1. Introduction

Multi-level modulation/demodulation techniques are expected to effectively accommodate the endlessly growing internet traffic on photonic networks by saving lambda and bandwidth resources, just like in wireless communication systems. M-ary quadrature amplitude modulation (M-QAM) is an important technique for attaining higher than four-bits-per-symbol modulation [1-4]. In M-QAM optical multi-level transmission, however, there are several technical issues to be overcome. First, the spectral linewidth requirements of the light source are extremely stringent. For example, for 16-QAM detection in optical phase-locked loop homodyne detection even with zero loop delay, a linewidth-to-bit-rate ratio of 1.4×10^-6 (14 kHz linewidth for 10-Gbit/s 16-QAM) is theoretically required [1]. Even when using digital carrier phase estimation, a 16-QAM system must attain a ratio of 3.0×10^-6 (30 kHz linewidth for 10 Gbit/s) [5]. For instance, 64-QAM and 128-QAM at 1-Gsymbol/s transmission have been experimentally demonstrated using a frequency-stabilized light source with 4-kHz linewidth [4]. Second, suppression of inter-symbol interference (ISI) in multi-level modulation should be seriously taken into account [1,6]. Specially-designed modulators, which can work in binary mode to suppress ISI, for 16-QAM with a modulation speed of more than 10 Gsymbol/s have been reported [7,8]. However, error-free performance has not yet been achieved. Finally, all of these reported M-QAM experiments have been demonstrated only in off-line mode; therefore, the constellation and bit-error-rate (BER) were not measured simultaneously in real time, and the obtained BER performance was limited to larger than 10^-5.

In this paper, we report, for the first time to the best of our knowledge, the demonstration of 2.5-Gsymbol/s (10-Gbit/s) 16-QAM transmission over 120 km of standard single-mode fiber (SSMF) without offline processing after detection and without any dispersion compensation, achieving a floor-less BER performance of less than 10^-9 using a distributed-feedback laser diode (DFB-LD) with a linewidth of 30 MHz. We employed an optical phase-noise canceling technique based on a polarization-multiplexed pilot-carrier to take advantage of the linewidth-tolerant characteristics of the scheme [9]. Some other phase-noise insensitive coherent systems in a binary modulation scheme were studied in [10]. The 16-QAM signal was modulated by a simple I-Q vector modulator with a branch for the pilot-carrier supply. ISI mainly caused in the modulator and driver electric amplifiers was compensated by applying a standard electronic pre-equalization (pre-EQ) technique, which is also applicable to the compensation of dispersion and optical nonlinearity [11,12].

