## Abstract

We propose a new 16QAM optical transmitter based on a combination of a dual-drive Mach-Zehnder modulator (DD-MZM) and a dual-parallel Mach-Zehnder modulator (DP-MZM) with electrical binary drive signals. For 16QAM generation, two arms of DD-MZM are independently driven by two different binary data, and consequently, the DD-MZM produces an offset square 4QAM of the 16QAM constellation. This 4QAM signal is then switched over other quadrants through a typical QPSK modulation scheme by the following DP-MZM, and resulting in 16QAM. By using the proposed transmitter together with a digital coherent receiver, we successfully demonstrate the 224-Gb/s and 320-Gb/s PDM-RZ-16QAM systems. Their OSNR sensitivities at 3.8x10^{−3} BER are observed to be 19.8 dB and 23 dB, respectively.

©2012 Optical Society of America

## 1. Introduction

To boost both transmission capacity and the spectral efficiency of dense wavelength-division-multiplexing (DWDM) systems, 16-ary quadrature amplitude modulation (16QAM) format is a promising candidate [1–8]. For 16QAM generation, several types of the optical transmitter have been proposed and demonstrated [1–8]. The 16QAM generation methods can be categorized according to structure: i) a single optical modulator together with electrical multilevel drive signals and ii) a combination of multiple optical modulators with electrical binary drive signals. In [1–4], the 16QAM signal can be generated by using either a dual-parallel Mach-Zehnder modulator (DP-MZM) or a dual-drive Mach-Zehnder modulator (DD-MZM) driven by the electrical multilevel signals (which are produced by direct addition of two independent binary data or a high-speed digital-to-analog converter (DAC)). Although, however, the DD-MZM-based method can be used to theoretically generate 16QAM, it suffers significantly from large phase deviation and/or constellation distortion induced by non-ideal drive signals [4]. In addition, these generation methods require the sophisticated high-speed electronics such as DACs and linear drive amplifiers for electrical multilevel signals. On the other hand, the second generation schemes using multiple optical modulators with electrical binary drive signals are attractive since they have a potential for easily increasing the symbol rate due to the use of all binary inputs. In [5–7], several 16QAM systems have been demonstrated by using the highly integrated modulators (based on a hybrid integration technology of silica planer lightwave circuits and LiNbO_{3} phase modulators) or a cascade of two DP-MZMs or a combination of DP-MZM, MZM, and phase modulator in tandem. However, they have shown only the feasibility of the high-baud-rate 16QAM system (>25 Gbaud) and did not evaluate quantitatively its performance (e.g. bit-error rate (BER), Q-factor, optical signal-to-noise-ratio (OSNR) sensitivity, etc.). Furthermore, these methods require several feedback circuits to control the bias points and/or splitting/combining ratios of modulators and they lead to high complexity.

In this paper, we experimentally demonstrate the feasibility of the 16QAM transmitter based on a cascade of DD-MZM and DP-MZM with electrical binary drive signals [8, 9] and assess the performances of the 28-Gbaud and 40-Gbaud 16QAM signals with polarization-division-multiplexing (PDM) technique. DD-MZM has been widely used to modulate the optical signal due to its additional degree of freedom (i.e. it can be operated in push-push or push-pull mode according to the modulator’s drive condition). Here we utilize DD-MZM to make the desired four symbols (4QAM) of a square 16QAM by applying two independent data to each drive arm of DD-MZM. Afterwards, this 4QAM signal is switched over other quadrants via the quadrature phase shift keying (QPSK) modulation scheme by the second DP-MZM, thus yielding 16QAM. In addition, to effectively compensate for the signal distortion induced by the transmitter imperfections, we introduce a phase-folded decision-directed (PF-DD) linear equalizer at the receiver [10]. By using the proposed transmitter together with a digital coherent receiver, we realize the 224-Gb/s and 320-Gb/s PDM-16QAM systems and evaluate their BER performances with respect to OSNR. The back-to-back OSNR sensitivities at 3.8x10^{−3} BER of 224-Gb/s and 320-Gb/s PDM-RZ-16QAM signals are measured to be 19.8 and 23 dB, respectively.

