Double Binary Convolutional Parallel Decoding Of Turbo Codes Research

Based on convolutional coding and iterative decoding, Turbo code is outstanding.Different from the classic one, Double Binary Turbo Code encoder inputs twoinformation bits at a time-slot, enhancing coding rate and decreasing the interleavingdepth. And IEEE802.16m chooses this code for its forward error correction.This paper introduces the application of DB-CTC in Data-Channel Coding whichincludes burst partition, FEC encoding, subblock interleaving, puncture and modulationat the transmitting terminal and has soft demodulation, iterative decoding and harddecision at the receiving terminal. A simulating system is built on these to testify theperformance of DB-CTC, and the Turbo code in LTE will be compared with DB-CTC toshow their difference in coding and decoding.DB-CTC encoder includes precoding and real coding to achieve tail-biting. But atthe decoder, the unknown circular states will adversely affect the performance of thecode. The paper deduced the computation of circular states and then introduced methodsbased on border metrics feedback to decrease the negative influence.When decoding DB-CTC with MAP or its simplified algorithm, the number ofbranch metrics is huge, and so are the computational complexity and redundancy.Branch metrics partition divides each branch metrics into two parts: information factorand parity factor, the memory for the branch metrics can be decreased by half and36add operations and8multiply operations can be reduced.Aiming at reducing the decoding time delay, this paper summarizes only to findthree parallel decoding algorithms in three distinct perspectives. Two-way paralleldecoding could save memories for state metrics by half, through having forward metricsand backward metrics be computed simultaneously; Segments parallel decoding couldincrease decoding speed by n times when code sequence is devided into n segments;Sliding-window (SW) parallel iteration decoding only needs memories for the length ofthe window and could approach the performance of serial iteration with less time delay.Furthermore, the advantages of these algorithms can be put together to produce twodimension or three dimension parallel decoding algorithms.