We performed long-haul WDM transmission experiments to compare 10 Gbit/s MSK and QPSK modulation with a channel grid of 12.5 GHz. A standard link setup with inline dispersion compensation was applied in combination with coherent detection and following offline signal processing. Both modulation formats showed nearly equal performance bridging about 4000 km at a BER of 10−3.
© 2012 OSA
To satisfy the demands for higher bandwidth of optical transmission systems, advanced optical modulation formats have been intensely investigated. For long haul transmission, the higher order modulation formats minimum shift keying (MSK) and differential quadrature phase shift keying (DQPSK) are aspirants to replace the conventional modulation formats differential binary phase shift keying (DBPSK) or on-off keying (OOK). They provide a doubled spectral efficiency at moderate additional complexity.
The authors experimentally demonstrated the performance of single channel optical minimum shift keying (MSK)  and compared MSK and DQPSK through numerical simulations . The results show that, with appropriate optical filtering, both modulation formats are applicable in a wavelength division multiplex (WDM) setup with a channel grid of 12.5 GHz at a data rate of 10 Gbit/s resulting in a symbol rate of 5 GBaud. This grid ensures the same spectral efficiency for the modulation formats in question, although the main lobe of MSK (15 GHz) is broader then the main lobe of DQPSK (10 GHz) as can be seen in Fig. 1 .
Because the channel spacing of 12.5 GHz can be generated through symmetrical splitting of ITU-T standard channels and optical interleavers are available on the market, 10 Gbit/s MSK or DQPSK may replace today’s 10 Gbit/s DBPSK format for long haul transmission links.
In the following, WDM experiments are presented for a comparison of MSK and QPSK. Unlike in previous investigations , we applied coherent detection instead of direct detection. This allows us to compare the performance of both modulation formats with the same receiver setup.
2. Experimental setup
The system setup used for this 5 GBaud QPSK and MSK experiment with WDM channel spacing of 12.5 GHz is illustrated in Fig. 2 . As laser sources, six external cavity lasers (ECL) with a linewidth of about 100 kHz spaced at 12.5 GHz were used. These six carriers were grouped in odd and even channels, individually modulated by two IQ-modulators (IQM) and then recombined by a 12.5 GHz interleaver.
The electrical driving signal is generated by an arbitrary waveform generator (AWG) operating at its maximum sample rate of 20 GSa/s. Due to the lack of a second AWG, the quadrature driving signal is generated out of the inphase signals, delayed by 11 symbols to ensure independent data, plus, in case of MSK an additional delay by half a symbol duration. The sinusoidal pulse shaping for MSK is done at 4 samples per symbol in the AWG. The driving signals for the second IQM are generated in the same manner out of the inverted data output of the AWG. These signals are delayed by approx. 1.25 ns to decorrelate the driving signals of the two individual IQMs. The electrical signals were amplified to Vpp ~5V, which is approx. Vπ of the IQMs. The IQMs were modulated symmetrically around the minimum power transmission point. Figure 3 shows the power transfer function of the interleaver on the left hand side and the electrical driving signals together with the optical WDM spectra at the transmitter output on the right hand side.
The transmission link is built as a recirculating loop comprising 80 km standard single mode fiber (SSMF, attenuation: 0.2 dB/km, chromatic dispersion (CD): 16.8 ps/(nm⋅km)) and a dispersion compensating fiber (DCF, 0.7 dB/km, CD: 103 ps/(nm⋅km)) for fully inline dispersion compensation. The launch powers into the fibers and the loop are controlled by optical attenuators. Optical filters with a 3-dB bandwidth of 3 nm are used to reduce the out-of-band noise produced by the Erbium-doped fiber amplifiers (EDFA). The launch powers were optimized for maximum transmission length. A polarization scrambler was applied, but not shown in Fig. 2.
The optical attenuator in front of the EDFA in the coherent receiver allows noise loading to set arbitrary OSNR values for back-to-back experiments. Another interleaver followed by a narrow optical channel filter is used to select the central channel of the even band for measuring. The optical front end is composed of a 90° hybrid and two balanced detectors. The transmitter ECL for the analyzed channel is used as local oscillator (LO) signal as well, however for decorrelation in a back-to-back setup 4 km SSMF is assembled between transmitter and receiver. Two polarization controllers (PC) are applied to maximize the optical power of the signal and the LO manually.
