Intensity-Modulation Full-Field Detection Optical Fast OFDM: Jian Zhao
Intensity-Modulation Full-Field Detection Optical Fast OFDM: Jian Zhao
Intensity-Modulation Full-Field Detection Optical Fast OFDM: Jian Zhao
demonstrated. However, coherent detection is expensive and is not suitable for cost-sensitive applications, such as 40/100/400G Ethernet, and access and short metro networks. On the other hand, direct-detection (DD) conventional OFDM has been extensively investigated. Both doublesideband (DSB) and single-sideband (SSB) DD OFDM systems have been proposed [2429]. DSB DD OFDM employing intensity modulation (IM) has the simplest implementation [2427]. However, it has significantly degraded CD tolerance. IM and DD were also applied in optical F-OFDM recently over either single-mode [810] or multimode fiber [1113] for cost-sensitive applications. SSB OFDM exhibits better performance than DSB OFDM; it is implemented either by inserting a spectral gap between the optical carrier and the signal [28] or by applying the information to the envelope of the SSB signal [29]. The former exhibits high performance approaching that using coherent detection, but increases the specification requirements in the sampling rate and component bandwidth. In certain cases, additional components, such as an electrical I-Q modulator, are needed. The latter scheme still exhibits reduced CD tolerance and requires an additional Hilbert transform. Furthermore, an I-Q optical modulator is required for the SSB systems unless an optical filter with sharp roll-off is used. Full-field detection (FD) [3036], which can extract the full optical field using an asymmetric MachZehnder interferometer (AMZI) and two photodiodes, is a promising scheme to balance the performance and the implementation complexity. This scheme was recently demonstrated in both onoff keying (OOK) [32] and offset differential quadrature phase-shift keying [33]. The recovered optical field allows subsequent near-ideal CD compensation. However, to the best of our knowledge, FD multicarrier systems (including both conventional OFDM and F-OFDM) have not been investigated yet. In this paper, we propose for the first time IM FD optical F-OFDM, and compare it with DD F-OFDM and FD conventional OFDM. We show that the proposed scheme, while avoiding the use of a coherent receiver, exhibits greatly improved CD tolerance over DD optical F-OFDM. It is also shown that optical F-OFDM using single-quadrature formats is more robust to the phase noise induced in full-field reconstruction, and so shows great performance advantages when compared to FD conventional OFDM at the same spectral efficiency. These advantages make the proposed scheme very promising for applications that demand low cost
2013 Optical Society of America
I. INTRODUCTION
ptical fast orthogonal frequency division multiplexing (F-OFDM) [114], with subcarrier spacing reduced to half of that in conventional OFDM, is a promising multicarrier scheme. The basic concept of F-OFDM was first proposed in wireless communications [1517] and has been studied thereafter [1823]. Recently, this concept was introduced to optical communications [1] and world-first experiments were performed to verify its implementation feasibility at both 1.55 [2] and 2 m optical wavelengths [3]. In coherent detection, optical F-OFDM exhibits greatly improved performance over conventional OFDM in frequency offset compensation [4]. In this scheme, the subcarrier multiplexing and demultiplexing are commonly implemented using a discrete cosine transform (DCT) pair. The designs of precise symbol synchronization [5] and guard interval (GI) [6] specific to optical F-OFDM were proposed. It was shown that, in contrast to conventional OFDM, a symmetric extension (SE) rather than cyclic-extension-based GI was required to enable chromatic dispersion (CD) compensation using one-tap equalizers. By using this design, transmission of coherent optical F-OFDM over 840 km single-mode fiber (SMF) was
Manuscript received December 18, 2012; revised February 13, 2013; accepted March 26, 2013; published April 18, 2013 (Doc. ID 181999). Jian Zhao is with Photonic Systems Group, Tyndall National Institute and Department of Physics, University College Cork, Lee Malting, Prospect Row, Cork, Ireland (e-mail: jian.zhao@tyndall.ie). http://dx.doi.org/10.1364/JOCN.5.000465
1943-0620/13/050465-10$15.00/0
Jian Zhao
but higher performance beyond the conventional IM DD solution, including long-reach 40/100/400G Ethernet (80 km), long-reach access networks (100 km), and short metro networks (300 km). The paper is organized as follows: we present the principle of the proposed scheme in Section II. Its feasibility is experimentally validated in Section III. In Section IV, extensive simulations are performed to study the performance limits and the advantages over other schemes. Finally, Section V summarizes the results.
