Forward error correction (FEC) allows a predetermined amount of errors to occur during long-haul transmission, and detects and corrects them at the receiving end. As optical components become increasingly complex to meet the needs of error-free 40-Gbit/s networks, FEC is emerging as an efficient and cost-effective solution.
The obvious goal of error-free optical transmission is to ensure that absolutely no errors occur during transmission. This requires extremely tight control and management of every component in high-speed optical links. The downside is a 40-Gbit/s solution that is simply not economically feasible and/or technically possible. Stringent tolerances on optical and electrical components significantly increase their manufacturing costs by making them very difficult to produce in volume.
As a result, the preferred approach to error-free transmission is to allow for a predetermined amount of errors and then actively detect and correct them at the receiving end. This technique is known as forward error correction (FEC). Although errors actually occur during transmission, they are corrected before reaching the end user. There are several approaches to achieving this, such as maintaining existing link distances with less precise transmission components, extending existing allowable 10-Gbit/s link distances, increasing the allowable bit rates, or any combination thereof.
By exploiting certain mathematical algorithms, FEC improves the effective optical signal-to-noise ratio (OSNR) by correcting received bit errors at the expense of bandwidth to achieve a system coding gain. This is achieved by introducing a certain level of redundancy in the service data with no required feedback mechanisms. Since the service data and FEC data propagate together in the same forward direction, it is thus referred to as "forward" error correction. Due to the mathematical relationship between OSNR and bit-error rate (BER), a higher OSNR yields a lower BER and vice versa.
By using FEC to correct the raw received BER, a "mathematical" increase in the effective system OSNR is achieved simply by extrapolation. Subsequently, FEC mathematically overcomes such debilitating transmission impairments as attenuation, dispersion, and noise while ensuring reliable, cost-effective link performance at 40 Gbit/s. As FEC is primarily a mathematical technique, it is easily and effectively embedded into high-speed hardware devices such as application-specific integrated circuits.
REED SOLOMON ALGORITHMS
Reed Solomon (RS) coding is a special case of Bose Chaudhuri Hocquenghem (BCH) codes—block-based error-correcting codes used to actively detect and correct received errors. The incurred errors are a direct result of the transmission environment and the line rates chosen by the system designer. There are generally three types of errors:
- Burst: a group of bit errors that occur sequentially in time. Polarization-mode dispersion (PMD) is a transmission impairment that causes burst errors.
- Impulse: large blocks of data containing many errors that are typically caused by major system failures such as a defective transmitter.
- Random: bit errors that are independent of each other and can be caused by numerous sources such as the additive noise in a given optical link.
Reed Solomon coding is designed to correct random errors, but it can also correct burst errors implementing block codes. The number and type of errors that can be corrected depend upon the RS code characteristics. Reed Solomon codes are linear block-code subsets of BCH codes specified as RS(n,k) with s-bit symbols. Encoders use k data symbols each s bits long, adding parity symbols making n-symbol code words. There are n - k parity symbols of s bits each. Reed Solomon decoders can correct up to t-symbol code-word errors where 2t = n - k. For example, G.709 standard FEC RS(255,239) code uses symbols each 8 bits in length with each code word containing 255 code word bytes with 239 data bytes and 16 parity bytes.
n = 255 total code word length of 8-bit symbols
k = 239 total transported data length of 8-bit symbols
s = 8 total symbol length in bits
t = 8 can correct 8 symbols with errors in a code word
RS(255, 239) decoders dynamically correct any 8-symbol errors in a received code word. With symbol size s, the maximum code word length n is n = 2s - 1, or 255 bytes.
Reed Solomon coding error-correcting capabilities can be significantly improved by interleaving. Interleaving optimizes the error-correcting code capabilities to the specific characteristics of the transmission environment and enhances the random-error-correcting capabilities of the chosen RS codes, making this technique more effective in bursty types of noise environments such as PMD-impaired fiber links. Strong interleaved FEC is a key enabler to overcoming PMD in most 40-Gbit/s networks.
Interleaving works by rearranging the encoded bits over a span of several predetermined block lengths. The interleaver span length is determined by the amount of error protection required, based upon the expected burst lengths to be incurred in the transmission medium (the optical fiber link). For successful coding and decoding to occur, transmitters and receivers must be compatible with respect to being aware of the FEC bit-arrangement schemes. The ultimate goal of interleaving is to distribute a long burst of bit errors such that they appear at the decoder (receiver) as independent random bit errors or shorter, more manageable burst errors. The actual cost of interleaving is due to the additional hardware related to the interleave process itself.