A digital storage oscilloscope with 16 GHz bandwidth digitizes the two photocurrents synchronously at 10 samples per symbol. The high oversampling was chosen to keep the opportunity to regain the optical signal most accurately in order to calculate, for example, the performance of direct detection receivers.
Data is recovered offline in a personal computer and the following signal processing blocks are applied: hybrid mean- and amplitude correction, low pass filtering, resampling to one (QPSK) respectively two (MSK) samples per symbol at the optimum sampling instant, phase estimation and correction  and data recovery. A frequency offset was no issue in this experiment, because the same ECL is utilized in the transmitter and receiver. The final error counting was averaged over 3 blocks of 10 million samples (3 million symbols), allowing a reliable bit error ratio (BER) measurement down to approx. 2⋅10−6.
3. Experimental results
Figure 4 shows the optical signal-to-noise ratio (OSNR) requirements of experiments and simulations in a back-to-back setup. While QPSK and MSK theoretically show equal performance , the required measured OSNR for MSK turned out to be slightly higher in the simulation as well as in the experimental investigation. This is due to the higher electrical bandwidth of MSK. Comparing the simulation results for the back-to-back case with the experimental performance, Fig. 4 shows approximately 2 dB implementation penalty for QPSK. While the implementation penalty is slightly higher for the MSK case, we observed nearly no additional OSNR penalty for the WDM case in comparison to a single channel setup for both modulation formats. The same system setup, including realistic interleaver filter functions, was applied for the simulations. The 2 dB penalty of the experimental results is due to imperfect electrical signal generation and bandwidth limitations.
Figure 5 shows the BER vs. transmission distance of 5 GBaud QPSK and MSK with WDM and 12.5 GHz channel spacing for the applied link setup with inline dispersion compensation. MSK bridges approx. 500 km more than QPSK at the same BER. Since the back-to-back performance of MSK is about 1 dB worse compared to QPSK, it implies a higher tolerance towards nonlinear degradation effects. This might be due to the constant optical power, even during symbol transitions. The implementation of an additional algorithm in the signal processing of the receiver for compensating fiber nonlinearities, particularly self phase modulation (SPM), a phase correction proportional to the received signal intensity , leads to an improvement of the performance with respect to achievable reach, especially for QPSK (Fig. 5).
Reducing SPM by nonlinearity compensation with signal processing leads to a distance gain of approx. 380 km for QPSK and 260 km for MSK at a BER of 10−3 (Fig. 5). Reducing the WDM degradation effects cross phase modulation (XPM) and four wave mixing (FWM) by switching off the modulation of the neighboring channels (odd band) and thereby doubling the channel spacing results in a gain of approx. 420 km for QPSK and 300 km for MSK, also at a BER of 10−3. These measurement results show, that the distance loss caused by SPM and XPM/FWM respectively is within the same order of magnitude.
The launch power was optimized for QPSK and a transmission distance of 3200 km. The best BER performance was achieved at launch powers of −10 dBm for the SSMF and −13 dBm for the DCF per channel (Fig. 6 ). Similar optimum launch powers were found for MSK. In this case, the corresponding received OSNR is ~7.6 dB for both modulation formats. It has been checked, that these optimized launch powers do not change significantly for greater distances. These launch powers have been kept constant for all transmission distances. The somewhat low optimum launch powers in comparison to previous results  are due to the fact that the WDM scenario causes additional cross phase modulation (XPM) penalty compared to the single channel case.
Both modulation formats achieve appealing distances of 4000 km across the investigated link configuration at a BER of 10−3. The spectral efficiency is doubled (0.8 bit/(s⋅Hz)) in comparison to DBPSK. Furthermore, by implementing polarization multiplexing another doubling of the spectral efficiency is feasible.
The experimental results show, that MSK performs somewhat better than QPSK in a dispersion-compensated long haul transmission setup. With compensation of fiber nonlinearities they perform nearly alike. Obviously, implementation aspects play a major role for the decision, which of the both modulation formats is suitable to replace DBPSK.