In the case that t t t is small, sint t t t t t. By recovering the phase difference in t, the full optical field can be reconstructed by Z V full t V t V t12 exp j asinV f tt dt . (3) Ideally, V full t is perfectly obtained given a sufficiently small t. However, in practice, similar to the FD OOK format [32], low-frequency components of the noise accumulate during the integration, and so have to be suppressed by the use of a high-pass electrical filter. The recovered signal consists of amplitude and phase, so the F-OFDM decoding is similar to that in coherent detection. By using the SE- instead of CP-based GI, the optical F-OFDM subcarriers can be demultiplexed without intercarrier interference [6]. CD results only in different constants multiplied to different subcarriers, and so can be compensated using one-tap equalizers, as shown in Fig. 1.
II. PRINCIPLE
The decoding principle of the proposed scheme is illustrated in Fig. 1. The optical signal is first processed by an AMZI with 2 differential phase shift and t differential time delay (DTD). The value of t is chosen to balance the precision of phase reconstruction and the noise impact. The optimized t value also depends on the dispersion. The two outputs of the AMZI are detected by two photodiodes, sampled by analog-to-digital converters (ADCs). In practical implementation, these samples are interpolated, allowing a high resolution for the identification of the start-of-frame (SOF) symbol and the determination of the DCT window. Additionally, the 2 differential phase shift of the AMZI results in similar received power on the two photodiodes. Consequently, the retiming stage can be applied to the two received signals separately, and the relative timing delay between them can also be automatically recovered. In the case where the receivers are AC coupled, biases are added before the full-field reconstruction. Assuming that the baseband optical field is jEtj expjt, where Et and t are the optical field and its phase, respectively, V t and V t in Fig. 1 are V t jEtj2 jEt tj2 2jEtj jEt tj cost t t AMZI ; (1)
where AMZI is the differential phase of the AMZI 2. If the t value is not large, jEt tj jEtj. We can derive V f t V t V t sint t t: V t V t (2)
Jian Zhao
a peak-to-peak voltage of 3.6 V. This was equivalent to 0.65 V . The bias was set to be 1.4 V away from the null point of the modulator, unless otherwise stated. The modulated optical signal was then amplified and transmitted over a recirculating loop comprising 60 km SMF with 14 dB fiber loss. The noise figure of the loop amplifier was 5 dB and another 0.8 nm optical bandpass filter (OBPF) was used in the loop to suppress the amplified spontaneous emission noise. The launch power per span was around 5 dBm. At the receiver, the optical signal was detected with an optically preamplified receiver and a variable optical attenuator (VOA) was used to vary the input power to an erbium-doped fiber amplifier (EDFA). The preamplifier was followed by an OBPF with a 3 dB bandwidth of 0.64 nm, a second EDFA, and another optical filter with a 3 dB bandwidth of 0.8 nm. Then the optical signal passed through a Kylia AMZI with 20 ps DTD and 2 differential phase shift. The two outputs of the AMZI were detected by two 7 GHz photodiodes. Both detected signals were sampled by a real-time oscilloscope at 25 GSs. In off-line processing (implemented in Matlab), the signals were upsampled. An algorithm similar to that in [5] was used for automatic symbol synchronization. The algorithm consisted of two stages. The first stage identified the SOF symbol while the second one precisely determined the DCT window. This method was dispersion transparent, and the performance was comparable to that in coherent detection where the standard variation of timing errors was less than 2.2 ps for OSNRs as low as 3 dB [5]. Note that, different from coherent detection, the output power of the two photodiodes was approximately the same. Consequently, the algorithm could be applied to the two received signals separately, and could simultaneously remove any temporal
misalignment between the signals due to the imbalanced path lengths of the two receiver chains. Because of the AC-coupled receivers, bias was added to the two signals before they were added and subtracted for full-field reconstruction. The decoding algorithms after full-field reconstruction included DCT and channel equalization. Conventional time-domain averaging over multiple training symbols (TSs) [37] was used for channel estimation, and the number of the TSs was 20 unless otherwise stated. By using SE-based GI, F-OFDM symbols could be demultiplexed by DCT without intercarrier interference [6]. Consequently, similar to conventional OFDM, one-tap equalizers could be used to compensate the dispersion. Note that frequency-domain equalization as used in FD OOK could be used to avoid the GI and reduce the overhead. However, this method increased the implementation complexity of CD compensation when compared to one-tap equalizers, and so was not considered in this paper. For each experimental curve in the following investigation, three sets of data were extracted independently, with each consisting of 1200 measured optical F-OFDM symbols, giving a total number of measured bits of 108 1200 3 388;800. The results indicated that the bit-error-rate (BER) variation was not significant for these three sets of data, so the recirculating loop, as well as the symbol synchronization and channel estimation algorithms, was relatively stable during the measurements. Note that the use of a synchronous polarization scrambler would further stabilize the BER measurements. Figure 3 depicts the performance of FD optical F-OFDM after 240 (triangles), 360 (squares), and 480 km (diamonds), in comparison with the back-to-back case (circles). The insets of Fig. 3 illustrate the recovered constellation
