An algorithm for converting floating-point computations to fixed-point in MATLAB based FPGA design

Abstract

Most practical FPGA designs of digital signal processing applications are limited to fixed-point arithmetic owing to the cost and complexiry of floating-point hardware. While mapping DSP applications onto FPGAs, a DSP algorithm designer, who often develops his applications in MATLAB, must determine the dynamic range and desired precision of input, intermediate and output signals in a design implementation to ensure that the algorithm fidelity criteria are met. The first step in a flow to map MATLAB applications into hardware is the conversion of the floating-point MATLAB algorithm into a fixed-point version. This paper describes an approach to automate this conversion, for mapping to FPGAs by profiling the expected inputs to estimate errors. Our algorithm attempts to minimize the hardware resources while constraining the quantization error within a specified limit