Trellis aggregation for high code rate turbo-code

Abstract

In high rate Turbo-code, the heavy punctionring makes long sequence of treillis without any bit of redondance. We show in [1] that for a m-states convolutional decoder, A redondance free section of length (m-1) * P, where P is an integer, can be exactly reduced to a section of treillis of length m-1. In [2], we show that for an M state Turbo decoder, among the L compressed trellis stages, only m = 3 or even m = 2 are necessary. The so-called m-min algorithm can either be used to increase the throughput for decoding a high rate turbo-code and/or to reduce its power consumption.

This work has been supported by the GIGADEC project (Brittany region funding), as well as the CPER project PALMYRE II (Brittany region and FEDER funding). The objective of the GIGADEC project was to define new architectures for very high speed Turbo Code (project in collaboration with Turbo-Concept and Telecom Bretagne.)

More information

The matlab file mentioned in [2] (reference [9]) can be download here: here)

In [2], table IV is extracted from this figures:

Figure 1: FER as a function of Eb/No. HSPA turbo-code of length K=5144 for several code rates. Sliding windows of size W is used combined with acquisition sequence of length W (combined with the Next Iteration Initialisation method).

Figure 2: FER as a function of Eb/No. HSPA turbo-code of length K=5144 and code rate r = 0.98. Several values of windows length are used.

Trellis aggregation for high code rate turbo-code

Abstract

More information

Related papers