VLSI Architecture for Broadband MVDR Beamformer

Background/Objectives: The main objective of this proposed work is to investigate the significance of adaptive beamforming technique and to develop an efficient VLSI architecture for broadband MVDR beamformer in the field of medical ultrasound imaging. Methods/Statistical Analysis: The proposed algorithm is Minimum Variance Distortionless Response (MVDR) for near field beamforming of broadband data which gives better contrast and resolution compared to conventional Delay and Sum (DAS) beamformers and is implemented in frequency domain. MVDR beamformer minimizes the output power by allowing the desired signal to pass undistorted with unity gain. The solution for this optimization problem, involves correlation matrix inversion, which is the challenging part of MVDR algorithm. Findings: MVDR algorithm is applied by finding the inverse of correlation matrix which is a complex matrix. Here four elements are used for simplicity. Calculation of inverse complex matrix is the challenging part of MVDR algorithm where different methods are being used. Here QR decomposition is used which follows givens rotation algorithm. The paper demonstrates the formal verification of the proposed work. Final result is compared with the golden reference model which is designed using FIELD II scanner in Matlab. From the results it’s seen that, MVDR beamformer gives a pencil like beamform which shows high resolution and better contrast when compared to DAS. The timing constraints and device utilization parameters are obtained from synthesis report of final architecture designed in FPGA. Conclusion/Improvements: MVDR algorithm gives a pencil like beamformer output with reduced main lobe width and reduced side lobe level. By upgrading number of elements from 4 to 64, it can be made real time.