You are on page 1of 10

FEC ensures reliable transmission

As the basic bearer technology of optical communication systems, WDM has seen a
rapid expansion in capacity alongside a similarly large increase in network traffic. During
this expansion, every increase in the rate of a single wavelength has brought about a
significant advance in communication technology:
The increase from 2.5G to 10G wavelengths resulted in an evolution from direct
modulation to external modulation and the use of DCMs for fiber dispersion
compensation.
The increase from 10G to 40G wavelengths marked a transition from OOK
modulation to PSK modulation.
The increase from 40G to 100G brought about the introduction of DSP-enabled
coherent communication technology, in which analog-to-digital converters (ADCs) work
at a rate higher than 56Gbit/s.
As WDM technology advanced, forward error correction (FEC) took on a key role in
ensuring reliable transmission of information and gradually became a mainstream
technology essential for optical communication systems. FEC technology for fiber
communication has also experienced several generations of evolution: from classical
hard decision codes to concatenated codes and, more recently, to soft-decision FEC,
which is advancing rapidly thanks to the use of 100G coherent communication
technology.
FEC is a coding technology widely used in communication systems. Using a classical
block code as an example, the FEC encoder at the transmit end uses kilo bits of
information as a block code. During the encoding, FEC adds n-k redundant check bits to
the information bits, constructing an n-bit codeword. After the codeword is transmitted to
the receive end over a channel, the FEC decoder detects and corrects bit errors during
decoding – if the errors are within the correction range. In this manner, FEC prevents
interference from channel transmission and improves the reliability of an optical
communication system. This shows that FEC can efficiently reduce the system BER with
only a small number of redundant overhead bits, thus extending the transmission
distance and reducing system costs.
FEC can efficiently improve system performance. According to the Shannon theorem,
FEC can ensure information is transmitted over a noisy channel at the maximum rate
(Shannon limit) without any bit errors. As shown in Fig. 2, the FEC correction
performance is subject to three factors: coding overhead, decision method, and code
scheme.

Coding overhead: The ratio of the number of check bits (n-k) to the number of
information bits (k). Theoretically, a higher overhead means a higher Shannon limit.
However, in practice this increase is not linear and, as overhead increases, improvement
in FEC performance is reduced. The selection of an overhead scheme is determined by
system design requirements.
Decision methods: FEC coding can be carried out with either of two decision methods:
hard decision and soft decision. The input to a hard-decision FEC decoder consists of a
single level of the binary bits 0 and 1. The low complexity but high maturity of hard
decision decoding makes it widely used in a variety of scenarios. On the other hand, the
input to a soft-decision FEC decoder is a multilevel quantization signal. While offering
the same coding rate as hard decision FEC, soft-decision FEC provides a higher coding
gain, albeit with a greatly increased processing complexity. Furthermore, as micro-
electronic technologies advance, soft-decision decoding that can support 100G
throughput is now becoming possible. With new developments in transport technology
and the advent of the 100G era, research into, and the applications of, soft-decision FEC
are maturing and will eventually be widely used in high-speed optical communication
systems that use coherent detection.
Code scheme: After the coding overhead and decision method are determined, an
optimal code scheme that ensures transmission performance close to the Shannon limit
becomes the major concern in a system using FEC. FEC using low-density parity-check
(LDPC) codes has gradually become the major FEC scheme for high-speed optical
communication because it provides excellent correction performance and allows a high
degree of parallelism.

