8b10b

8b/10b encoding is a binary line code used to transmit 8-bit data as 10-bit symbols. It was developed for high-speed serial communications to provide DC balance and reliable clock recovery while limiting long runs of identical bits. In the encoding, each 8-bit input is split into a 5-bit and a 3-bit portion. The 5-bit portion is encoded into 6 bits (5b/6b), and the 3-bit portion is encoded into 4 bits (3b/4b). The resulting 6 and 4 bits are concatenated to form a 10-bit output. A running disparity mechanism tracks the difference between number of 1s and 0s to keep the overall signal near zero DC level across long transmissions, aiding power efficiency and optical/electrical signaling.

Not all 10-bit patterns are used for data; several patterns are designated as control or special codes (K-codes) to carry control information or align the receiver. A well-known example is the K28.5 code, used as a comma character to facilitate bit alignment in a stream. The encoding therefore supports both data and control information, while constraining the symbol set to maintain DC balance and a minimum density of signal transitions.

Performance and tradeoffs: 8b/10b introduces a 25% overhead, since 8 data bits become 10 bits per symbol. It provides predictable timing and robust decoding, but at the expense of reduced raw data rate compared to more modern schemes that use higher-efficiency encodings. It remains widely associated with early PCIe generations, SATA, Fibre Channel, and certain other high-speed serial interfaces, where its balance of simplicity, balance, and transition density suited existing hardware.