Implementation of distributed arithmetic-based symmetrical 2-D block finite impulse response filter architectures

Area of Science:

• Digital Signal Processing
• VLSI Design
• Filter Architectures

Background:

• Efficient implementation of two-dimensional (2-D) Finite Impulse Response (FIR) filters is crucial for various signal processing applications.
• Existing architectures often face challenges in optimizing area, power consumption, and throughput.
• Exploiting filter coefficient symmetries can lead to reduced hardware complexity.

Purpose of the Study:

• To propose and implement efficient 2-D FIR filter architectures utilizing block processing and exploiting filter coefficient symmetries (diagonal and quadrantal).
• To reduce the number of multipliers and optimize memory usage for improved area and power efficiency.
• To evaluate the performance of the proposed architectures compared to existing methods.

Main Methods:

• Development of 2-D FIR filter architectures incorporating block processing for enhanced throughput.
• Utilization of distributed arithmetic (DA)-based multiplication with dual-port memory-based lookup tables (DP-MLUTs) to minimize area and power.
• Implementation using Verilog HDL and synthesis in 45nm technology, with performance analysis including area, delay, and power consumption.

Main Results:

• The proposed architectures demonstrate significant improvements in power, area, Area Delay Product (ADP), and Power Delay Product (PDP) compared to existing designs.
• The 2-D block Quadrantal Symmetry Filter (QSF) with block size 4 achieved reductions of 58.94% in power, 59.5% in area, 48.44% in ADP, and 47.78% in PDP.
• Memory reuse and sharing techniques were employed to reduce register count and circuit complexity.

Conclusions:

• A novel, area-power-efficient DA-based 2-D block FIR filter architecture with symmetries has been successfully realized.
• The incorporation of symmetries and DP-LUT-based multipliers effectively minimizes hardware resources.
• The proposed filter architecture is highly suitable for applications requiring fixed filter coefficients due to its efficiency.