## 2. Proposed transmitter

The configuration of the proposed transmitter is illustrated in Fig. 1
. The transmitter is comprised of DD-MZM and DP-MZM in tandem and each modulator is driven by electrical binary data (Data1~Data4) where *v _{1}*(

*t*)~

*v*(

_{4}*t*) denote the binary data taken values between 0.5 and −0.5 and

*A*~

_{1}*A*represent their amplitudes. First, DD-MZM modulates the amplitude and phase of the incident light with four symbol points, as shown in Fig. 2 . Since individual phase modulators of DD-MZM can operate independently when we apply two independent data (Data1 and Data2) to two arms of DD-MZM, we can manipulate the amplitude and phase of the optical signal by adjusting their driving voltages (

_{4}*A*and

_{1}*A*). If we assume that the splitting/combining ratios of DD-MZM are even, the transfer function

_{2}*H*(

*t*) of DD-MZM is expressed as [8, 9]

*V*and

_{π}*v*represent the switching voltage and bias voltage of DD-MZM, respectively. By using this mathematical relation, we can find the modulation conditions of DD-MZM to generate four symbol points of 16QAM. When

_{b}*A*=

_{1}*A*= 0.3

_{2}*V*and

_{π}*v*= 0.5

_{b}*V*, we can yield a square 4QAM placed in the 1st quadrant of the complex plane, as shown in Fig. 2(a). In other words, each phase modulator of DD-MZM modulates the phase of the incident light with 0.3π radians and two waves in both arms are then combined with quadrature phase. After QPSK modulation by the second DP-MZM driven with Data3 and Data4, the 16QAM signal is finally generated as shown in Figs. 2(a) and 2(c). Figures 2(c) and 2(d) show the modulated signal’s trajectory including all transitions between symbols. Furthermore, we can find another possible solution of DD-MZM operation to generate the 16QAM signal. If

_{π}*A*=

_{1}*A*= 0.7

_{2}*V*and

_{π}*v*= 0.5

_{b}*V*, DD-MZM modulates the optical signal with four alternate symbols of 16QAM, as illustrated in Fig. 2(b). In the same manner, through the QPSK modulation scheme, we can also produce a square 16QAM, as shown in Figs. 2(b) and 2(d). In addition, Figs. 2(e) and 2(f) represent the simulated results regarding the optical filtering effect when the 3-dB bandwidth of the 4th-order Gaussian shaped bandpass filter is ~1.4 times the symbol rate. As shown in the simulated results of Figs. 2(e) and 2(f), the performance of the second case can be more sensitive to bandwidth limitation due to its larger chirp than the first case. Thus, we select the first modulation condition of DD-MZM (i.e.

_{π}*A*=

_{1}*A*= 0.3

_{2}*V*and

_{π}*v*= 0.5

_{b}*V*) for the experimental verification of the proposed transmitter.

_{π}## 3. Experiments and results

Figure 3
shows the experimental setup to evaluate the performance of 224-Gb/s PDM-RZ-16QAM. To make four 28-Gbaud drive signals, four independent data of 7-Gb/s pseudo-random bit sequence (PRBS) with a pattern length of 2^{15}-1 from a pulse pattern generator (PPG) were multiplexed electrically and then each reshaped data by D-type flip flop (DFF) was applied to each modulator with a different electrical delay. First, an external cavity laser (ECL, linewidth < 200 kHz) operating at 193.4 THz was fed into a conventional MZM for 50% return-to-zero (RZ) pulse carving and then it was modulated with 112-Gb/s RZ-16QAM by using the proposed transmitter as described in previous section. RZ pulse shaping is beneficial to improve the tolerance to inter-symbol interference [11]. The modulated RZ-16QAM signal was split into two paths and then orthogonally recombined with an optical delay of several tens of symbol periods for polarization multiplexing. The −3-dB optical bandwidths of DD-MZM and DP-MZM used in our experiment were ~35 GHz and ~27 GHz, respectively. The optical eye diagrams of 112-Gb/s 16QAM before and after RZ pulse carving and 224-Gb/s PDM-RZ-16QAM and the optical spectrum of 224-Gb/s PDM-RZ-16QAM are found in Fig. 4
. By RZ pulse carving, we could obtain clearer 16QAM signal than non-return-to-zero (NRZ) pulse shape.

