpp and CR are chosen so that a minimum Q-factor penalty is obtained. For all the four
modulation formats, larger values of ER and DAC resolution would promise a better system
performance. For a set Q-factor penalty (1 dB for instance), using higher order modulation formats
would increase the requirements of the DAC resolution and the ER of the child MZM. For
commercially available optical I/Q modulators [12] with 20-dB ER, to achieve a 1-dB or less Q-factor
penalty, the DAC resolution required for 4-QAM, 16-QAM, 32-QAM, and 64-QAM modulation
formats are less than 4, 4.7, 5.3, and 6.3, respectively. The DAC requirement will be even more
stringent when no predistortion algorithm is used, especially in higher-order modulation formats

Vol. 6, No. 3, June 2014 7200207

IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

Fig. 3. Contour plots of Q-factor penalty versus the resolution of the DAC and the ER of the child MZM
with (solid line) predistortion and without (dashed line) predistortion for (a) 4-QAM, (b) 16-QAM,
(c) 32-QAM, and (d) 64-QAM.

Fig. 4. Contour plots of Q-factor penalty against the Vpp of the driving signal and the ER of the child
MZM for (a) 4-QAM, (b) 16-QAM, (c) 32-QAM, and (d) 64-QAM.

scenarios. For instance, without the usage of the predistortion algorithm, the required DAC
resolution will be 6.7 bits for 64-QAM modulated signals.
Fig. 4(a)–(d) shows the contour plot of Q-factor penalty as a function of Vpp and ER for 4-QAM,
16-QAM, 32-QAM, and 64-QAM, respectively. The DAC resolution was set to 5 bits: a typical ENOB

Vol. 6, No. 3, June 2014 7200207

IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

Fig. 5. Contour plots of Q-factor penalty against the CR of the driving signal and the ER of the child MZM
for (a) 4-QAM, (b) 16-QAM, (c) 32-QAM, and (d) 64-QAM.

that today’s state-of-the-art high-speed DAC can achieve [20]. For each combination of Vpp and ER
value, the CR value was chosen so that a minimum Q-factor penalty was obtained. The predis-
tortion algorithm was enabled. As shown in Fig. 4, we have the following observations: 1) for 4-QAM,
a larger Vpp always promises better performance as long as it is less than V ; 2) for 16-QAM,
32-QAM, and 64-QAM, there exists an optimum Vpp G V in order to achieve the minimum Q-factor
penalty for each ER value; 3) for 16-QAM, 32-QAM, and 64-QAM, as the ER value increases, the
optimum Vpp value decreases; 4) when infinite ER value is used, a smaller Vpp always promises
better performance, which agrees with previous observations [6]. For commercially available optical
I/Q modulators [12] with 20-dB ER, the optimum Vpp values for 4-QAM, 16-QAM, 32-QAM, and
64-QAM are about 1, 0.9, 0.8, and 0.7 times V , respectively.
We also numerically investigated the optimum CR with respect to minimize the Q-penalty for
different ER values, as shown in Fig. 5. For each combination of CR and ER value, the Vpp value
was chosen so that a minimum Q-factor penalty was obtained. The predistortion algorithm was also
enabled. The DAC resolution was assumed to be 5 bits. It is shown that, for all the four modulation
formats, larger CR values (less clipping) are required when modulators with larger ER values are
used. For commercially available optical I/Q modulators [12] with 20-dB ER, CR of about 2.7, 3, 3.2,
and 3.4 dB are found to offer the minimum Q-factor penalty for 4-QAM, 16-QAM, 32-QAM, and
64-QAM, respectively.

4. Conclusion
In a CO-OFDM system, signal processing including digital hard clipping and predistortion could be
employed to combat the large PAPR of OFDM signals and compensate the nonlinearity of the
optical modulators at the transmitter side. For the four modulation formats we have investigated in
this paper, the CR and the peak-to-peak voltage (Vpp) could be optimized with respect to minimize
the Q-factor penalty as shown in Figs. 4 and 5. Unlike previous work using modulators with infinite
ER, the optimum Vpp and CR values have been shown to vary with different ERs and modulation
levels.

