ICACT20210252 Question.1
Questioner: lijingwei10@huawei.com 2021-02-09 오후 8:28:34 |
ICACT20210252 Answer.1
Answer by Auhor ra-maneekut@kddi-research.jp 2021-02-09 오후 8:28:34
Chrome  Click!! |
| Does the constellation gain have a theoretical upper limit? And What is the gap between your results and the theoretical upper limit? |
The theoretical upper bound of the constellation shaping can be obtained from mutual information (MI) which is the metric for symbol-wise (SW) soft decision (SD) decoder. Since we optimized the constellation based on MAP-BER using bit-wise hard decision decoder, there are some gaps between those two decision method. By the way, the gap in achieved information rate (AIR) vs SNR curve between MI and HD-GMI (converted from MAP-BER) are less than 1 dB roughly. |
ICACT20210252 Question.2
Questioner: lijingwei10@huawei.com 2021-02-09 오후 9:11:43 |
ICACT20210252 Answer.2
Answer by Auhor ra-maneekut@kddi-research.jp 2021-02-09 오후 9:11:43
Chrome Click!! |
| For the long-haul transmission experiment over 800 km, it gets better shaping gain,why it happens? And do you have other ways to improve this perfermance further? |
The shaping gain using MAP detection is strongly depended on the entropy H(P) that's why we can see large shaping gain for the case of low H(P). Furthermore, the distance is chosen based on the target SNR, it may suffer from the deterioration in the ASE side more than the fiber nonlinearity side that's why we can see the shaping gain in MAP-BER vs SNR curve. MAP detection gives good performance in AWGN regime. |
ICACT20210252 Question.3
Questioner: lijingwei10@huawei.com 2021-02-09 오후 9:19:25 |
ICACT20210252 Answer.3
Answer by Auhor ra-maneekut@kddi-research.jp 2021-02-09 오후 9:19:25
Chrome Click!! |
| Is your shaping gain influenced by the modulation format? |
In this work, only one modulation format is used, the 64-QAM. We compared the PS-64QAM (PS over square 64QAM) and the HPGS-64QAM (PS over GS-64QAM). Therefore, the shaping gain we obtained are from the shaping in both probabilistic and geometric spaces which can slightly improve the performance better than probabilistic shaping only. |
ICACT20210252 Question.4
Questioner: gueye.ibrahima1@esp.sn 2021-02-10 오후 9:45:47 |
ICACT20210252 Answer.4
Answer by Auhor ra-maneekut@kddi-research.jp 2021-02-10 오후 9:45:47
Chrome Click!! |
| Hello, thank you for the presentation, what are the crystals you used to make the comparison between the optimized constellations and the regular square 64-QAM constellations. |
If my understanding is correct, the questioner ask that what metric that I use to compare both constellations ? In this case, we consider hard decision decoding method so we use the MAP-BER as the metric. The threshold of BER is at 1e-2 which is the level of FEC limit assuming 20% overhead. The difference in SNR at that threshold is obtained as the shaping gain between 2 constellations. |
