On the theory and performance of trellis termination methods for turbo codes

Author

Hokfelt, Johan ; Edfors, Ove ; Maseng, Torleiv

Author_Institution

Fiberless Soc. AB, Lund, Sweden

Volume

19

Issue

5

fYear

2001

fDate

5/1/2001 12:00:00 AM

Firstpage

838

Lastpage

847

Abstract

The performance of a turbo code can be severely degraded if no trellis termination is employed. This paper investigates the implications of the choice of trellis termination method for turbo codes, and explains the origin of the performance degradation often experienced without trellis termination. An efficient method to derive the distance spectrum of turbo codes for different trellis termination methods is presented. Further, we present interleaver design rules that are tailored to each termination method. Using interleavers designed with these restrictions, we demonstrate that the performance difference between various termination methods is very small, including no trellis termination at all. For example, we demonstrate a turbo code with a 500-bit interleaver that exhibits no sign of an error floor for frame error rates as low as 10^-8, even though no trellis termination is employed

Keywords

convolutional codes; decoding; error statistics; interleaved codes; spectral analysis; trellis codes; turbo codes; decoder; distance spectrum; error floor; feedforward convolutional encoders; frame error rates; interleaver design rules; performance degradation; recursive convolutional encoders; trellis termination methods; turbo codes; Convolutional codes; Degradation; Design methodology; Error analysis; Error correction; Tail; Termination of employment; Turbo codes;

fLanguage

English

Journal_Title

Selected Areas in Communications, IEEE Journal on

Publisher

ieee

ISSN

0733-8716

Type

jour

DOI

10.1109/49.924868

Filename

924868