-2
Log10(BER)
-3
-4
14
16
18
20
22
24
0 km
1 0 -1 -2 -2 -1
360 km
1 0 -1
480 km
-2 -2
-1
Jian Zhao
diagrams at around 21 dB OSNR. The investigated system required 18 dB OSNR for a BER of 103. We expect an improved back-to-back performance by using a narrower OBPF after the preamplifier (as investigated in the next section) and reduced PAPR with pre-coding. It is also shown that the penalties after 360 and 480 km were less than 1 and 2 dB, respectively. These penalties might have been caused by the imperfectly recovered phase during full-field reconstruction, and the insufficient GI length. At 0 km, the signal phase was close to zero, and so would not have a significant impact on the signal performance. However, as the fiber length increased (240360 km), the role of the reconstructed phase became important and the requirement for precise phase estimation was more stringent. In the experiment, imperfect bias value, t of the AMZI, thermal noise, and limited bandwidth of photodiodes could be potential sources to reduce the estimation precision. For 480 km, the CD value was larger, and the additional 1 dB penalty was likely induced by the insufficient GI length. Figure 4 gives the BER measurements for more fiber lengths. The OSNR values for 60, 120, 180, 240, 300, 360, 420, and 480 km are around 22.6, 22.3, 22, 21.6, 21.4, 21.1, 20.8, and 20.7 dB, respectively. As expected, the BER degraded gradually as the transmission distance increased, due to the degraded OSNR and transmission penalty. In particular, the insufficient GI length might result in the BER degradation for 420 and 480 km. However, a BER of 6 104 was still achieved after 480 km at 20.7 dB OSNR. In the experiment, a bias of 1.4 V with respect to the null point of the modulator and a 1.5 GHz high-pass filter after the integrator in phase reconstruction were used. The effect of this bias on the performance is shown in Fig. 5. In principle, a bias closer to the null point results in signal distortion arising from clipping after square-law detection, while a bias farther away from the null point would increase the DC component and the associated power consumption. It is observed from Fig. 5 that the performance was optimized for a bias range of 1.5 to 1 V for both 0 and 360 km. The optimal bias for 360 km was slightly closer to the null point when compared to the 0 km case due to the dispersed/broadened signal.
-3
-2
Log10(BER)
-3
-4
-3
-2.5
-2
-1.5
-1
-0.5
As described in Section II, a high-pass filter is required to suppress the low-frequency noise amplification. Figures 6(a) and 6(b) depict the spectra of the reconstructed V f and the output of the integrator (before the electrical high-pass filter). Statistically, V f should have very few low-frequency signal components, as can been seen in Fig. 6(a). The DC component in the figure might have been induced by the imperfect bias added to V t and V t. However, Fig. 6(b) shows that the low-frequency components of the reconstructed phase were very large. In fact, the integrator has a frequency transfer function of 12 jf , where f is the frequency. It is clear that the components with small f values dominate during integration, with a scaling factor of 1j2 f j. Therefore, any noise or inaccuracy in the low-frequency components would accumulate and eventually limit the system performance. Figure 7 depicts the performance versus the bandwidth of the high-pass electrical filter to suppress the lowfrequency components. The figure confirms the necessity of the high-pass filter. In the absence of the high-pass filter or when the filter bandwidth was not sufficiently large, the low-frequency components of the noise accumulated during integration, which severely degraded the performance. On the other hand, the filter also removed the low-frequency components of the signal, which resulted in signal distortion when the filter bandwidth was too large. The optimal filter bandwidths for both distances were 1.5 GHz. Figure 8 investigates the fiber nonlinearity effect when the transmission distance is 360 km. It can be seen that signal power (into the fiber) of 1 dBm resulted in similar performance as that of 5 dBm. It also suggests that, at 5 dBm as used in the experiment, the nonlinearity effect was negligible. When the input power was increased beyond 3 dBm, the performance started to be degraded prominently. In the experiment, the coefficients of the one-tap equalizers were obtained by time-domain averaging using TSs, the number of which should be sufficient to mitigate the noise effect. Figure 9 depicts the BER versus the number of TSs after 360 and 480 km. The performance was nearoptimal when the number of TSs was larger than five for both distances. It was equivalent to 35 ns adaptation time. In practice, additional channel tracking might also be needed after the initial training.
