Fighting Noise with Math
Every communication channel corrupts some bits. Rather than hoping for the best or demanding retransmission, forward error correction (FEC) adds carefully structured redundancy so the receiver can reconstruct the original message even when some bits flip. The key insight, proved by Shannon in 1948, is that reliable communication is possible at any rate below the channel capacity — you just need a good enough code.
From Hamming to LDPC
Richard Hamming's 1950 code could correct single-bit errors in 7-bit blocks — revolutionary for early computers. Reed-Solomon codes (1960) handle burst errors and still protect CDs and QR codes. Convolutional codes with Viterbi decoding dominated satellite and cellular systems for decades. The modern era began with turbo codes (1993) and LDPC codes (rediscovered 1996), which approach Shannon's limit to within a fraction of a decibel.
The Rate-Performance Trade-off
Every FEC code trades throughput for reliability. A rate-1/2 code sends two channel bits for every data bit, halving the raw throughput but enabling dramatic error reduction. A rate-7/8 code adds minimal redundancy and provides modest correction. The optimal rate depends on the channel: deep-space links use rate-1/6, while fiber optics use rate-0.9+ because the raw BER is already very low.
Visualizing Error Correction
This simulation shows the error correction process in action: random errors are injected into transmitted blocks at the specified raw BER, and the decoder attempts to correct them. You can see how increasing the correction capability t or decreasing the code rate R drives down the output BER, and observe the threshold effect where the decoder transitions from overwhelmed to virtually error-free.