A Brief Analysis of SD-FEC

Forward error correction (FEC) is widely used in optical communications to improve error correction, enhance system reliability, and extend optical transmission distance. It is also used to reduce optical transmitter power and system costs. In response to the rapid growth of optical communications, the ITU-T has started researching FEC coding and has recommended ITU-T G.707, G.975, G.709, and G.975.1. As optical transmission systems evolve towards longer transmission distance, greater capacity, and speeds of 100G or  beyond, chromatic dispersion, polarization mode dispersion, and nonlinear effects are seriously affecting smooth evolution. High-performance FEC codes are therefore being developed so that higher net coding gain (NCG) and better error correction can be achieved.

An Efficient FEC Technique

The OSNR tolerance for 10G NRZ is less than 12 dB when the pre-FEC BER is 2 × 10-3. However, the OSNR tolerance for 100G PM-QPSK is around 15.5 dB when pre-FEC BER is 2 × 10-3. When the same FEC is used, the transmission distance of 100G is less than half that of 10G. It is therefore necessary to introduce a highly efficient FEC technique.

Adaptive forward-error correction (AFEC) has been widely used in 10G and 40G DWDM systems and an NCG of about 8.5 dB can be achieved. The Optical Internetworking Forum (OIF) suggests that soft-decision forward-error correction (SD-FEC) with a redundancy of 18−20% be used in a 100G DWDM system. The NCG can reach up to 10.5 dB, and the line rate is approximately 126 Gbit/s. With efficient SD-FEC, the OSNR tolerance for 100G PM-QPSK is around 13 dB. This ensures that 100G transmission systems cover almost the same distance as 10G systems.

FEC Classification

FEC codes can have block or convolutional structures. Block codes include hamming, reed-solomon (RS), and BCH codes and have been extensively used in optical communications. Most block codes are constructed in the Galois field and thus have a strict algebraic structure. An algebra-based hard-decision decoding algorithm is used for block codes. Convolutional codes have a dynamic structure that can be described by a finite state machine. A soft-decision decoding algorithm is used for convolutional codes. Because convolutional codes do not support parallel decoding architecture, they have a long decoding delay. Convolutional codes are therefore seldom used in optical communications.

FEC coding can be done using hard decision and soft decision. Hard-decision decoding is based on traditional error correcting. A demodulator makes the best hard-decision on the channel outputs, and the demodulator sends the decision results to a hard-decision decoder. The decoder receives code streams, typically 0 or 1 in a binary code, and corrects errors by using the algebraic structure.

Soft-decision decoding uses the waveform information that is output by channels. A real number is output by a matched filter, and the demodulator sends this to a soft-decision decoder. The decoder needs not only 0 or 1 code streams but also soft information to indicate the reliability of these input code streams. The further the code value is from the decision threshold, the more reliable the signal is, and vice versa.

Because a soft-decision decoder has more channel information than a hard-decision decoder, it can use the information through probability decoding and obtain higher coding gains than a hard-decision decoder.

FEC Evolution

First generation

First generation FEC uses hard-decision block codes. The typical representative is the RS (255, 239) code with a 6.69% overhead. When an output BER is 1E-13, the RS code yields a net coding gain of about 6 dB. RS (255, 239) codes have been recommended for long-haul optical transmission as defined by ITU-T G.709 and G.975.

Second generation 

Second generation FEC uses hard-decision concatenated codes combined with interleaving and iterative decoding techniques to improve FEC capability. The ITU-T G.975.1 standard has defined eight second-generation FEC algorithms with 6.69% overhead. When an output BER is 1E-15, most FEC algorithms yield a net coding gain of more than 8 dB and support 10G and even 40G long-haul transmission systems.

Third generation

Third generation FEC uses soft-decision. As the single-channel rate evolves from 40G to 100G, a coherent receiver is the necessary to develop 100G long-haul transmission equipment. Coherent receiving technology in optical communication systems and the rapid growth in integrated circuit technology make the application of soft-decision FEC possible. When an output BER is 1E-15, a soft-decision FEC scheme with 15−20% overhead yields a net coding gain of about 11 dB. This is enough to support long-haul 100G and beyond transmission. Soft-decision FEC often uses turbo product codes and low density parity check codes.

ZTE 100G SD-FEC

ZTE's 100G SD-FEC scheme has the following features:

• an innovative, full soft-decision FEC algorithm that achieves higher gains, higher integration and lower power consumption
• brand-new, optimized algorithm and architecture. The FEC scheme with 15% overhead now has strong error correction performance, allows an input BER threshold of between 1.8E-2 and 2E-2, and effectively prevents line errors.
• 100% soft-decision decoding. No concatenated hard-decision FEC codes are necessary, and this greatly reduces decoding delay
• innovative, optimized codeword structure and decoding algorithm that provides ultralow error floor
• soft-decision FEC scheme with 15% overhead that offers higher transmission efficiency and better wave filtering performance than an FEC scheme with 20% overhead.