Log10(BER)
-4
Jian Zhao
20 0 -20 -40 -60 -20 -10 0 10 20
(a)
(b)
-10
10
20
Frequency (GHz)
Frequency (GHz)
Fig. 6. Spectra of (a) the recovered V f t and (b) its integration, V p t, versus the frequency at 480 km.
-2
-3
-4 0 1 2 3 4
using commercially available 2556 GSs ADCs. The IDCT/DCT used 128 points, of which 108 subcarriers (Nos. 3110) were used for data transmission. The sampling rate of the digital-to-analog converter (DAC)/ADC was 20 GSs. A 9 GHz third-order Gaussian-shaped lowpass filter was used after the DAC to remove the aliasing. The laser linewidth was 1 MHz, unless otherwise stated. The peak-to-peak voltage of the signal was 0.65 V . The fiber had a CD value of 16 psnmkm and a loss of 0.2 dBkm. The bandwidth of the third-order Gaussianshaped OBPF at the receiver was optimized and was found to be around 20 GHz. The DTD of the AMZI was 20 ps and the receiver electrical filter bandwidth was 10 GHz. Both 16.8 Gbits BPSK and 33.6 Gbits four amplitude phaseshift keying (4ASK) FD F-OFDM systems were simulated. For comparison, BPSK/4ASK DD optical F-OFDM and QPSK/16 quadrature amplitude modulation (16QAM) FD conventional OFDM were also simulated with the same parameters as FD F-OFDM. In the first case, the signal at the receiver was directly detected. In the digital domain, the square root was applied to the detected signal. The signal was then serial-to-parallel converted, demultiplexed by DCT, equalized by one-tap equalizers, and decoded. In the second case, as F-OFDM had a subcarrier spacing reduced to half of that in conventional OFDM, QPSK/16QAM was employed in the conventional OFDM to obtain the same spectral efficiency [1]. This can also be understood as follows: the 128-point DCT in F-OFDM could directly generate a real time-domain signal with 108 subcarriers carrying BPSK /4ASK data. On the other hand, in order
-2
Log10(BER)
-2
360 km 480 km
Log10(BER)
Log10(BER)
-3
-3
-4
-4
16
18
20
22
24
26
10
20
30
40
Jian Zhao
to produce a real time-domain signal for IM, the subcarriers in conventional OFDM had to be Hermitian symmetric. For a fixed number of samples (or time period) per symbol, the numbers of subcarriers for DC, data, and zeropadding in conventional OFDM were only 1, 54, and 9, respectively. Consequently, QPSK/16QAM should be used in conventional OFDM to obtain the same spectral efficiency as BPSK/4ASK optical F-OFDM. Figure 10 shows the required OSNR to achieve a BER of 103 versus transmission distance for (a) BPSK and (b) 4ASK F-OFDM with different GI lengths. The performance of DD optical F-OFDM with the GI length of 12 is also illustrated for comparison (circles). In Fig. 10(a), the curve for FD F-OFDM with system parameters the same as those in the experiment is also plotted. It can be seen from Fig. 10(a) that, with the experimental parameters, the simulated back-to-back receiver sensitivity at 103 was around 17 dB. This was close to the experimental result (18 dB). The additional 1 dB penalty might be due to the imperfect differential phase shift of the AMZI, the bias of the MZM, and timing error arising from symbol synchronization. In the simulation, the transmission reach at 2 dB penalty was 600 km, which was also close to that in the experiment (480 km). When the system parameters were optimized, 3.5 dB improvement in the back-to-back receiver sensitivity was observed in BPSK FD F-OFDM.
