‘Soft decision’ FEC benefits 100G
One of the fundamental limitations in designing optical transport networks is optical signal-to-noise ratio (OSNR).
By RANDY EISENACH
One of the fundamental limitations in designing optical transport networks is optical signal-to-noise ratio (OSNR). WDM networks must operate above their OSNR limit to ensure error-free operation. The OSNR limit is one of the key parameters that determine how far a wavelength can travel before regeneration. Depending on whether a ROADM is designed for metro, long-haul, or ultra-long-haul applications, 10G wavelengths can be transported 800 to 2,000 km without any unusual measures before regeneration is required.
At data rates above 10 Gbps, however, advanced modulation schemes are needed to achieve similar reach. These modulation formats minimize the effects of such optical impairments as chromatic and polarization-mode dispersion as well as ensure the optical signal fits within the ITU 50-GHz grid used on modern DWDM systems. The downside of these higher data rates and advanced modulation schemes is that they require substantially better OSNR performance than do conventional 10-Gbps transmissions. At 100 Gbps, the minimum OSNR required is +10 dB higher than for 10-Gbps wavelengths.
Without some type of correction or compensation, the OSNR requirements would limit 100G optical transport to extremely short distances. Fortunately, sophisticated forward error correction (FEC) techniques, particularly “soft decision” FEC, can extend the reach of 100G signals to much longer, more usable distances.
Basics of FEC
FEC is a method of encoding a signal with additional error detection and correction overhead information (i.e., parity bytes) so that optical receivers can detect and correct errors that occur in the transmission path. FEC dramatically lowers the bit-error rate (BER) and extends the distances that optical signals can be transmitted without regeneration.
There are a number of FEC algorithms available that vary in complexity, strength, and performance. One of the more common and standardized first generation FECs is Reed-Solomon (255, 239). Reed-Solomon adds slightly less than 7% overhead for the FEC bytes and provides about 6-dB net coding gain. In high-speed optical networks, a 6-dB gain is a very significant performance improvement – approximately quadrupling the distance between regenerators.
In addition to Reed-Solomon FEC, many vendors offer stronger second generation FEC schemes as a provisionable parameter on 10G and 40G optical interfaces. These “ultra” FEC and “enhanced” FEC (EFEC) algorithms still use 7% overhead but implement stronger, more complex encoding and decoding algorithms that provide an additional 2- to 3-dB coding gain over Reed-Solomon.
While first generation Reed-Solomon FEC and second generation EFEC have provided substantial performance improvements for 10G and 40G wavelengths, even stronger, more complex third generation FEC algorithms are needed at 100G to achieve optimal performance.
Soft decision FEC
At 100G rates, leading optical suppliers are implementing third generation FEC capabilities to extend performance and overall optical distances even further. These third generation FECs are based on even more powerful encoding and decoding algorithms, iterative coding, and something referred to as soft decision FEC (SD-FEC). In a hard decision FEC implementation, the decoding block makes a firm decision based on the incoming signal and provides a single bit of information (a “1” or “0”) to the FEC decoder. A signal is received and compared to a threshold; anything above the threshold is a “1” and anything below the threshold is a “0.”
A soft decision decoder uses additional data bits to provide a finer, more granular indication of the incoming signal. In other words, the decoder not only determines whether the incoming signal is a “1” or “0” based on the threshold, but also provides a “confidence factor” in the decision. The confidence factor provides an indication of how far the signal is above or below the threshold crossing.
The use of confidence or “probability” bits along with the stronger, more complex third generation FEC coding algorithms enables the SD-FEC decoder to provide 1–2 dB of additional net coding gain. In practice, a 3-bit confidence estimation normally provides most of the theoretically achievable performance improvement. While 1–2-dB coding gain doesn’t sound like much, it can translate into a 20% to 40% improvement in overall achievable distances, which is a very substantial improvement at 100G.
One tradeoff with these more advanced FECs is they require ~20% overhead for the FEC bytes, more than twice the ~7% overhead of first and second generation FECs. The higher 20% FEC overhead translates to slightly higher optical data rates, which are already operating at the edges of currently available technology at 100G.
Implementing 100G SD-FEC
While the mathematics behind SD-FEC algorithms have been known for many years and used in the wireless industry, it is only recently that SD-FEC has gained interest for use on high-speed optical signals. Numerous technology and ASIC limitations prevented implementation of third generation SD-FEC in optical applications. In other words, the semiconductors weren’t fast enough and didn’t have enough processing power or memory to support SD-FEC at 100G optical rates.
Take, for example, the high-speed analog-to-digital converters (ADCs) used inside a 100G receiver. These devices operate at an incredible 56 gigasamples per second (Gsa/sec) and just became generally available in 2011. SD-FEC requires the use of even higher-speed ADCs, operating at 63 Gsa/sec to implement the SD-FEC process-ing, along with an equally fast and powerful SD-FEC silicon implementation. Fortunately, such component limitations are now part of the past, meaning that SD-FEC for 100G optical signals has become a reality.
Ready for use
As backbone speeds increase from 10G to 100G per wavelength, the OSNR requirements increase by +10 dB. Without some type of compensation or correction, 100G optical distances would be very limited and uneconomical.
First and second generation FEC algorithms have been used at both 10G and 40G to lower the BER and improve overall distances. Soft decision FEC is a third generation encoding algorithm that enables longer distances and fewer regenerations on 100G optical networks.
Randy Eisenach is a WDM product marketing manager at Fujitsu Network Communications Inc.