On the other hand, we implemented an intradyne digital coherent receiver, as shown in Fig. 3. A free-running ECL (linewidth = ~150 kHz) was utilized as a local oscillator (LO) and it was mixed with the modulated signal in a dual-polarization 90° optical hybrid. The eight outputs of the optical hybrid were detected and sampled by four balanced photo-detectors (BPDs) and analog-to-digital converters (ADCs). We utilized a digital storage oscilloscope as an ADC with the sampling rate of 80 GS/s and bandwidth of 33 GHz (It was achieved by interleaving two 40-GS/s channels.). The number of sampled data was around one million. After the sampling skew correction, the sampled data were processed offline as follows. First, we compensated for the in-phase/quadrature (IQ) imperfection of the receiver’s front-end by using the Gram-Schmidt orthogonal projection (GSOP) method and performed low pass filtering (LPF) to filter out unwanted noises. After resampling process of two samples per symbol, the half-symbol-spaced linear equalization and polarization demultiplexing were executed simultaneously. For pre-equalization, a 20-tap butterfly-structure equalizer adapted constant-modulus algorithm (CMA) was implemented and then switched to radius-directed equalization (RDE). We then implemented a standard 150-tap decision-directed linear equalizer followed by a decision-directed phase-locked loop (PLL) for frequency offset compensation and carrier phase estimation [12]. To mitigate the impacts of the signal distortion induced by the transmitter imperfections (e.g. amplitude mismatch of drive signals and error in bias voltage, etc.), the 350-tap PF-DD linear equalizer was applied [10]. As illustrated in Fig. 5 , the PF-DD linear equalizer first folded the phase of the input signal up by using the quadrant-based decision. Then, the symbols in each quadrant of the input signal were overlapped at the 1st quadrant. This phase-folded signal was equalized by a typical decision-directed linear equalizer with a new reference signal set in the 1st quadrant. After equalization, the overlapped phases were unfolded back. Using the PF-DD linear equalizer, we could effectively compensate for the signal distortion induced by transmitter imperfections [10]. Finally, BER was calculated after decision.

Figure 6
shows the measured BER curves of the generated 16QAM signals as a function of OSNR in 0.1 nm bandwidth. For OSNR change, we emulated numerically the amplified spontaneous emission (ASE) noises in the receiver. Since the proposed transmitter utilizes DD-MZM driven by two independent data, the generated 16QAM signal has inherently a frequency chirp, as shown in Fig. 2, and the signal’s performance can be affected by this chirp. However, RZ pulse carving is effective in removing the performance degradation induced by frequency chirp, as represented in NRZ and RZ results of Fig. 6 (triangles vs. circles). Figure 6 also shows the BER measurements and recovered constellation diagrams of the 224-Gb/s PDM-RZ-16QAM signal before and after PF-DD equalization (dashed: without PF-DD equalization, solid: with PF-DD equalization). Thanks to PF-DD equalization, the signal distortion was effectively compensated and before noise loading, the error-free results under the given number of samples were observed in both single-polarization and PDM cases. In Fig. 6, the black solid line indicates the theoretical limit of 28-Gbaud 16QAM in a single polarization case [13] and the square and circle symbols represent the measured BERs of 112-Gb/s RZ-16QAM and 224-Gb/s PDM-RZ-16QAM, respectively. The OSNR sensitivity (@ BER = 3.8x10^{−3} which corresponds to the threshold BER of advanced forward error correction (FEC) with 7% overhead [14]) of 112-Gb/s RZ-16QAM was measured at 16.7 dB and the difference from the theoretical limit was only ~1.1 dB. This OSNR sensitivity outperforms conventional DP-MZM-based methods reported in [2, 3] due to the use of RZ-pulse carving and PF-DD equalization. In addition, the OSNR penalty induced by PDM adaptation was almost negligible because the OSNR difference between single-polarization and PDM was 3.1 dB.

We also evaluated the feasibility of 40-Gbaud 16QAM generation. For this evaluation, we generated 320-Gb/s PDM-RZ-16QAM by using the same experimental setup of Fig. 3 except for the symbol rate and the clock frequency for RZ pulse carving. Figure 7
shows the optical eye diagrams of 160-Gb/s NRZ/RZ-16QAM and 320-Gb/s PDM-RZ-16QAM and the optical spectrum of 320-Gb/s PDM-RZ-16QAM. Although the optical carrier in the optical spectrum of 320-Gb/s PDM-RZ-16QAM remained due to the bandwidth limitation of modulators and drive amplifiers in the transmitter, we obtained quite clear eye diagrams by RZ pulse carving. We also measured the BER performances of the 160-Gb/s RZ-16QAM and 320-Gb/s PDM-RZ-16QAM signals. Figure 8
shows the measured BER curves (dashed: without PF-DD equalization, solid: with PF-DD equalization) and the recovered constellation diagrams of 320-Gb/s PDM-RZ-16QAM signal before and after PF-DD equalization. Before noise loading (OSNR > 35 dB), we observed the error floor at BER of ~4x10^{−4} in both single-polarization and PDM cases. At the FEC threshold BER of 3.8x10^{−3}, the OSNR sensitivities of 40-Gbaud 16QAM signals with single-polarization and PDM were measured at 19.8 dB and 23 dB, respectively. Even in the case of 40-Gbaud symbol rate with single polarization, the implementation penalty at 3.8x10^{−3} BER was observed to be only ~2.6 dB compared to its theoretical limit.

