Open Access
Add to BibSonomy
Abstract
Continuous real-time measurements are demonstrated from a 200Gb/s format configurable CFP transceiver that uses dual-polarization probabilistic-shaped 16QAM (DP-PS16QAM) modulation. Placed in a 50GHz coherent DWDM transmission system, DP-PS16QAM achieves a back-to-back 1.8dB OSNR gain over uniform DP-16QAM. It also transports over 1940km with EDFA-only amplification, thus doubling propagation distance of uniform DP-16QAM. Furthermore, a 1Tb/s super-channel consisting of five 200Gb/s DP-PS16QAM sub-carriers is placed in a 200GHz grid, and it achieves over 1600km transmission and 5b/s/Hz SE with a raw SE at 6.86b/s/Hz.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Coherent detection has enabled unprecedented growth in data rates of optical communication. With rapidly increasing demands for bandwidth, single carrier 200Gb/s for dense wavelength division multiplexing (DWDM) optical transmission systems are deployed commercially, and are gaining market share. Operators look for solutions that deliver the lowest cost per bit per Hz and per kilometer with minimum interruption on in-service network. Propagation distance, spectral efficiency (SE), and tolerance to optical filtering are three key requirements. However, the commonly adopted modulation formats for 200Gb/s transmission such as DP-16QAM, DP-8QAM, and DP-QPSK fall short at propagation distance (DP-16QAM) [1], at optical filtering tolerance (DP-8QAM), and at SE (DP-QPSK) [2].
Constellation shaping (CS), which can achieve a gain of up to 1.53 dB in signal-to-noise ratio (SNR) over additive white Gaussian noise (AWGN) channel [3,4], has attracted enormous attention in optical transmission recently. It has been studied and experimentally demonstrated for several rate-reach records using offline digital signal processing (DSP) [5–7]. Potentially, CS provides a solution for single carrier 200Gb/s that might meet operator’s three key requirements and hence, hardware supporting real-time CS is urgently demanded.
Aiming for long-haul single carrier 200Gb/s transmission that can be placed in 50GHz DWDM system, we selected and implemented a probabilistic-shaped 16QAM (PS16QAM) such that the source entropy is optimized to match the channel condition. The selected PS16QAM requires minimum modification on DSP algorithms and ASIC implementation architecture of uniform 16QAM, and transmits 3.5 information bits per symbol. To shape the constellation points, equal-probable information bits are passed through an efficient distribution matcher (DM) which is developed with similar operation principles as constant composition distribution matcher (CCDM) [8] that allows block to block parallel processing. Figure 1 shows the plot of the implemented DP-PS16QAM constellation.
In this paper, we report the first demonstration of real-time measured back-to-back and transmission performance gains obtained from CS and also long-term error-free performance of DP-PS16QAM transceivers. We transmit DP-PS16QAM over 1920km through 50GHz spaced DWDM link that consists 24 spans of G.652 at an 80km span length with EDFA-only for amplification, which doubles the propagation distance of uniform 16QAM. We also construct a 1Tb/s super-channel with 5x39.2GBaud 200Gb/s DP-PS16QAM sub-carriers, placed in a 200GHz grid with an achieved SE of 5b/s/Hz, and transmit it over 1600km. Considering the overhead from network, FEC, and DSP, the raw data rate for 1Tb/s super-channel is 6.86b/s/Hz.
2. Probabilistic-shaped 16QAM
In this section, we briefly review probabilistic constellation shaping and its theoretical gain and compare it to uniform 16QAM constellation. Over a linear AWGN channel, to minimize the average transmitted energy for a given rate, one can shape constellation points according to Maxwell-Boltzmann (M-B) distribution to achieve up to 1.53dB shaping gain. With the input power-dependent interference in nonlinear optical channel, even more gain can be achieved through a proper shaping design [9].
The three requirements for network operators in propagation distance, SE, and tolerance to optical filtering for long-haul single carrier 200Gb/s transmission in 50GHz DWDM system are considered here to select the information rate using probabilistic shaping. We used one-dimensional amplitude shaping with M-B distribution to design a PS16QAM constellation at 3.5 bits per complex QAM symbol for each polarization. The selected shaped constellation is depicted in Fig. 1.
