Implementation of FPGA based Fast DOA Estimator using Unitary MUSIC Algorithm
M. Kim · Medical Entomology and Zoology · 2003
This paper proposes the practical implementation of DOA estimation system using FPGA (Field Programmable Gate Array) that is a key technique in the realization of the DOA-based adaptive array antenna for cellular wireless basestation. It incor- porates spectral unitary MUSIC (MUltiple SIgnal Classification) algorithm, which is one of the representative super resolution DOA estimation techniques (1). This paper describes the way of DSP design and real hardware implementation of the unitary MUSIC algorithm. This system achieves the high performance in the eigenvalue decomposition (EVD) and MUSIC angular spectra computation with Cyclic Jacobi processor based on CORDIC (COordinate Rotation DIgital Computer) (2) and spatial DFT (Discrete Fourier Transform), respectively. All DSP functions are computed by only fixed-point operation with finite bit-length to meet the requirements of fast processing and low power consumption due to the simplified and optimized architecture. Exploiting adaptive array antenna technologies, the wireless system capacity will be dramatically increased and the harmful effects by multipath fading can be combated as well. From the theoretical point of view, many useful algorithms of the adaptive array antenna techniques need DOAs (Directions Of Arrival) of desired and interferer signals in advance. Of course, the practical researches often have used Wiener-solution based algorithms like LMS and RLS employing a temporal reference signal instead of the DOA informations, while the DOA-based systems exploit the exact DOAs in a beamformer to separate the desired signal from interferers spatially. However the DOA-based systems have many advantages over the conven- tional temporal reference based solutions. For exaple, they are more applicable to the downlink solution thanks to the exact directional information. And the performance of DOA-based beamforming is superior to that of other types of algorithms for small angular spread, while it has time-consuming task of DOA estimation (3). In order to implement such a DOA- based system, the most time-consuming DOA estimation step shoud be processed as fast as possible. But such processing has been very difficult to realize in the practical systems from the lack of cost effective digital processing devices to solve the hard computational burden. We believe that general Von Neumann architecture processors can never usually meet the requirements of the fast and compact architecture and low power consumption at the same time. Thus, in this paper, the FPGA based DSP (Digital Signal Processing) design and hardware implementation of the fast DOA estimator will be presented. It can be applied to cel- lular wireless basestation for DOA-based beamforming and a realtime DOA monitoring system usefully, if it is tuned up appropriately correspoding to the appied environment. It incorporates unitary MUSIC (MUltiple SIgnal Classification) algorithm, which is one of the representative super resolution DOA estimation techniques. MUSIC based algorithm has many advantages in the real hardware implementation due to its simplicity compared with other well-known subspace based techniques like ESPRIT. However, there still remains the computational complication of the complex number arithmetic, which is a great distress to the fast and compact architecture. With a unitary transform, the eigendecomposition of the cor- relation (or covariance) matrix in the MUSIC algorithm can be solved with real number only (1) (6). The unitary MUSIC processor (UMP) performs all DSP functions with only fixed- point operation with finite bit-length in order to meet the requirement of fast processing and low power consumption by simplified and optimized architecture. This system performs the fast computation of EVD and MUSIC angular spectra with Cyclic Jacobi processor based on CORDIC (COordinate Rotation DIgital Computer) (2) and spatial DFT (Discrete Fourier Transform), respectively.