## 4. Conclusion

We have proposed and demonstrated an optical 16QAM transmitter consisted of DD-MZM and DP-MZM in tandem with electrical binary drive data. This technique has a potential for increasing the symbol rate of 16QAM due to the use of all binary inputs and its structure is much simpler than other previous schemes using binary drive data. Although this transmitter has a frequency chirp induced by DD-MZM, the results show that the impact of frequency chirp induced by DD-MZM is removed effectively by RZ pulse carving. In addition, by the PF-DD linear equalizer, the signal distortions induced by the transmitter imperfections is effectively compensated. We have experimentally verified the feasibility of the proposed transmitter by evaluating the performances of the 224-Gb/s and 320-Gb/s PDM-RZ-16QAM systems. The OSNR sensitivities of 28-Gbaud and 40-Gbaud PDM-RZ-16QAM signals (at the FEC threshold BER of 3.8x10^{−3}) are found to be 19.8 dB and 23 dB, respectively, which are 1.2 dB and 2.8 dB off the theoretical limits.

## References and links

**1. **P. J. Winzer, A. H. Gnauck, S. Chandrasekhar, S. Draving, J. Evangelista, and B. Zhu, “Generation and 1200-km transmission of 448-Gb/s ETDM 56-Gbaud PDM 16-QAM using a single I/Q modulator,” in *Proceedings of ECOC2010*, paper PD2.2 (2010).

**2. **A. H. Gnauck, P. J. Winzer, S. Chandrasekhar, X. Liu, B. Zhu, and D. W. Peckham, “Spectrally efficient long-haul WDM transmission using 224-Gb/s polarization-multiplexed 16-QAM,” J. Lightwave Technol. **29**(4), 373–377 (2011). [CrossRef]

**3. **M. S. Alfiad, M. Kuschnerov, S. L. Jansen, T. Wuth, D. van den Borne, and H. de Waardt, “11x224-Gb/s POLMUX-RZ-16QAM transmission over 670 km of SSMF with 50-GHz channel spacing,” IEEE Photon. Technol. Lett. **22**(15), 1150–1152 (2010). [CrossRef]

**4. **M. Nakamura, Y. Kamio, and T. Miyazaki, “Experimental demonstration of 16-QAM transmission with a single dual-drive Mach-Zehnder modulator,” in *Proceedings of OFC/NFOEC2011*, paper JThA42 (2011).

**5. **A. Sano, H. Masuda, T. Kobayashi, M. Fujiwara, K. Horikoshi, E. Yoshida, Y. Miyamoto, M. Matsui, M. Mizoguchi, H. Yamazaki, Y. Sakamaki, and H. Ishii, “69.1-Tb/s (432 x 171-Gb/s) C- and extended L-band transmission over 240 km using PDM-16-QAM modulation and digital coherent detection,” in *Proceedings of OFC/NFOEC2010*, paper PDPB7 (2010).

**6. **G.-W. Lu, M. Sköld, P. Johannisson, J. Zhao, M. Sjödin, H. Sunnerud, M. Westlund, A. Ellis, and P. A. Andrekson, “40-Gbaud 16-QAM transmitter using tandem IQ modulators with binary driving electronic signals,” Opt. Express **18**(22), 23062–23069 (2010). [CrossRef] [PubMed]

**7. **X. Zhou and J. Yu, “200-Gb/s PDM-16QAM generation using a new synthesizing method,” in *Proceedings of ECOC2009*, paper 10.3.5 (2009).

**8. **H. Y. Choi, T. Tsuritani, and I. Morita, “A novel transmitter for 320-Gb/s PDM-RZ-16QAM generation using electrical binary drive signals,” in *Proceedings of ECOC2012*, paper Tu.4.A.2 (2012).

**9. **H. Y. Choi, T. Tsuritani, and I. Morita, “BER-adaptive flexible-format transmitter for elastic optical networks,” Opt. Express **20**(17), 18652–18658 (2012). [CrossRef] [PubMed]

**10. **H. Y. Choi, T. Tsuritani, H. Takahashi, W. Peng, and I. Morita, “Generation and detection of 240-Gb/s PDM-64QAM using optical binary synthesizing approach and phase-folded decision-directed equalization,” Opt. Express (to be published).

**11. **H. Y. Choi, T. Tsuritani, and I. Morita, “Effects of LN modulator chirp on performance of digital coherent optical transmission system,” in *Proceedings of COIN2012*, paper TuF.2 (2012).

**12. **G. Picchi and G. Prati, “Blind equalization and carrier recovery using a “stop-and-go” decision-directed algorithm,” IEEE Trans. Commun. **35**(9), 877–887 (1987). [CrossRef]

**13. **J. G. Proakis, *Digital Communications*, 4th ed. (McGraw-Hill, New York, 2001).

**14. **ITU-T Recommendation G.975.1 2004, Appendix I.9.