The theoretical limits for linear regime are depicted in Fig. 2 for both uniformly distributed 16QAM and PS16QAM. Figure 2(a) shows the achievable information rates (AIR) for both constellations along with the Shannon limit. The generalized mutual information (GMI) is used for calculating the spectral efficiency versus SNR values. Assuming one-dimensional amplitude shaping for the base constellation in m2-QAM, and for a given FEC of rate Rc, the information rate of the probabilistic shaped constellation can be written as [10]
where 2H(PA) is the complex constellation entropy and m = 4 for 16QAM. Since we are comparing PS16QAM performance with uniform 16QAM assuming a similar FEC, we select the corresponding points for a 25% FEC overhead in Fig. 2(a). The information rates for both constellations and the corresponding gap are depicted here. The lower gap to Shannon limit demonstrates the potential gains from probabilistic shaped constellation according to the non-uniform probability. As it can be seen, without considering any implementation loss, there is a 2.1dB performance gain in terms of SNR by using the shaped constellation at different spectral efficiency with the FEC rate of 0.8 (overhead 25%).As mentioned before, the pure shaping gain is limited to 1.53dB. However, the comparison here is performed for two different spectral efficiencies which will results in different baudrates for the given data rate. Considering the impact of baudrate in optical communication systems, we use optical signal-to-noise ratio (OSNR) as the performance measure to compare the two constellations with different baudrates. Moreover, for a specific FEC overhead, one can employ normalized generalized mutual information (NGMI) [11] versus OSNR as a universal measure to compare the performance of constellations with different baudrates and spectral efficiencies. Figure 2(b) shows the NGMI values for different OSNRs for 200Gb/s DP-16QAM and DP-PS16QAM constellations. The threshold of 0.8 for 25% overhead FEC is also depicted where it shows that 1.5dB OSNR gain is theoretically achievable by using the DP-PS16QAM compared to uniform DP-16QAM.
3. Test-bed setup
In this section, we introduce the test-bed that we used for our measurements. The employed C form-factor pluggable (CFP) optical transceiver has multi-format flex-rate capability and it supports 100G DP-QPSK, 200Gb/s uniform DP-16QAM, and 200Gb/s DP-PS16QAM and 200Gb/s DP-8QAM. The following test-bed setup is used to carry out two sets of measurement:
- (I) Verification of real-time performance of probabilistic shaping using DP-PS16QAM and its propagation performance compared to uniform DP-16QAM and DP-8QAM;
- (II) Test and performance comparison for 1Tb/s super-channel propagation.
Test-bed configurations are shown in Fig. 3. The five CFP transceivers are configured such that they can support QPSK, 8QAM, 16QAM, and PS16QAM. Each transceiver is plugged into a line card as shown in Fig. 3(a). The optical link consists of 24 spans of G.652 optical fibre with a 1x2 optical switch placed at the end of every even number of span. The span loss is set to 23dB by attenuators connected at the end of each 80km fibre. Optical noise is loaded at the receiver end and OSNR at the receiver is measured with an OSA, The lowest required optical signal to noise ratio (ROSNR) at which the transceiver can still operate error-free is used as metric for performance measurement and comparison. Moreover, the carrier frequency of test channel TX3 is tuned to 193.578THz.
Fig. 3 Test-bed setup. (a) Line card hosting a CFP transceiver supporting DP-PS16QAM; (b) Optical spectrum with 40 channels (back-to-back); (c) Optical spectrum for a 1Tb/s super-channel and 39 loaded channels (back-to-back).
For DP-PS16QAM performance measurements in test (I), a total of 40 transceivers are placed on 50GHz grids as shown in Fig. 3(b) where TX1 is tuned to 193.478THz, TX 2 to 193.528THz, TX4 to 193.628THz, and TX 5 to 193.678THz.
