100G相干SD-FEC调制解调器：业界首例CMOS芯片与系统验证
需积分: 9 23 浏览量
更新于2024-09-10
收藏 1.57MB PDF 举报
本文档标题为"RT-100G Transceiver with SDFEC：一项创新技术的介绍"，主要探讨了一款先进的120 Gb/s相干极化复用四相移键控（QPSK）光收发器，特别强调了其软决策前向纠错（SD-FEC）编码的应用。这项突破性的技术基于Turbo Product Code，首次实现了40纳米互补金属氧化物半导体（CMOS）应用特定集成电路（ASIC）的集成设计，集成了模拟到数字转换、数字信号处理以及SD-FEC功能。该器件遵循了光网络论坛（Optical Internetworking Forum，OIF）的多源协议标准，确保了设备间的兼容性。
文章的核心内容聚焦在了对这款100Gbps实时传输模块的详细测试与验证。作者团队，包括Nelson、Zhang、Birk等人，通过一系列长时间距离的实验，如超过1000公里的长途和超长距离试验（达到3760公里），深入检验了这个新型Transceiver在实际网络环境中的性能和鲁棒性。他们证实了这款设备在面对诸如光纤非线性、高达60,000 ps/nm的色散以及极化模态色散等系统挑战时展现出的高度容忍度。
软决策前向纠错（SD-FEC）是关键的纠错机制，它能在接收端利用概率统计方法，即使在存在噪声和错误的情况下也能更准确地识别并纠正数据包中的错误。这对于100Gbps级别的高速数据传输至关重要，因为高速率下更容易出现错误，而SD-FEC的引入显著提高了数据的可靠性和传输效率。
此外，文章还可能涵盖了该器件的功耗分析、集成度优势、温度和供电影响的考量，以及与传统FEC方法的对比，以突出SD-FEC在提升信号质量、降低误码率方面的显著效果。这篇论文不仅介绍了技术原理，也提供了实用的工程实现和性能评估结果，为高性能光通信系统的未来设计和发展提供了有价值的技术参考。
Nelson et al. VOL. 4, NO. 11/NOVEMBER 2012/J. OPT. COMMUN. NETW. B133
Fig. 2. (Color online) Block diagram of the 100 Gb/s WDM test
setup over the uncompensated 1000 km link. The tunable laser,
N7788B polarization controller (PC) and polarimeter were added for
the experiments of Section V. Inset: the 40-channel spectrum at the
booster output.
Fig. 3. (Color online) Measurements of pre-FEC BER versus OSNR
for the 100 Gb/s MSA transceiver as channel 20 of the 40 × 100 Gb/s
1000 km system, with +3 to +6 dBm/ch launch powers into the spans.
used in the link; hence, the MSA transceiver compensated
for approximately 17,000 ps/nm of chromatic dispersion. After
each span, a two-stage erbium-doped fiber amplifier (EDFA)
was used to amplify the optical signal power. To optimize the
gain tilt, variable optical attenuators were introduced at the
mid-stages of EDFAs 1, 2, 4, 5, 7, 8, and 9, while 50 GHz
grid wavelength blockers at the mid-stages of EDFAs 3 and 6
allowed per-channel equalization. The total launch power into
each span could be varied from +19 to +21 dBm, while still
maintaining a flat channel spectrum to within ±1.5 dB at the
input to each span. Following the tenth fiber span, the test
channel was noise loaded, and its OSNR was measured with an
optical spectrum analyzer (OSA). Finally, a second commercial
50 GHz WSS was used to de-multiplex the test channel before
it was sent to the receiver input of the 100G transceiver.
Figure 3 presents BER curves after 1000 km WDM trans-
mission for launch powers of +3, +4, +5 and +6 dBm/channel
into the fiber spans. (To reach +6 dBm/ch, 20 channels in the
center of the band were transmitted.) At +3 dBm/ch, the OSNR
Fig. 4. (Color online) Comparison of the measured pre-FEC BER
versus OSNR curves from Fig. 3 with the curves measured when the
three channels on both sides of channel 20 were turned off.
penalty relative to back-to-back is 1.0 dB for a BER of 1.9×10
−2
(the SD-FEC threshold). Moreover, the received OSNR after
1000 km is 22.8 dB, indicating that sufficient margin remains
for transmission distances of 2000 km. As shown in Fig. 3, the
OSNR penalty increases to 3 dB for +6 dBm/ch; however, at
that launch power, the BER floor is still well below the FEC
threshold. To test the impact of the neighboring 100G channels,
BER curves were also measured at the different launch powers
when the six nearest neighbors (i.e., ch 17–19 and 21–23)
were turned off. To maintain the same per-channel power and
EDFA loading, three additional channels were turned on at
the short and long wavelength ends of the spectrum. Figure 4
shows a comparison of the BER curves for the channel under
test with neighboring 50-GHz-spaced, 100G channels and with
nearest neighbors spaced 200 GHz from the channel under test
(i.e., with ch 17–19 and 21–23 turned off). At +3 dBm/ch, the
required OSNR difference is only 0.3 dB, while the difference
is 1.5 dB for +6 dBm/ch. Note that, given the SSMF dispersion
and the uncompensated link, we believe the 2 Gbaud lower
symbol rate of the loading channels compared with that of
the channel under test will have only a minor impact on
the measured WDM penalties. In fact, the 28 Gbaud loading
channels walk through the channel under test at a slightly
slower rate than 30 Gbaud loading channels would, and thus
the measured WDM penalties may be slightly higher than the
actual WDM penalties would be with loading channels at the
same 30 Gbaud symbol rate as the channel under test.
As mentioned previously, the SD-FEC has been designed to
convert a pre-FEC BER of 1.9 × 10
−2
to less than 1 × 10
−15
after SD-FEC decoding, corresponding to more than 11 dB net
coding gain. In the experiments, the measured FEC threshold
is in good agreement with prediction. Figure 5 presents the
recorded post-FEC errors after SD-FEC decoding as a function
of measured pre-FEC BER for the different launch powers,
including the points from the curves of Fig. 3. Comparing the
curve for back-to-back with those after transmission verifies
that the FEC threshold—and thereby the coding gain—is
essentially maintained in the presence of fiber nonlinearities.
The FEC threshold was measured to change by <0.15 dBQ
at power levels up to +6 dBm per channel, the maximum
launch power attainable in this setup. To test repeatability,
BER curves for 3 dBm/ch and 5 dBm/ch were measured twice,
with slightly different received OSNRs.
剩余10页未读，继续阅读
157 浏览量
2023-05-19 上传
267 浏览量
2022-07-15 上传
428 浏览量
143 浏览量
2022-09-23 上传
131 浏览量
langbin418
- 粉丝: 16
- 资源: 16
我的内容管理
展开
