Bit-level soft-decision decoding of Reed-Solomon codes

HE practical importance of Reed-Solomon codes is well established (see [1]-[7]). The application of Reed-Solomon codes spans from deep-space communication standard [2] -[4] and over the air teletext broadcast [5] to frequency-hop spreadspectrum systems [6] and optical communication [7]. Harddecision decoders for RS codes are readily available using algebraic decoding algorithms. Such decoders have been implemented and operate at rates above 120 Mb/s. Soft-decision decoding of RS codes is, however, an entirely different matter. The advantage of soft decision over hard-decision decoding is adequately established in many works [8]-[IO] (for graphs of soft-decision coding gain versus SNR for various RS codes see, for example, [IO]). Yet even though the decoder can be often supplied with reliable softdecision data relatively easily [I], the high complexity of optimal soft decoding makes full utilization of such data prohibitive. In fact, the available soft-decoding algorithms, such as Forney's generalized minimum distance decoding [ I l l and others 1121, [13], make use of soft-decision information only on the byte level. Namely, the confidence values of the received bits are processed in one way or another [12] to generate the average confidence value of the symbol, which is then transferred to the decoder. Thus, the bit-level soft-decision information is lost. In this context Berlekamp et ai. [ l ] state that "the major drawback with RS codes (for satellite use) is that the present generation of decoders do not make full use of bit-based soft-decision information." In this paper, we develop Reed-Solomon decoder that makes use of bit softdecision information. The proposed decoding algorithms are in many