Third-generation FEC enhancing 100G transmission


Research on fiber applications for FEC got under way recently, and it was only in 1988
that Gover applied FEC to fiber communication for the first time. The various types of
FEC used in fiber communication can be classified into three generations, based on their
developmental history.
First-generation FEC: It uses classical hard-decision codewords; for example, Hamming,
BCH, and RS codes. The most classical representative of these is the RS (255,239)
codeword, providing a 6.69% overhead. When the input BER is 1.4E-4 and the output
BER is 1E-13, the RS (255,239) code yields a net coding gain of 5.8 dB. RS (255,239)
codes are recommended for long-haul transmission as defined by ITU-T G.709. They
match STM-16 frame formats well and are widely used. In 1996, RS (255,239) codes
were successfully used in trans-Pacific and trans-Atlantic communication systems as
long as 7000 km and provided data rates as high as 5Gbit/s.
Second-generation FEC: Besides hard-decision codewords, second-generation FEC
uses concatenated codes alongside interleaving, iterative, and convolutional decoding
techniques, which increases FEC coding gain significantly and allows the technology to
support 10G and even 40G transmission systems. In many cases second-generation
FEC yields a net coding gain higher than 8 dB. The FEC scheme recommended by ITU-
T G .975.1 can be regarded as being representative of second-generation FEC.
Most live 10G systems use the second-generation hard-decision FEC schemes;
however, hard-decision FEC usually uses small overheads and cannot provide high
performance. So, hard-decision FEC with larger overheads is demanded to support
smooth upgrade of live systems. For example, 10G submarine transmission systems
currently use hard-decision FEC with 6.69% overhead recommended by ITU-T G .975.1.
If the systems use hard-decision FEC with 20% overhead, the net coding gain can be
increased by 1.5 dB, thereby significantly improving the system performance.
Third-generation FEC: The use of coherent detection in optical communication systems
makes possible the application of soft-decision FEC. A soft-decision FEC scheme with
20% or larger overhead (such as schemes using Turbo codes, LDPC codes, or TPC
codes) yields a net coding gain of over 10 dB, which provides solid support for ultra long-
haul transmission of 40G, 100G, and even 400G data.
Current 40G systems use hard-decision FEC schemes with high overheads that promote
smooth upgrades. They may also use coherent detection and soft-decision FEC
schemes to provide more powerful transmission performance. As 100G and 400G fiber
transmission systems use coherent detection, soft-decision FEC is therefore a very
appropriate partner technology. Soft-decision FEC uses concatenated codes, turbo
codes, and interleaved codes, and provides net a coding gain of higher than 11 dB.
Therefore, it can completely address the requirements of 100G and 400G systems (see
Fig. 3).
Soft decision decoding principles
A hard-decision FEC decoder receives data streams consisting only of the binary digits 0
and 1. Hard-decision decoding will normally be performed based on the algebraic code
format. With this decoding mode, statistical characteristics of channel interference in a
signal are lost.

To fully utilize the information in a received waveform and improve the decision accuracy
of the decoder, sampling and quantization can be performed for the received signal (see
Fig. 4). Using this sampling information, the decoder provides higher decoding accuracy
and therefore greatly improves system performance. When working at the same rate,
soft-decision FEC provides a 1.5 dB higher net coding gain than hard-decision FEC.
Soft-decision FEC ensures better performance and a bit throughput that is many times
that of hard-decision FEC. However, it can be implemented only when a high-speed
ADC is used to perform sampling and quantization. In addition, the soft-decision FEC
algorithm is very complex because it must consider the changes in noise probability
distribution caused by channel performance deterioration. Fortunately, the rapid
development of integrated circuits has made commercial use of soft-decision FEC a
reality.
LDPC principles
In 1962, Gallager first came up with LDPC codes and defined the construction of each
LDPC code and the hard-decision decoding algorithm. Due to the high complexity of
LDPC codes and the low capability of computer processors at the time, LDPC codes
were ignored in optical fiber communication systems for some years. Then in 1996,
MacKay and R. Neal rediscovered LDPC codes and demonstrated that they can achieve
error performance close to the Shannon limit. Since then, a great deal of research effort
has gone into the design and construction of LDPC codes.
An LDPC code uses a sparse parity check matrix, which gives the rise to the name
LDPC. Such a matrix uses codes with sparse bits (that is, bits where 1s are less common than
0s) and therefore ensures a relatively long Euclidean distance between codewords.