24
From Figs. 10(a) and 10(b), the required OSNRs at a BER of 103 for BPSK and 4ASK were around 13.5 and 21.5 dB, respectively. It is also seen that the supported transmission reach increased as the length of the GI increased. With the GI length of 12, the transmission reaches at 3 dB OSNR penalty were 600 and 480 km for BPSK and 4ASK, respectively. In contrast, DD optical F-OFDM exhibited limited dispersion tolerance. At 3 dB OSNR penalty, only 30 km could be supported for both 16.8 Gbits BPSK and 33.6 Gbits 4ASK. To further illustrate the advantage of the proposed scheme, Fig. 11 compares the performance of FD optical F-OFDM and FD conventional OFDM for (a) BPSK F-OFDM and QPSK OFDM and (b) 4ASK F-OFDM and 16QAM OFDM. Figures 12(a)12(d) depict the recovered constellation diagrams of (a) BPSK F-OFDM, (b) QPSK OFDM, (c) 4ASK F-OFDM, and (d) 16QAM OFDM at 0 km. From these figures, it can be clearly seen that, at the same spectral efficiency, F-OFDM outperformed conventional OFDM, and around 3.5 dB penalties were observed for FD conventional OFDM in Figs. 11(a) and 11(b). The performance advantage of F-OFDM in FD may be due to the fact that F-OFDM with single-quadrature formats is more robust to the phase noise induced in the full-field reconstruction. Note that this phase noise might not be slow-varying, so would introduce intercarrier interference as well as phase rotation.
24
20
20
16
16
(a)
12 0 200 400 600 800 1000
28
28
24
24
(b)
20 0 200 400 600
Jian Zhao
2 2
-1
-1
-2 -2
-1
-2 -2
-1
-2
-2
-4 -4
-2
-4 -4
-2
Constellation diagrams of FD F-OFDM and conventional OFDM at 0 km. (a) and (b) 20 dB OSNR. (c) and (d) 25 dB OSNR.
Figure 13 depicts the OSNR penalty versus the bandwidth of the high-pass electrical filter used for noise suppression. The OSNR penalty is with respect to the back-to-back receiver sensitivity of FD 4ASK F-OFDM at 103 . At back-to-back, the system performance was insensitive to the filter bandwidth provided that the low-frequency noise amplification was suppressed. This is because the received signal was real (i.e., the phase was close to zero) and any distortion on the reconstructed phase due to a wider filter bandwidth would not result in degraded performance. At 360 km, as the received signal was complex after dispersion, optimal filter bandwidth was required to balance the noise suppression and the distortion arising from the filtered signal phase components.
10
The optimal filter bandwidth was around 11.5 GHz, matching that in the experiment (Fig. 7). It is also seen that 16QAM conventional OFDM was more sensitive to the filter bandwidth, especially when the low-frequency noise amplification dominated (low filter bandwidth region). This confirms the benefit of optical F-OFDM over conventional OFDM when FD is applied. In contrast to coherent detection, the proposed scheme does not require a high-specification laser. Figure 14 shows the OSNR penalty versus the laser linewidth. As expected, FD was insensitive to the laser linewidth. Lasers with linewidth of less than 2 MHz resulted in negligible penalty, indicating the possible use of a cost-effective laser.
10
8 6 4 2 0 -2 -1 10
10
10
Jian Zhao
3 2
1 0
-1
10
20
30
40
50
10
20
30
40
DGD (ps)
Fig. 17. OSNR penalty versus DGD for FD 4ASK F-OFDM.
We have also investigated the flexibility to choose the AMZI DTD. Figure 15 illustrates that the optimal range of the AMZI DTD is large (20 ps for both 0 and 360 km). In principle [Eqs. (2) and (3)], a smaller DTD can provide more accurate estimation. However, when the DTD is small, the value of V f is small such that the effect of thermal noise increases. The optimal DTD should balance the precision of phase reconstruction and the noise impact. Additionally, it is shown from Fig. 15 that the optimal value also depended on the fiber length. In the presence of CD, the signal was broadened, so the optimal DTD increased accordingly. Figure 16 further investigates the performance sensitivity to the AMZI differential phase. Ideally, the required AMZI phase should be 2. However, this phase cannot be perfectly obtained in practice and may also drift over time. The figure shows that, at 2 dB OSNR penalty, the tolerance range was as wide as 35 35 and 45 45 for 0 and 360 km, respectively. This requirement can be readily achieved by commercial optical devices. The principle of FD [Eqs. (1)(3)] is based on single polarization without polarization mode dispersion (PMD), so the PMD tolerance of the proposed scheme is investigated in Fig. 17. The figure indicates that, at 1 dB OSNR penalty with respect to 0 ps differential group delay (DGD), the scheme could tolerate 20 ps DGD at both 0 and 360 km
8
for 33.6 Gbits 4-ASK F-OFDM. When rescaled to 56 Gbits, the DGD tolerance would still be larger than 10 ps. This is sufficient to most applications that the proposed scheme targets, including long-reach 100/400GE, long-reach access networks (100 km), and short metro networks (300 km). Note that the proposed scheme may possibly be extended to support two polarizations together with enhanced PMD tolerance, based on recently reported work for polarization multiplexed FD QPSK [36]. Finally, although all the experimental and numerical results are based on the single-channel investigation, we believe that the proposed scheme can be readily applied to wavelength division multiplexing configurations. The conclusions should be still valid unless the channel spacing is close to the symbol rate per channel (commonly called optical superchannel). In that case, the channel orthogonality in the optical domain should be considered. However, it is beyond the scope of this paper.