For test (II), a 1Tb/s super-channel is formed as in Fig. 3(c) with TX1 tuned to 193.498THz, TX2 to 193.538THz, TX4 to 193.618THz, and TX5 to 193.658THz. The other 39 transceivers are configured to support 200Gb/s DP-16QAM.
4. 200Gb/s DP-PS16QAM results and analysis
The transceiver can be software configured to support uniform 16QAM and PS16QAM, as it provides a direct verification on the impact of probabilistic shaping. For measurements, pseudo-random bit sequence (PRBS) patterns with length of 231-1 has been used. The pre-FEC BER threshold is similar for both constellations where post-FEC BER is less than 10−15. The measured back-to-back ROSNRs are 15.8dB for DP-PS16QAM and 17.6dB for DP-16QAM. Therefore, the real-time transceiver achieved an OSNR gain of 1.8dB with the shaped constellation. This performance gap can be verified according to the theoretical limits discussed in Section 2 (potential 1.5dB OSNR gain as in Fig. 2), and also by considering the impact of hardware implementation loss on both constellations where the uniform DP-16QAM has 0.4-0.5dB higher loss on the measured line card compared to DP-PS16QAM.We successfully transmit the DP-PS16QAM through 1920km EDFA-only G.652 optical link (24 80-km spans that test-bed can support), with an extra 1dB available OSNR margin. For a direct comparison, we configured TX1 to TX 5 to DP-16QAM and were only able to propagate 960km at a 1dB OSNR margin. Propagation penalties, in terms of increased ROSNR, of 34GBaud/s DP-16QAM, 46GBaud/s DP-8QAM, and 39.2Gbaud/s DP-PS16QAM are measured by configuring the transceivers to those three formats and setting the power into fibre to 1dBm per channel. The launch power per channel is selected according to the EDFA total output power for field applications with 50GHz channel spacing in full C-band DWDM with 96 channels. As shown in Fig. 4(a), at targeted below 2dB propagation penalty, DP-PS16QAM doubles the transmission distance of DP-16QAM (24 spans versus 12 spans, respectively).
Fig. 4 (a) ROSNR penalties of DP-PS16QAM, DP-16QAM, and DP-8QAM versus number of spans transmitted at 1dBm per channel launch power; (b) ROSNR penalties versus number of cascaded WSS.
To verify if 200Gb/s DP-PS16QAM suits the 50GHz application, we transmit it through cascaded wavelength selective switch (WSS) to commercially be used for 50GHz DWDM system. Figure 4(b) shows the ROSNR penalties versus the number of WSS for the three formats. DP-PS16QAM has 1dB penalty after propagating through 10 WSSs, and performs much better than DP-8QAM which can only transmit through 3 WSSs.
5. 1Tb/s super-channel results
In this section, we discuss the results for 1Tbps Super-channel with five 200Gb/s DP-PS16QAM sub-carriers that was introduced in Section 3. Figure 5 shows how we formed the 1Tb/s super-channel. The measured crosstalk penalty of the central sub-carrier TX3 which has the most crosstalk from adjacent sub-carriers is depicted in Fig. 6(a); with each sub-carrier’s roll-off factor (alpha) set to 0.2, 0.1, and 0.05. Next, we select alpha as 0.05 to perform measurements for ROSNR penalty and link OSNR margin after 1200km and 1600km transmission. The summary is shown in Fig. 6(b). This measurement proves that the 1Tb/s super-channel with a SE of 5bit/s/Hz can transmit over 1600km at 1.5dB OSNR margin.
Fig. 6 (a) Central sub-carrier ROSNR penalty due to crosstalk with alpha set to 0.2, 0.1, and 0.05; (b) ROSNR penalty (solid line) and OSNR margin (dash line) after 1200km and 1600km transmission.
With the power into fibre per loaded channel set to 1dBm, we verify the long term stability of the central sub-carrier TX3 at the output of 20 spans by monitoring pre-FEC BER and after-FEC error. More than 1016 bits have been received throughout the monitoring period. No after-FEC error was observed during fourteen hours monitoring period and Fig. 7 shows pre-FEC and after-FEC BER over this period.