2. Experimental setup

Figure 1(a) shows the experimental setup for the 10-Gbit/s 16-QAM transmission. A DFB-LD with an emission wavelength of 1551 nm and a linewidth of 30 MHz was employed as a light source. We used a pilot-carrier vector modulator, where only the TM polarization component was modulated through an I-Q vector modulator for 16-QAM, and the TE component served to generate a polarization-multiplexed pilot-carrier for optical phase-noise cancellation in a self-homodyne receiver [9]. Though the pilot-carrier can be used to carry independent information as experimentally demonstrated in [9], it was kept unmodulated for simplicity in this experiment. For this purpose, the modulator consisted of polarization beam splitters (PBSs), a dual-electrode I-Q vector modulator in the TM path, and a TE path for the pilot-carrier, as shown in Fig. 1(b). All of these elements were monolithically integrated in a LiNbO₃ waveguide. The optical insertion loss and half-wave voltage of the modulator were 11 dB and 3 V, respectively. To attain best performance, the optical power of the pilot-carrier was adjusted to be about 50 % of the total modulator output power, by a first polarization controller (PC1) [13]. Two sets of complementary four-level electrical 2.5-Gsymbol/s 16-QAM signals, ±Data-I and ±Data-Q (−Data-I,Q is complementary logic to +Data-I,Q), were applied to the modulator from an arbitrary waveform generator (AWG). 16-QAM patterns for Data-I and Data-Q were generated from two sets of two binary 2⁹-1 pseudo-random bit sequence (PRBS) patterns having 40-bit mutual delay to obtain de-correlation. Here, electronic pre-EQ by digital equalization with a 16-tap transversal filter was applied to Data-I and Data-Q independently to suppress ISI distortion caused by non-ideal frequency characteristics of the modulator and electric devices. The gain coefficients of the 16 taps were determined by a least mean squares (LMS) algorithm. The 120-km fiber transmission line consisted of two 60-km lengths of SSMF without any dispersion compensation (17 ps/nm/km), with a coupled optical power of +6 dBm. The polarization mode dispersion (PMD) coefficient of the transmission fiber was 0.05 ps/km^0.5. At the receiver side, an optical preamplifier was followed by an optical band pass filter (OBPF) with 1-nm pass bandwidth, and the signal was divided by a 3-dB coupler into in-phase (I) and quadrature (Q) arms, each including manual polarization controllers (PC2, PC3), PBSs, and balanced photo-detectors (PDs) with RF pre-amplifiers for self-homodyne detection. PC2 and PC3 were used not only to control the polarizations, but also to make π/2 phase difference between I and Q arms. PBSs work to coherently mix the two polarization components of optical signal and pilot-carrier in the same manner as common homodyne does [14]. The symbol-error-rate (SER) at 2.5 Gsymbol/s was measured with an error detector (ED) in real time by adjusting the threshold levels, Th_a, Th_b, and Th_c, one by one, for each I- and Q-component, as shown in Fig. 2. The total SER of the 16-QAM transmission can be obtained by summing the measured error rates at all threshold levels of the I- and Q-components. Assuming Gray-coding, the BER is one-fourth of the total SER because 16-QAM is four-bits-per-symbol modulation. We simultaneously observed the constellation and eye-diagram using an oscilloscope.

 

Fig. 1. (a) Experimental setup for 10-Gbit/s (2.5-Gsymbol/s) 16-QAM self-homodyne transmission and (b) pilot-carrier vector modulator.

Download Full Size | PDF

 

Fig. 2. 16-QAM constellation and threshold levels to characterize the SER performance.

Download Full Size | PDF

 

Fig. 3. Constellation and eye-diagrams of the received 10-Gbit/s 16-QAM signals using EC-LD (a), DFB-LD (b), and DFB-LD with 120-km SSMF transmission (c).

