Turbo code

Turbo code
Name	Turbo code
Invented	1993
Inventors	Claude Berrou, Alain Glavieux, Punya Thitimajshima
Field	Coding theory
Applications	Wireless communications, deep-space communication, storage

Turbo code is an error-correction coding technique that approaches the theoretical limits predicted by Claude Shannon for reliable data transmission. Developed in the early 1990s, it dramatically improved practical performance in channels such as those encountered by European Space Agency missions and mobile networks led by companies like Nokia and Ericsson. Turbo code spurred rapid advances in digital communications standards including work by the 3GPP and institutions such as NASA and European Telecommunications Standards Institute.

History and development

Turbo code was introduced in 1993 by a team including Claude Berrou, Alain Glavieux, and Punya Thitimajshima while affiliated with institutions linked to Institut National Polytechnique de Grenoble and collaborating laboratories. The invention occurred amid parallel advances by researchers at Bell Labs and teams working on concatenated codes, including influences from the Reed–Solomon work adopted in Voyager program data systems. Early demonstrations compared turbo performance to the Viterbi algorithm implementations used in standards such as Global System for Mobile Communications and prompted revisions in proposals under organizations like International Telecommunication Union and European Space Agency. Recognition of turbo coding led to awards and acknowledgement across conferences held by IEEE and societies such as ACM.

Principles and architecture

Turbo code architecture employs parallel concatenation of two or more convolutional encoders separated by an interleaver; this design builds on concepts from earlier concatenated coding schemes proposed by researchers associated with Richard Hamming-inspired error control. The interleaver, often a pseudo-random mapping, breaks correlation patterns analogous to interleavers used in systems developed by AT&T Bell Laboratories and implemented in hardware by firms like Broadcom. Decoding relies on iterative exchange of probabilistic soft information between component decoders, extending algorithms with roots in the Baum–Welch algorithm family and influenced by belief propagation techniques studied in contexts such as Factor graph theory presented by Frank R. Kschischang. The conceptual leap connected ideas prevalent in signal processing at institutions including MIT and Caltech.

Encoding and decoding algorithms

Encoding uses simple recursive systematic convolutional (RSC) encoders, a structure derived from early convolutional designs tested in projects at NASA Jet Propulsion Laboratory and industrial research at Siemens. The interleaver permutations are chosen to optimize weight spectra, an approach informed by combinatorial analysis from groups such as École Polytechnique collaborators. Decoding employs soft-input soft-output (SISO) modules often implemented with the BCJR algorithm (named for Bahl, Cocke, Jelinek, and Raviv) alongside log-domain variants used in implementations by companies like Qualcomm. Iterative decoding exchanges extrinsic information until convergence criteria—studied in the context of density evolution at laboratories such as Bell Labs Research—are met; complexity trade-offs influenced hardware realizations by Intel and embedded designs in chips by ARM Holdings.

Performance and analysis

Turbo codes achieve bit-error-rate performance near the Shannon limit under additive white Gaussian noise channels; this result was validated in simulations often reproduced in academic groups at University of California, Berkeley and University of Cambridge. Analytical tools include extrinsic information transfer (EXIT) charts developed by researchers associated with EPFL and asymptotic weight enumerator analyses reminiscent of techniques used by teams at University of Illinois Urbana–Champaign. Finite-length performance depends strongly on interleaver design and constituent encoder memory, considerations that guided standards committees like 3GPP and agencies such as European Space Agency when selecting codes for missions and mobile networks.

Implementations and applications

Turbo codes were implemented in software libraries maintained by academic consortia including groups at ETH Zurich and deployed in commercial baseband processors by vendors such as Nokia and Qualcomm. Applications include deep-space links supported by European Space Agency and related missions where forward error correction competes with LDPC code choices evaluated by NASA teams. Mobile telecommunications adopted turbo coding in third-generation cellular standards coordinated by 3GPP and device ecosystems developed by firms like Samsung and LG Electronics. Storage and broadcasting systems evaluated turbo-like schemes in trials led by Thomson Multimedia and research centers at Fraunhofer Society.

Variants and extensions

Extensions of turbo concepts led to serial concatenation, hybrid concatenated schemes, and turbo-like architectures such as repeat-accumulate codes developed in academic labs at Massachusetts Institute of Technology and University of Edinburgh. Low-density parity-check codes, advocated by proponents in the David MacKay community and vetted at institutions like University of Cambridge, emerged as a competitive class with shared analysis tools including belief propagation and EXIT charts. Other variants include adaptive interleavers inspired by algorithms from Stanford University research and multi-dimensional concatenations explored in collaborations with industry partners like Nokia and Ericsson.

Category:Error detection and correction