V. CONCLUSION
We have proposed for the first time IM FD optical F-OFDM and investigated its performance advantages over DD optical F-OFDM and FD conventional OFDM. A 16.8 Gbits BPSK FD optical F-OFDM system over 480 km without optical dispersion compensation has been experimentally demonstrated to verify the implementation feasibility. BPSK/4ASK FD optical F-OFDM systems with optimized system parameters have been numerically investigated, and compared with BPSK/4ASK DD F-OFDM, and QPSK/16QAM FD conventional OFDM at the same spectral efficiency. It is shown that the proposed scheme, by recovering the full optical field without a coherent receiver, exhibits greatly improved CD tolerance over DD F-OFDM. This scheme is also more robust to the phase noise induced in full-field reconstruction, and so shows significant performance advantages over FD conventional OFDM. These advantages make the proposed scheme very promising for applications that demand low cost but higher performance beyond the conventional IM DD solution, such as high-speed long-reach 40/100/400G Ethernet, long-reach access networks, and short metro networks.
0 30 60 90 120 150
Jian Zhao
ACKNOWLEDGMENTS
This work was supported by Science Foundation Ireland under grant nos. 11/SIRG/I2124 and 06/IN/I969 and the EU Seventh Framework Program under grant agreement 318415 (FOX-C).
of the interference between odd and even terms in optical fast OFDM scheme based on interleaved multiplexing, IEEE Photon. Technol. Lett., vol. 24, pp. 11601162, 2012. [15] M. R. D. Rodrigues and I. Darwazeh, Fast OFDM: a proposal for doubling the data rate of OFDM schemes, in Int. Conf. Telecommunications, 2002, pp. 484487. [16] F. Xiong, M-ary amplitude shift keying OFDM system, IEEE Trans. Commun., vol. 51, pp. 16381642, 2003. [17] G. Mandyam, On the discrete cosine transform and OFDM systems, in Int. Conf. Acoustics, Speech, and Signal Processing, 2003, pp. 544547. [18] Y. Yeh and S. Chen, Efficient channel estimation based on discrete cosine transform, in Int. Conf. Acoustics, Speech, and Signal Processing, 2003, pp. 676679. [19] P. Tan and N. C. Beaulieu, A comparison of DCT-based OFDM and DFT-based OFDM in frequency offset and fading channels, IEEE Trans. Commun., vol. 54, pp. 21132125, 2006. [20] N. Al-Dhahir, H. Minn, and S. Satish, Optimum DCT-based multicarrier transceivers for frequency-selective channels, IEEE Trans. Commun., vol. 54, pp. 911921, 2006. [21] F. Gao, T. Cui, A. Nallanathan, and C. Tellambura, Maximum likelihood based estimation of frequency and phase offset in DCT OFDM systems under non-circular transmissions: algorithms, analysis, and comparisons, IEEE Trans. Commun., vol. 56, pp. 14251429, 2008. [22] R. G. Clegg, S. Isam, I. Kanaras, and I. Darwazeh, A practical system for improved efficiency in frequency division multiplexed wireless networks, IET Commun., vol. 6, pp. 449457, 2012. [23] P. N. Whatmough, M. R. Perrett, S. Isam, and I. Darwazeh, VLSI architecture for a reconfigurable spectrally efficient FDM baseband transmitter, IEEE Trans. Circuits Syst., vol. 59, pp. 11071118, 2012. [24] J. L. Wei, X. Q. Jin, and J. M. Tang, Influence of directly modulated DFB lasers on the transmission performance of carrier-suppressed single-sideband optical OFDM signals over IMDD SMF systems, J. Lightwave Technol., vol. 27, pp. 24122419, 2009. [25] J. Xiao, J. Yu, X. Li, Q. Tang, H. Chen, F. Li, Z. Cao, and L. Chen, Hadamard transform combined with companding transform techniques for PAPR reduction in an optical direct-detection OFDM system, J. Opt. Commun. Netw., vol. 