6. Summary
To the best of our knowledge, for the first time in real-time, we prove that DP-PS16QAM has OSNR gain of 1.8dB over uniform DP-16QAM by using a format configurable CFP transceiver. We placed 200Gb/s DP-PS16QAM in a 50GHz DWDM grid and it achieved over 1920km transmission with EDFA-only amplification, which doubles the propagation distance of uniform DP-16QAM. Moreover, a 1Tb/s super-channel, consisting of five 200Gb/s DP-PS16QAM sub-carriers with alpha 0.05 is placed in a 200GHz grid and it achieves over 1600km transmission and 5b/s/Hz SE with a raw SE at 6.86b/s/Hz. Experimental results prove that 200Gb/s DP-PS16QAM is an excellent candidate for the 50GHz DWDM system, and meets the three key requirements: propagation distance, spectral efficiency, and optical filtering tolerance.
References
1. Y. R. Zhou, Y. Rong, K. Smith, R. Payne, A. Lord, G. Whalley, T. Bennett, E. Maniloff, S. Alexander, and D. Boymel, “Real-time gridless 800G super-channel transport field trial over 410km using coherent DP-16 QAM,” in Optical Fiber Communication Conference (Optical Society of America, 2014), paper Tu2B.3.
2. Z. Zhang, C. Li, J. Chen, T. Ding, Y. Wang, H. Xiang, Z. Xiao, L. Li, M. Si, and X. Cui, “Coherent transceiver operating at 61-Gbaud/s,” Opt. Express 23(15), 18988–18995 (2015). [CrossRef] [PubMed]
3. G. D. Forney, “Trellis shaping,” IEEE Trans. Inf. Theory 38(2), 281–300 (1992). [CrossRef]
4. F. R. Kschischang and S. Pasupathy, “Optimal nonuniform signaling for Gaussian channels,” IEEE Trans. Inf. Theory 39(3), 913–929 (1993). [CrossRef]
5. F. Buchali, G. Böcherer, W. Idler, L. Schmalen, P. Schulte, and F. Steiner, “Experimental demonstration of capacity increase and rate-adaptation by probabilistically shaped 64-QAM,” in European Conference on Optical Communication (ECOC) (IEEE, 2015), paper PDP.3.4,. [CrossRef]
6. S. Chandrasekhar, B. Li, J. Cho, X. Chen, E. Burrows, G. Raybon, and P. Winzer, “High-spectral-efficiency transmission of PDM 256-QAM with parallel probabilistic shaping at record rate-reach trade-offs,” in European Conference on Optical Communication (ECOC) (IEEE, 2016), paper Th3C.1.
7. J. Cho, X. Chen, S. Chandrasekhar, G. Raybon, R. Dar, L. Schmalen, and E. Burrows, “Trans-Atlantic field trial using probabilistically shaped 64-QAM at high spectral efficiencies and single-carrier real-time 250-Gb/s 16-QAM,” in Optical Fiber Communication Conference (Optical Society of America, 2017), paper Th5B.3,. [CrossRef]
8. P. Schulte and G. Böcherer, “Constant composition distribution matching,” IEEE Trans. Inf. Theory 62(1), 430–434 (2016). [CrossRef]
9. R. Dar, M. Feder, A. Mecozzi, and M. Shtaif, “On shaping gain in the nonlinear fiber-optic channel,” in IEEE International Symposium on Information Theory (IEEE, 2014), pp. 2794–2798. [CrossRef]
10. G. Böcherer, F. Steiner, and P. Schulte, “Bandwidth efficient and rate-matched low-density parity-check coded modulation,” IEEE Trans. Commun. 63(12), 4651–4665 (2015). [CrossRef]
11. J. Cho, L. Schmalen, and P. j. Winzer, “Normalized generalized mutual information as a forward error correction threshold for probabilistically shaped QAM,” in European Conference on Optical Communication (ECOC) (IEEE, 2017), paper M2D.2. [CrossRef]