Download Full Size | PDF

3. Results and discussion

First, for comparison, we conducted an experiment using an external-cavity laser diode (EC-LD), having a linewidth of 200 kHz, as the light source in a back-to-back (BtB) configuration without the SSMF. The observed constellation and eye-diagrams of the I- and Q-components for the EC-LD setup are shown in Fig. 3(a). Clear signal-point spacing and four-level eye-opening were observed, thanks to the suppression of ISI by the pre-EQ. The received optical power was set at −20 dBm (the received optical power includes the power of the pilot-carrier). The eye-opening of the Q-component was degraded by about 20 % as compared with that of the I-component. We believe this was caused chiefly by un-balancing of the sub-Mach-Zehnder for the Q-component in the vector modulator. Figure 3(b) shows the observed constellation and eye-diagrams of the self-homodyne detection in the BtB configuration using the 30-MHz DFB-LD. It should be noted that clear signal-point spacing and eye-opening were observed in spite of the large phase noise of the DFB-LD; this is thanks to the phase-noise cancelling capability of the pilot-carrier scheme. Figure 3(c) shows the constellation and eye-diagrams after transmission through the 120-km SSMF. The eye-diagrams were distorted slightly, mainly by the accumulated dispersion of the SSMF. Nevertheless, a clear eye-opening was obtained due to the bandwidth efficient 16-QAM format at 2.5 Gsymbol/s. Figure 4 shows the optical spectrum (0.01-nm resolution) of the 10-Gbit/s 16-QAM optical signal measured at the output of the modulator. We observed the spectral-efficient spectrum with a 20-dB-down full bandwidth of only 5 GHz, achieving the spectral efficiency of 2 bit/s/Hz. Higher spectral efficiency for multi-carrier transmission can be achieved by polarization-division multiplexing where the polarization-multiplexed pilot-carrier is modulated. Furthermore, tight optical filtering supported by forward-error correction (FEC) will improve the spectral efficiency. Figure 5 shows the SER measured at each threshold level explained in Fig. 2 above. The SER performance using the EC-LD in the BtB configuration is shown in Fig. 5(a). Negligibly small SER-performance difference was observed for all threshold levels, Th_a to Th_c in each I- and Q-component, whereas the error-performance of the Q-component was worse than that of the I-component by less than 4 dB because of the smaller eye-opening explained above. Figure 5(b) shows the SER performance using the DFB-LD in the BtB configuration. We could not observe any error floor at all threshold levels in spite of the large phase-noise of the DFB-LD. However, we observed comparatively large error-performance degradation in the outer inter-symbol transition at the threshold levels Th_a and Th_c, which is apparent from the constellation in Fig. 3(b). The SER performance after the 120-km SSMF transmission is shown in Fig. 5(c); we did not observe any notable large penalty because of the clear eye-opening, as shown in Fig. 3(c). Finally, the BER performance of the 10-Gbit/s 16-QAM signals was obtained from the measured SERs and is plotted in Fig. 6. It should be noted that floor-less BER characteristics of less than 10^-9 were achieved in all cases. The power penalty for using the 30-MHz DFB-LD in the BtB configuration was less than 2 dB at a BER of 10^-9 compared with the performance using the 200-kHz EC-LD. Furthermore, the power penalty for the 120-km SSMF transmission without any dispersion compensation was less than 1 dB at a BER of 10^-9. In our previous report [13], the proposed scheme with quadrature phase-shift-keying (QPSK) data format showed dispersion and PMD tolerance which was comparable with that of differential QPSK at 10-Gsymbol/s. One different point from conventional schemes is that coherency between the signal and pilot-carrier affects the performance of the system. However, it is not a serious problem because PMD in a long-haul transmission line is usually smaller enough than coherent time of a DFB-LD even when the linewidth is as large as 30 MHz. Therefore, longer transmission is possible if the dispersion is compensated. The experimental results clearly show the advantage of the linewidth-tolerant characteristics of the pilot-carrier scheme, even in the case of 16-QAM.

 

Fig. 4. Optical spectrum for 10-Gbit/s 16-QAM optical signal using DFB-LD.

Download Full Size | PDF

 

Fig. 5. SER characteristics of the 10-Gbit/s 16-QAM signals using EC-LD (a), DFB-LD (b), and DFB-LD with 120-km SSMF transmission (c).

Download Full Size | PDF

 

Fig. 6. BER characteristics of the 10-Gbit/s 16-QAM signals using EC-LD, DFB-LD, and DFB-LD with 120-km SSMF transmission.

Download Full Size | PDF

4. Conclusion

We experimentally demonstrated 10-Gbit/s 16-QAM transmission using a DFB-LD with a linewidth of 30 MHz for the first time. Thanks to the combination of the phase-noise cancelling capability of the self-homodyne technique and a standard electronic pre-equalization technique, we confirmed clear 16-QAM constellations and error-free characteristics with a BER of less than 10^-9, as well as a receiver sensitivity penalty of less than 2 dB compared with the case using the EC-LD with 200-kHz linewidth. Furthermore, we also successfully demonstrated error-free transmission over 120-km SSMF without any dispersion compensation. We evaluated BER characteristics of the 10-Gbit/s 16-QAM signals from all six (three different thresholds for each of the I and Q components) SERs measured in real-time at 2.5 Gsymbol/s. Our proposed scheme will allow the use of inexpensive light sources and less-complex transmitters with fewer driver amplifiers for M-QAM multi-bit-per-symbol transmission systems.