Analytical estimation of transition activity for DSP architectures

We present an analytical method to estimate the average number of transitions, T, for a given DSP architecture. First, an exact relation between the transition activity, probability, and autocorrelation for a 1-bit signal is derived. Next, we estimate word-level signal transition activity using word-level signal statistics, the signal generation model, and the signal encoding. Input signal statistics are propagated through the DSP operators in a system and then employed to estimate T. Simulations with filters result in errors in T between 1% and 12%. Also, the transpose form is shown to have fewer signal transitions than the direct form for the same input.