Teaching DSP software development: from design to fixed-point implementations

In this paper, a digital signal processing (DSP) software development process is described. It starts from the conceptual algorithm design and computer simulation using MATLAB, Simulink, or floating-point C programs. The finite-word-length analysis using MATLAB fixed-point functions or Simulink follows with fixed-point blockset. After verification of the algorithm, a fixed-point C program is developed for a specific fixed-point DSP processor. Software efficiency can be further improved by using mixed C-and-assembly programs, intrinsic functions, and optimized assembly routines in DSP libraries. This integrated software-development process enables students and engineers to understand and appreciate the important differences between floating-point simulations and fixed-point implementation considerations and applications.