Authors:
César Benavente-Peces
1
;
Andreas Ahrens
2
;
José M. Pardo-Martín
1
and
Fco. Javier Ortega-González
1
Affiliations:
1
Universidad Politécnica de Madrid, Spain
;
2
University of Technology Business and Design, Germany
Keyword(s):
Multiple Input Multiple Output (MIMO) systems, Singular Value Decomposition (SVD), Bit Allocation, Power Allocation, Wireless Transmission, Finite Word Length, Fixed Point Arithmetic.
Related
Ontology
Subjects/Areas/Topics:
Biomedical Engineering
;
Biomedical Instruments and Devices
;
Detection, Decoding and Diversity Techniques
;
Emerging Technologies
;
MIMO Systems and Techniques
;
Telecommunications
;
Wireless and Mobile Technologies
;
Wireless Information Networks and Systems
Abstract:
This paper is devoted to analyze the error met when computing the singular value decomposition (SVD) of the gain channel matrix in a MIMO communication system and using fixed-point arithmetic for the calculations. The study is focused in the case of the SVD implementation for modulation-mode and power assignment in non-frequency selective MIMO, and M-ary Quadrature Amplitude Modulation (QAM). It is demonstrated that not necessarily all MIMO layers must be activated. The combination of the CORDIC algorithm and look-up tables seems to be a good solution for this task since it can efficiently compute the singular value decomposition. The paper highlights the characterization of computation errors and shows the performance losses produced by the use of approximations in the eigenvalues and eigenvectors calculation, including the effects produced in the power assignment process.