LDPC codes with soft decision decoding are based on belief propagation (BP), such as
the sum-product algorithm. This algorithm uses the channel information obtained by
quantifying level signals and the parity check matrix to establish the operational
relationship between bits inside a codeword. Based on this operational relationship, the
algorithm determines the "belief" level of bits. On the Tanner graph, the algorithm
transfers the belief information between adjacent variable and check nodes. As shown in
the graph, through several iterations the algorithm improves the decision accuracy.
Like turbo product codes, LDPC codes achieve good error performance close to the
Shannon limit and provide powerful correctional capability even at a high BER. LDPC
codes also show litter error floor and their decoding complexity linearly only increases
with code length, resulting in just moderate complexity in calculations. In addition, LDPC
codes support parallel decoding, which makes them suitable for high-speed
communication systems. As such, LDPC codes have been used in various fields,
including optical communication, wireless communication, deep space communication,
and magnetic recording.
In the Chang'e-2 lunar probe of the second phase of the Chinese Lunar Exploration
Program, LDPC coding was at the core of six engineering objectives and four
innovations.
In November 2010, Tsinghua University successfully finished testing its LDPC-based
telemetry, tracking & control (TT&C) channel coding scheme. This marks the first
application of the LDPC coding technique in China's aerospace field. Thanks to the
strong correctional capability and high coding gain, it is thought that LDPC codes can
further improve channel margin and drastically improve the reliability of satellite TT&C.
All these advantages make LDPC codes a good choice for future deep space
exploration.
In the optical communication field, 100G is regarded by the industry as the most
important transmission rate of the next generation. With the speedy development of
electronic technology and commercialization of high-speed ADC (with sampling rates
higher than 56Gbit/s) and DSP devices, 100G coherent transmission technology has
become the mainstream 100G transmission technology, thanks to its high OSNR
sensitivity, as well as strong CD and PMD compensation capability. This trend makes
LDPC-based soft-decision FEC a good candidate for 100G long-haul transmission.
Although 100G coherent transmission technology greatly outperforms non-coherent
transmission technology, which has been widely used in 10G and 40G transmission
systems, it does have some problems, such as short transmission distance and limited
coverage. LDPC-based soft-decision FEC provides a promising solution to further
improve the performance of 100G transmissions.

LDPC-based soft-decision FEC


The performance of soft-decision FEC depends greatly on the code construction and
decoding algorithm. It can be improved to a large extent by using a rational code
construction, an efficient decoding algorithm, interleaving, concatenated, and
convolutional decoding techniques, and full use of redundancy.
Huawei's block interleaved convolutional code (BICC), which is an LDPC-based high-
end FEC scheme, combines interleaving and concatenated decoding techniques into the
conventional LDPC codes and yields a coding gain of higher than 11.5 dB. Compared
with a conventional hard-decision FEC scheme, BICC improves coding gain by 2.6 dB,
providing levels that are sufficient to support 100G high-speed optical transmission.
Huawei's 100G soft-decision FEC scheme has the following characteristics:
Higher coding gain and integration alongside lower power consumption.
100% soft-decision decoding without using concatenated hard-decision FEC codes,
greatly reducing transmission latency.
Unique architecture, implemented by cascaded pipelines, greatly reduces
complexity when soft decision decoding.
Innovative soft decision architecture, which is scalable and flexible to provide
different power consumption specifications and performances.
Twenty percent or higher overhead, together with spectrum compression at the
transmit end, which reduces the transmission penalty caused by rate increases at the
transmit end and ensures the desired coding gain for high overhead soft-decision FEC.
Soft decision combined with Huawei's unique DSP technique provides differentiated
system solutions.

FEC in the future


As FEC technology develops, code performance is approaching the Shannon limit.
However, the application of FEC technology in optical transmission systems must take
system requirements into consideration. Users have to plan a code scheme that uses a
rational overhead and decision method, but involves low design complexity.
Hard-decision FEC will be still used in the optical communication field for a long time.
System designers have never stopped researching ways to combine multiple techniques
to develop high-performance hard-decision FEC schemes, so that the FEC schemes can
support upgrade of existing 10G and 40G systems.
For 100G, 400G, and even 1T high-speed optical transmissions, though, soft-decision
FEC will make continuous improvements in increasing throughput and net coding gain,
reducing error floor and the complexity of decoding algorithms. New generations will
adapt better to non-AWGN non-linear channels, to address the requirements of higher
transmission rates, and soft-decision FEC with high performance but moderate
complexity will become the mainstream FEC scheme used to advance the optical
transport applications and developments.

You might also like