In this paper, an efficient technique to evaluate temporal correlation and transition activity at high level in DSP systems is presented. The method is based on the spectral distribution of signals and has the advantage to require information well known by designer, simple and easy to apply even for complex circuits. With only the knowledge of the input signal spectrum and the transfer function from the input to the circuit nodes allows an efficient estimation of the correltion coefficients. A model for computing transition activity in terms only of spectral distribution is also developed. The spectral analysis gives new criteria to obtain power optimization. Experimental results validating the method are also included in this papera

Peer Reviewed