Reduced Complexity Decoding of Space Time Trellis Codes in the Frequency Selective Channel
Degree GrantorUniversity of Canterbury
In this work a new iterative approach has been suggested for decoding Space Time Trellis Codes (STTCs) in the frequency selective Multiple Input Multiple Output (MIMO) channel. The objective of this thesis has been to investigate the performance of the approach and determine what parameters affect its performance. The proposed method uses the Partitioned Viterbi Algorithm (PVA) as an equalizer for the MIMO system. The equaliser provides soft outputs which are then used by a STTC decoder to estimate the transmitted data sequence. To reduce error propagation between the two Viterbi based algorithms, an interleaver is introduced.
To further increase the performance, an iterative approach is used, where the decoded data sequence is re-encoded and used by the PVA to improve interference cancellation. This is similar in concept to turbo equalisation. Both the PVA equaliser and STTC decoder have been adapted to provide soft outputs using a Soft Output Viterbi Algorithm (SOVA).
Simulations of a NT = 2 transmit antenna, NR = 2 receive antenna MIMO system have been performed for both 4PSK and 8PSK constellations. It is shown that the iterative procedure achieves a performance within 2.0dB of Maximum Likelihood (ML) decoding, at a FER of 10-2. However, the iterative approach suffers a small diversity loss. It is also shown that the complexity of the iterative approach is far lower than ML decoding. For example, a 16 state 4PSK STTC can be decoded using the iterative approach, with equal performance and less complexity, than a 4 state 4PSK with ML decoding.
It is also shown that for a large diversity system (rNR>3), where is the rank of the STTC, that codes designed using the trace, or Euclidean distance criteria, suggested by Chen et.al.  are superior to the rank and determinant criteria used by Tarokh
et.al.  and Baro et.al. . Other factors investigated that affect the performance are the choice of interleaver, number of iterations, soft information and the size of the MIMO system. The use of a PVA or similar equaliser, coupled with an outer code could be used to increase the effective user bandwidth of the MIMO channel. Standard convolutional codes could be used with the equaliser to improve performance in an iterative approach. More research is required to investigate the performance and complexity of such an approach.