Looking at the transmitter, the differential encoding for direct detection systems is less complex for MSK than for QPSK. It can be accomplished just as in a DBPSK system as a feedback XOR gate. A DQPSK transmitter needs parallel implementation with 10 gates . However, considering the effort to implement FEC, the additional effort for differential encoding is not relevant. On the other hand, the necessity of the sinusoidal weighting of the driving signals for MSK increases the complexity of the transmitter and the bandwidth requirements of the components. Shifting this procedure to the optical domain prevents the drawback of enhanced bandwidth requirements at the cost of an extra optical modulator, but lately a highly integrated MSK transmitter has been demonstrated using a so called quad Mach-Zehnder inphase quadrature modulator [7, 8].
The hardware effort of a coherent receiver is identical for both modulation formats. In direct detection systems a standard DBPSK receiver composed of a delay line interferometer (DLI) and a balanced detector (BD) can be applied for MSK. A DQPSK receiver requires twice this hardware. Due to the 50% broader spectrum in comparison to QPSK, MSK is more critical with respect to optical filtering or laser emission frequency maladjustments.
Overall, for direct detection systems, MSK has an advantage over QPSK because the compensation of nonlinearities on the receiver side is not possible and MSK performs better under these circumstances . Furthermore the receiver for MSK is easier to implement. For coherent detection, this advantage is reduced. The receivers are identical and the performance of QPSK gains through the mitigation of nonlinear phase distortion. Coherent optical transmission systems have the big advantage, that optical inline dispersion compensation is no longer necessary. The overall system noise is reduced because there is no more need for optical amplifiers to compensate for the attenuation of the DCF and longer distances are achievable.
We have successfully demonstrated long haul WDM transmission of optical QPSK and MSK at a data rate of 10 Gbit/s at a channel spacing of 12.5 GHz, resulting in a doubled spectral efficiency compared to 5 Gbit/s DBPSK. A distance of 4000 km over a fully dispersion-compensated EDFA-amplified link with 80 km SSMF transmission sections has been bridged using coherent detection and offline processing for both modulation formats, indicating a slight performance advantage for MSK at the same spectral efficiency. While MSK is easier to implement in a direct detection system, in a coherent detection setup the complexity of both formats is comparable.
This work was supported by the BMOD project of the German Federal Ministry of Education and Research (BMBF).
References and links
1. M. Rohde, R. Freund, C. Caspar, A. Hachmeister, and M. Gruner, “Long haul transmission of optical minimum shift keying format,“ in Proc. Eur. Conf. of Optical Comm, Sept. 2008, paper Mo.4.E.4.
2. A. Hachmeister, M. Rohde, and R. Freund, “Long haul transmission of optical minimum shift keying format with narrow channel spacing,” in Proc. Asia Comm. and Phot. Conf., Nov. 2009, paper ThT6.
3. M. Seimetz, “Digital phase estimation,” in High-Order Modulation for Optical Fiber Transmission (Springer, 2009), 99–111.
4. F. Xiong, Digital Modulation Techniques (Artech House, 2006).
5. K.-P. Ho and J. M. Kahn, “Electronic compensation technique to mitigate nonlinear phase noise,” J. Lightwave Technol. 22(3), 779–783 (2004). [CrossRef]
6. R. A. Griffin and A. C. Carter, “Optical differential quadrature phase shift key (oDQPSK) for high capacity optical transmission,” in Proc. Optical Fiber Comm. Conf., Mar. 2002, paper WX6.
7. T. Sakamoto, G. W. Lu, A. Chiba, T. Kawanishi, and T. Miyazaki, “Coherent demodulation of 10-Gb/s optical minimum shift keying,” in Proc. Opticl. Fiber Comm. Conf., Mar. 2010, paper JThA2.
8. G. W. Lu, T. Sakamoto, A. Chiba, T. Kawanishi, T. Miyazaki, K. Higuma, and J. Ichikawa, “80-Gb/s optical MSK generation using a monolithically integrated quad Mach-Zehnder IQ modulator,” in Proc. Optical Fiber Comm. Conf., Mar. 2010, paper OWN5.