Vol. 6, No. 3, June 2014 7200207

IEEE Photonics Journal CO-OFDM Transmitter With Limited ER

References
[1] W. Shieh, X. Yi, Y. Ma, and Q. Yang, BCoherent optical OFDM: Has its time come? [Invited],[ J. Opt. Netw., vol. 7, no. 3,
pp. 234–255, Mar. 2008.
[2] J. Proakis and M. Salehi, Digital Communications, 5th ed. New York, NY, USA: McGraw-Hill, 2007.
[3] Y. Tang, W. Shieh, X. Yi, and R. Evans, BOptimum design for RF-to-optical up-converter in coherent optical OFDM
systems,[ IEEE Photon. Technol. Lett., vol. 19, no. 7, pp. 483–485, Apr. 2007.
[4] E. Giacoumidis, J. L. Wei, X. L. Yang, A. Tsokanos, and J. M. Tang, BAdaptive-modulation-enabled WDM impairment
reduction in multichannel optical OFDM transmission systems for next-generation PONs,[ IEEE Photon. J., vol. 2,
no. 2, pp. 130–140, Apr. 2010.
[5] T. Takagi et al., BDynamic routing and frequency slot assignment for elastic optical path networks that adopt distance
adaptive modulation,[ presented at the Opt. Fiber Commun. Conf., Los Angeles, CA, USA, 2011, Paper OTuI7.
[6] Y. Tang, K.-P. Ho, and W. Shieh, BCoherent optical OFDM transmitter design employing predistortion,[ IEEE Photon.
Technol. Lett., vol. 20, no. 11, pp. 954–956, Jun. 2008.
[7] D. J. Fernandes Barros and J. Kahn, BOptical modulator optimization for orthogonal frequency-division multiplexing,[
J. Lightw. Technol., vol. 27, no. 13, pp. 2370–2378, Jul. 2009.
[8] Y. Bao et al., BNonlinearity mitigation for high-speed optical OFDM transmitters using digital pre-distortion,[ Opt. Exp.,
vol. 21, no. 6, pp. 7354–7361, Mar. 2013.
[9] T. Dennis and P. A. Williams, BChirp characterization of external modulators with finite extinction ratio using linear
optical sampling,[ IEEE Photon. Technol. Lett., vol. 22, no. 9, pp. 646–648, May 2010.
[10] M. C. R. Medeiros, M. P. Thakur, P. Laurencio, and J. E. Mitchell, BTransmission limitations of WiMax over fibre
transmission employing optical up-conversion schemes,[ in Proc. 14th ICTON, 2012, pp. 1–4.
[11] T. Sugihara, T. Yoshida, and K. Ishida, BEffect of modulator bias Control in the presence of a finite extinction ratio in
DQPSK pre-equalization systems,[ IEEE Photon. Technol. Lett., vol. 24, no. 5, pp. 371–373, Mar. 2012.
[12] 40 Gbps LN Modulator: Fujitsu Optical Components Limited. [Online]. Available: http://jp.fujitsu.com/group/foc/en/
services/optical-devices/40gln/
[13] D. Fonseca, A. V. T. Cartaxo, and P. Monteiro, BModelling and experimental validation of an x-cut four phase
modulators structure,[ IEE Proc. Optoelectron., vol. 153, no. 4, pp. 145–151, Aug. 2006.
[14] W. S. Chang, RF Photonic Technology in Optical Fiber Links. New York, NY, USA: Cambridge Univ. Press, 2002.
[15] J. Lee et al., BNovel pilot scheme for transmitter IQ mismatch compensation in CO-OFDM system,[ IEEE Photon.
Technol. Lett., vol. 24, no. 17, pp. 1543–1545, Sep. 2012.
[16] J. Armstrong, BPeak-to-average power reduction for OFDM by repeated clipping and frequency domain filtering,[
Electron. Lett., vol. 38, no. 5, pp. 246–247, Feb. 28, 2002.
[17] A. S. Alaraimi and T. Hashimoto, BPAPR and OOBP of OFDM and their improvement by using self cancellation
codings,[ in Proc. 14th Int. Symp. WPMC, Oct. 3–7, 2011, pp. 1–5.
[18] S. G. Kang, J. G. Kim, and E. K. Joo, BA novel subblock partition scheme for partial transmit sequence OFDM,[ IEEE
Trans. Broadcast., vol. 45, no. 3, pp. 333–338, Sep. 1999.
[19] P. K. Frenger and N. Svensson, BParallel combinatory OFDM signaling,[ IEEE Trans. Commun., vol. 47, no. 4,
pp. 558–567, Apr. 1999.
[20] DAC I and DAC II. [Online]. Available: http://www.micram.com/index.php/products/vega
[21] 100 G Ultra Long Haul DWDM Framework Document. [Online]. Available: http://www.oiforum.com/public/documents/
OIF-FD- 100G-DWDM-01.0.pdf
[22] 100 G Forward Error Correction White Paper. [Online]. Available: http://www.oiforum.com/public/documents/
OIF_FEC_100G- 01.0.pdf
[23] B. Zhu et al., BSpace-, Wavelength-, Polarization-Division Multiplexed Transmission of 56-Tb/s over a 76.8-km Seven-
Core Fiber,[ presented at the Opt. Fiber Commun. Conf., Los Angeles, CA, USA, 2011, Paper PDPB7.
[24] Y. Gao, A. P. T. Lau, and C. Lu, BModulation-format-independent carrier phase estimation for square M-QAM systems,[
IEEE Photon. Technol. Lett., vol. 25, no. 11, pp. 1073–1076, Jun. 2013.
[25] D. Rafique et al., BFlex-grid optical networks: Spectrum allocation and nonlinear dynamics of super-channels,[ Opt. Exp.,
vol. 21, no. 26, pp. 32184–32191, 2013.
[26] H. Cheng et al., BPilot-symbols-aided cycle slip mitigation for DP-16QAM optical communication systems,[ Opt. Exp.,
vol. 21, no. 19, pp. 22166–22172, Sep. 2013.

Vol. 6, No. 3, June 2014 7200207