Quantization analysis tool for fixed-point implementation of real time algorithms on the TMS320C5000

Digital Signal Processing (DSP) chips are getting faster, with higher available performance and lower power consumption. Simultaneously, the complexity of algorithms that are being implemented on these devices is increasing rapidly. Currently, Texas Instruments (TI) has targeted the TMS320C5000, fixed point (16 bits), family of DSPs to be the most power efficient DSP engine. Implementing complex algorithms on a fixed-point platform is not a trivial task. Fixed-point programs require variables to have a quantization format, and prevent (minimize) overflow and underflow quantization errors The process of converting a successful floating-point implementation of an algorithm to fixed-point can be a long and tedious process. This paper presents software tool (Fast Quantization Tool (FaQT)), which simplifies the process of converting an algorithm from floating-point to fixed-point. The paper also presents the application of FaQT to a Noise Reduction (NR) algorithm used in speech processing for cell-phone and other audio applications.