4, pp. 709714, 2012. [26] L. Chen, Y. Qiao, Y. Zhao, and Y. Ji, Wide range frequency offset estimation method for a DD-OFDM-PON downstream system, J. Opt. Commun. Netw., vol. 4, pp. 565570, 2012. [27] E. Giacoumidis, A. Kavatzikidis, A. Tsokanos, J. M. Tang, and I. Tomkos, Adaptive loading algorithms for IMDD optical OFDM PON systems using directly modulated lasers, J. Opt. Commun. Netw., vol. 4, pp. 769778, 2012. [28] B. J. C. Schmidt, A. J. Lowery, and J. Armstrong, Experimental demonstrations of electronic dispersion compensation for long-haul transmission using direct-detection optical OFDM, J. Lightwave Technol., vol. 26, pp. 196203, 2008. [29] M. Schuster, S. Randel, C. A. Bunge, S. C. J. Lee, F. Breyer, B. Spinnler, and K. Petermann, Spectrally efficient compatible single-sideband modulation for OFDM transmission with direct detection, IEEE Photon. Technol. Lett., vol. 20, pp. 670672, 2008. [30] A. D. Ellis and M. E. McCarthy, Receiver-side electronic dispersion compensation using passive optical field detection
REFERENCES
[1] J. Zhao and A. D. Ellis, A novel optical fast OFDM with reduced channel spacing equal to half of the symbol rate per carrier, in Optical Fiber Communication Conf., 2010, paper OMR1. [2] S. K. Ibrahim, J. Zhao, D. Rafique, J. ODowd, and A. D. Ellis, Demonstration of world-first experimental optical fast OFDM system at 7.174 Gbits and 14.348 Gbits, in European Conf. Optical Communication, 2010, paper PDP3.4. [3] N. MacSuibhne, Z. Li, B. Baeuerle, J. Zhao, J. P. Wooler, S. U. Alam, F. Poletti, M. N. Petrovich, A. M. Heidt, I. P. Giles, D. J. Giles, B. Palsdottir, L. Gruner-Nielsen, R. Phelan, J. OCarroll, B. Kelly, D. Murphy, A. D. Ellis, D. J. Richardson, and F. C. G. Gunning, Wavelength division multiplexing at 2 m, in European Conf. Optical Communication, 2012, post-deadline paper Th.3.A.3. [4] J. Zhao and A. D. Ellis, Advantage of optical fast OFDM over OFDM in residual frequency offset compensation, IEEE Photon. Technol. Lett., vol. 24, pp. 22842287, 2012. [5] J. Zhao, S. K. Ibrahim, D. Rafique, P. Gunning, and A. D. Ellis, Symbol synchronization exploiting the symmetric property in optical fast OFDM, IEEE Photon. Technol. Lett., vol. 23, pp. 594596, 2011. [6] J. Zhao and A. D. Ellis, Transmission of 4-ASK optical fast OFDM with chromatic dispersion compensation, IEEE Photon. Technol. Lett., vol. 24, pp. 3436, 2012. [7] J. Zhao and H. Shams, Fast dispersion estimation in coherent optical 16QAM fast OFDM systems, Opt. Express, vol. 21, pp. 25002505, 2013. [8] C. Lei, H. Chen, M. Chen, and S. Xie, A high spectral efficiency optical OFDM scheme based on interleaved multiplexing, Opt. Express, vol. 18, pp. 2614926154, 2010. [9] C. Lei, H. Chen, M. Chen, and S. Xie, 16 10 Gbs symmetric WDM-FOFDM-PON realization with colorless ONUs, Opt. Express, vol. 19, pp. 1527515280, 2011. [10] E. Giacoumidis, S. K. Ibrahim, J. Zhao, J. M. Tang, A. D. Ellis, and I. Tomkos, Experimental demonstration of cost-effective intensity-modulation and direct-detection optical fast-OFDM over 40 km SMF transmission, in Proc. Optical Fiber Communication Conf. and Expo., 2012, paper JW2A.65. [11] E. Giacoumidis, I. Tomkos, and J. M. Tang, Performance of optical fast-OFDM in MMF-based links, in Optical Fiber Communication Conf., San Diego, CA, 2011, paper OWU3. [12] E. Giacoumidis, S. K. Ibrahim, J. Zhao, J. M. Tang, A. D. Ellis, and I. Tomkos, Experimental and theoretical investigations of intensity-modulation and direct-detection optical fastOFDM over MMF-links, IEEE Photon. Technol. Lett., vol. 24, pp. 5254, 2012. [13] E. Giacoumidis, A. Tsokanos, C. Mouchos, G. Zardas, C. Alves, J. L. Wei, J. M. Tang, C. Gosset, Y. Jaouen, and I. Tomkos, Extensive comparison of optical fast OFDM and conventional OFDM for local and access networks, J. Opt. Commun. Netw., vol. 4, pp. 724733, 2012. [14] W. Long, J. Zhang, D. Wang, J. Han, S. Chen, A. M. Han, G. Gao, H. Leng, L. Liu, W. Zhu, Y. Zhao, and W. Gu, Mitigation
