Computer aided implementation of complex algorithms on DSP's using automatic scaling
K. Kassapoglou, Martin Vetterli · 2005
A methodology for transforming complex floating-point algorithms into correct fixed-point DSP programs is presented. In particular, an automatic scaling scheme leading to overflow-free programs is described. Depending on the application, scaling generated automatically may either perfectly fit, or be modified in order to substantially improve accuracy. Each case is illustrated by an example (FFT and Recursive Least-Squares algorithms).