What is FEC?
2019-12-03 | 4 min read
Data center operators have begun migrating their networks to 400 gigabit Ethernet (GE), and research and development into 800G and terabit speeds is under way. Faster throughput is no longer just a matter of increasing network speed. New breakthrough technologies, such as four-level pulse amplitude modulation (PAM4) signaling, are the solution to increasing network bandwidth while maintaining the data center footprint. This is necessary as most data centers are already at maximum capacity.
The use of PAM4 signaling has changed Ethernet test requirements. Due to the reduced signal to noise ratio (SNR), PAM4 links do not necessarily run error-free and require forward error correction (FEC). FEC is an advanced coding technique that sends the required information to correct errors through the link along with the payload data. The benefit of “forward” in FEC is that there is no need to retransmit information to correct errors detected at the receiver.
FEC is especially beneficial in applications where retransmissions are costly or impossible, such as one-way communication links and transitions to multiple receivers in data centers. The FEC technique controls errors in data transmissions over unreliable or noisy channels. FEC transforms transmitted data (message sequences) into code words that include redundant data. The receiver decoder uses the added redundancy to detect and correct errors that may occur in the message, minimizing the need to retransmit.
System tests need to change as new measurements are required to support 400G, PAM4, and FEC. The key measurements affected by FEC test implications include the transmitter (output) and receiver (input) burst error counting and analysis. Testing with stressed FEC-encoded data and measuring BER in links that do not by design run error-free is also important. The infographic below illustrates the 400GE FEC encode/decode processing for both the transmit and receive paths.
Infographic: 400GE FEC Encode/Decode Processing
Specifications for all 400GE systems now require FEC. The Institute of Electrical and Electronics Engineers (IEEE) standard for 100GE backplane first defined the RS(544, 514), or KP4, and RS(544, 528), or KR4, FEC Reed-Solomon codes for Ethernet test. Reed-Solomon codes use symbols and codewords, rather than bytes and frames, to correct symbol errors rather than bit errors. Note that in the case of FEC, the term “symbol” does not equate to PAM4 symbols. All RS FEC decoding is done at the bit level after the conversion of the PAM4 gray code to a non-return-to-zero (NRZ) bitstream.
Before FEC, engineers tested signal transmissions by checking for packet errors to ensure that the bit error ratio (BER) was better than 10-12 (or 10-15 depending on the interface). However, even with the use of FEC, engineers must still check for BER and bit errors. Frame loss ratio (FLR), a ratio of lost packets to the total number of packets, is a new IEEE metric for uncorrectable code words in high-speed Ethernet links with FEC. Since errors are pro-actively corrected, it is less obvious whether the transmissions are “good”. Engineers need to understand FEC test implications and look deeper into test results to achieve a healthy system.
Find out more about PAM4 FEC test challenges at: PAM4 High-Speed Digital System Design Solutions