474 J. OPT. COMMUN. NETW./VOL. 5, NO. 5/MAY 2013 for low cost 10 Gbits 600 km-reach applications, in Proc. Optical Fiber Communication Conf., 2006, paper OTuE4. [31] N. Kikuchi and S. Sasaki, Highly sensitive optical multilevel transmission of arbitrary quadrature-amplitude modulation signals with direct detection, J. Lightwave Technol., vol. 28, pp. 123130, 2010. [32] J. Zhao, M. E. McCarthy, and A. D. Ellis, Electronic dispersion compensation using full optical field reconstruction in 10 Gbits OOK based systems, Opt. Express, vol. 16, pp. 1535315365, 2008. [33] J. Zhao and A. D. Ellis, Full-field detection based multi-chip MLSE for offset-DQPSK modulation format, in European Conf. Optical Communication, 2011, paper Tu.5.A.2. [34] X. Liu, S. Chandrasekhar, and A. Leven, Digital selfcoherent detection, Opt. Express, vol. 16, pp. 792803, 2008. [35] I. Tselniker, M. Nazarathy, S. B. Ezra, J. Li, and J. Leuthold, Self-coherent complex field reconstruction with in-phase and quadrature delay detection without a direct-detection branch, Opt. Express, vol. 20, pp. 1545215473, 2012. [36] J. Li, R. Schmogrow, D. Hillerkuss, P. C. Schindler, M. Nazerathy, C. S. Langhorst, S. B. Ezra, I. Tselniker, C. Koos, W. Freude, and J. Leuthold, A self-coherent receiver for detection of PolMUX coherent signals, Opt. Express, vol. 20, pp. 2141321433, 2012. [37] B. Inan, S. Adhikari, O. Karakaya, P. Kainzmaier, M. Mocker, H. V. Kirchbauer, N. Hanik, and S. L. Jansen, Real-time
Jian Zhao 93.8 Gbs polarization-multiplexed OFDM transmitter with 1024-point IFFT, Opt. Express, vol. 19, pp. B64B68, 2011.
Jian Zhao (M08) received a B.Eng. degree from the University of Science and Technology of China in 2002, and M.Phil. and Ph.D. degrees from the Chinese University of Hong Kong in 2004 and 2007, respectively. He joined the Photonic Systems Group at the Tyndall National Institute, Cork, Ireland, as a Postdoctoral Researcher in August 2007, and is currently a Staff Researcher and Research Fellow. His current research interests include optical OFDM, advanced modulation and detection schemes, communication and information theory, and digital signal processing. He proposed, and together with his colleagues, demonstrated the world-first optical F-OFDM at 1550 and 2000 nm optical wavelengths. He led a team that developed a 10 G real-time electronicdispersion-compensation integrated receiver prototype supporting 900 km transmission and was the runner-up of the Alcatel-Lucent (UK & Ireland) Innovation Competition. Since 2009, he has been a Principal Investigator or a Partner with more than 1M funding support from Enterprise Ireland, Science Foundation Ireland, and EU FP7. He has published more than 75 papers (around 50 of which are senior authored) in peer-reviewed journals and conferences, and two book chapters. He serves as a reviewer for various journals (Journal of Lightwave Technology, IEEE Photonics Technology Letters, IEEE Transactions on Vehicular Technology, Optics Express, Optics Letters, Electronics Letters, etc.), and as a session cochair or TPC member for several IEEE/IET international conferences. He is a member of IEEE.