MIMO-OFDM LTE System based on a parallel IFFT/FFT on a multiprocessor platform

This paper proposes a software workflow used to develop and evaluate a real time MIMO-OFDM LTE communication system. The central focus in this work is on OFDM modulation/demodulation functions which induce most of the processing time. To guarantee low latency, low energy dissipation, and high bandwidth requirements, we propose a multicore framework based on a parallel IFFT algorithm. We perform an accurate design space exploration by varying the number of tiles and by changing some NoC parameters. Based on our DSE study, the latency, bandwidth, and total energy dissipation are analyzed to select the best architecture design for the OFDM LTE system. In the worst case of a 20 MHz channel bandwidth, the best architecture selected for the OFDM LTE system is implemented with a hybrid network-on-chip using 16 tiles computing IFFT tasks. It leads to 84% reduction in latency, 73% increase in bandwidth in comparison with traditional OFDM LTE system using a single processing tile for the IFFT task.