In this dissertation the aim was to investigate the usage of algorithms found in computer
science and apply suitable algorithms to the problem of decoding multiple-input multipleoutput
(MIMO) space-time-frequency block coded signals. It was found that the sphere
decoder is a specific implementation of the A* tree search algorithm that is well known in
computer science. Based on this knowledge, the sphere decoder was extended to include
a priori information in the maximum a posteriori probability (MAP) joint decoding of the
STFC block coded MIMO signals. The added complexity the addition of a priori information
has on the sphere decoder was investigated and compared to the sphere decoder without
a priori information. To mitigate the potential additional complexity several algorithms that
determine the order in which the symbols are decoded were investigated. Three new algorithms
incorporating a priori information were developed and compared with two existing
algorithms. The existing algorithms compared against are sorting based on the norms of the
channel matrix columns and the sorted QR decomposition.
Additionally, the zero forcing (ZF) and minimum mean squared error (MMSE) decoderswith and without decision feedback (DF) were also extended to include a priori information.
The developed method of incorporating a priori information was compared to an existing
algorithm based on receive vector translation (RVT). The limitation of RVT to quadrature
phase shift keying (QPSK) and binary shift keying (BPSK) constellations was also shown in
its derivation. The impact of the various symbol sorting algorithms initially developed for
the sphere decoder on these decoders was also investigated. The developed a priori decoders
operate in the log domain and as such accept a priori information in log-likelihood ratios
(LLRs). In order to output LLRs to the forward error correcting (FEC) code, use of the
max-log approximation, occasionally referred to as hard-to-soft decoding, was made.
In order to test the developed decoders, an iterative turbo decoder structure was used together
with an LDPC decoder to decode threaded algebraic space-time (TAST) codes in a Rayleigh
faded MIMO channel. Two variables that have the greatest impact on the performance of the
turbo decoder were identified: the hard limit value of the LLRs to the LDPC decoder and the
number of independently faded bits in the LDPC code.
Dissertation (MEng)--University of Pretoria, 2013.