How do turbo codes achieve near-Shannon-limit performance?

Turbo codes use parallel concatenation of two recursive systematic convolutional (RSC) encoders separated by an interleaver. Key innovations: Interleaver decorrelates inputs to second encoder, enabling iterative decoding; Soft-output decoders (BCJR algorithm) exchange extrinsic information between component decoders; Iterations refine reliability estimates, approaching MAP detection. Design considerations: Interleaver design (random, S-random), puncturing for rate adjustment, and convergence analysis (EXIT charts). Turbo codes come within 0.5dB of Shannon limit at moderate block lengths. Used in 3G, 4G, and deep-space communication.