FEC CODING GAINS
Forward error correction ensures that the probability of errors remaining in the decoded (corrected) data is lower than the probability of errors if FEC were not employed. An improved BER can be achieved by increasing the transmitter launch power to yield a higher OSNR (fewer received errors) or by using FEC at the original (lower) launch power and correcting the received errors. Subsequently, FEC allows an optical link to achieve a comparable BER at transmitter powers.
Numerous commercial FEC devices are available for a variety of transmission environments. However, commercial ICs that incorporate the coding gains required for successful and economic 40-Gbit/s line rate transmission do not yet exist. Initial 40-Gbit/s FEC devices will be proprietary in nature.
How are the generated FEC codes transported across an optical network from the ingress to egress points? Two methods are available: in-band FEC or out-of-band FEC. In-band FEC is commonly implemented in SONET/SDH optical networks by exploiting the undefined overhead bytes to carry FEC codes. The caveat is that there is a fixed amount of undefined overhead bytes available. This ultimately limits the maximum achievable coding gains, which are insufficient for 40-Gbit/s optical networks. Although the maximum in-band FEC coding gain is indeed limited, it is sufficient for most 10-Gbit/s-based optical networks currenlty deployed.
Out-of-band FEC is the transport vehicle of choice for 40-Gbit/s-based optical networks because of the substantial coding gains required. Owing to significant achievable out-of-band FEC coding gains, debilitating transmission impairments such as PMD, chromatic dispersion, nonlinear effects, and attenuation can be economically and reliably managed. As a result, power levels can be lowered in certain applications to significantly lower nonlinear effects while simultaneously compensating for PMD in the vast majority of post-1994 fiber plants. Since a 40-Gbit/s optical pulse actually contains four times less energy than a 10-Gbit/s optical pulse, a minimal 6-dB gain is required at the receiver. This additional required gain is offset using strong out-of-band FEC schemes with high resulting coding gains.
The most obvious advantage of FEC is the cost-effective achievement of significant net effective coding gains (NECGs). These NECGs represent a key technology enabler for 40-Gbit/s optical networks. Transmission impairments are managed in a more cost-effective manner than purely optically based management techniques.
Forward error correction also can be used for network fault prognosis. By monitoring a declining received BER, service providers can locate and proactively repair faults before they begin to affect service. Because of NECG margins, customers would be unaware of increasing bit errors because they are corrected before reaching the receiving end to maintain guaranteed service-level agreements. Since customers only receive a corrected BER, service providers can proactively repair impending failures before they become noticeable to end users. On shorter links, FEC avoids having to implement complex and costly active 40-Gbit/s PMD compensators in certain applications. Thus, out-of-band FEC can be seen as a critical component of reliable and cost-effective networks at 40-Gbit/s line rates.
However, unlike in-band FEC, out-of-band FEC actually increases the signal line rate (see table). This slight increase in line rate is often referred to as the line-rate "tax." However, unlike most taxes, this one is acceptable to most users. The goal of the system designer is to achieve an optimal balance between increased signal line rates due to FEC and a decreasing allowable reach (see figure). Higher line rates also mandate faster and wider band receivers due to greater eye closure from increased dispersion at higher line rates.
Still, these disadvantages do not outweigh the advantages achieved when using out-of-band FEC. If 40-Gbit/s networks are to be successfully commercialized, they must respect existing standard hut spacing with increasing distances between regenerators, thereby necessitating out-of-band FEC. As no standard yet exists for 40-Gbit/s FEC coding, vendors will implement proprietary solutions in the interim.
The idea of error detection and correction using various FEC schemes is not a new concept by any account. However, its application with respect to the next generation of 40-Gbit/s optical networks is significant. It is a proven technology that will enable the successful commercialization of 40-Gbit/s optical networks in the very near term.
Brian Lavallée is senior manager of systems engineering in Nortel Networks' Optical Internet business,2351 Boulevard Alfred Nobel St-Laurent, Quebec City, H4S2A9, Canada. He can be reached at firstname.lastname@example.org.