Title :
Efficient fixed-point implementation of linear equalization for cooperative MIMO systems
Author :
Mehlhose, Matthias ; Schiffermuller, Stefan
Author_Institution :
Fraunhofer Inst. for Telecommun., Heinrich-Hertz-Inst. (HHI), Berlin, Germany
Abstract :
We describe an efficient fixed-point calculation of the weight matrices for a linear MMSE equalizer in a MIMO-OFDM system intended for the large numbers of antennas typical for cooperative base stations. Our real-time implementation is numerically stable also in critical channels with reduced rank. Implementation loss due to the 16 bit fixed-point arithmetic is noticeable only in high SNR range above 30 dB. A speed optimization technique is proposed splitting the algorithm in several dot products which can be highly optimized. Using 3 parallel C64x+ DSP cores running at 1 GHz each, equalization of a 12 × 12 MIMO-OFDM system with 1200 subcarriers can be realized in less than 1 millisecond. Compared to a floating-point implementation, at the same performance power consumption can be reduced by a factor of 6.
Keywords :
MIMO communication; OFDM modulation; antenna arrays; cooperative communication; digital signal processing chips; equalisers; fixed point arithmetic; least mean squares methods; optimisation; power consumption; telecommunication power management; MIMO-OFDM system; antenna arrays; cooperative MIMO systems; cooperative base stations; fixed-point arithmetic; floating-point implementation; frequency 1 GHz; linear MMSE equalizer; linear equalization; parallel C64x+ DSP cores; power consumption; speed optimization; weight matrices; word length 16 bit; Artificial neural networks; Complexity theory; Matrix decomposition;
Conference_Titel :
Signal Processing Conference, 2009 17th European
Conference_Location :
Glasgow
Print_ISBN :
978-161-7388-76-7