Turbokoodien

Turbokoodien, or turbo codes, are a class of high-performance error-correcting codes used in digital communications. Introduced in the early 1990s, they approach the Shannon limit of channel capacity when used with long block lengths. A typical turbo code is formed by two recursive systematic convolutional encoders connected in parallel by a random interleaver. The input bits are transmitted along with two parity streams produced by the encoders, yielding a rate that can be adjusted by puncturing or by choosing the encoder and interleaver parameters.

Decoding is performed iteratively using soft-input soft-output decoders, often based on the BCJR algorithm (MAP) for each constituent encoder. The decoders exchange log-likelihood ratio information via the interleaver, repeatedly refining the estimates of the transmitted bits. This iterative process yields significant performance gains over traditional convolutional codes at practical code lengths.

Turbo codes were introduced independently in 1993 by Claude Berrou, Alain Glavieux, and P. Thitimajai at the Institut National des Télécommunications. Their breakthroughs catalyzed rapid standardization efforts in mobile communications and data transmission; they became central components of 3G and later wireless standards, and influenced subsequent families of codes such as LDPC codes. Variants include parallel turbo codes and rate-compatible turbo codes; later work explored serial turbo codes and improved interleaver designs.

While turbo codes deliver near-capacity performance in many settings, they require iterative decoding with substantial computation and memory, which can increase latency and power consumption. Design choices such as interleaver design, constituent encoders, and puncturing patterns influence error floors and practical performance. Ongoing research continues to optimize turbo codes and their variants for diverse channels and hardware constraints.