We here address the problem of computing the correlation of a PWM signal that carries information and is possibly chaotically jittered in order to avoid high peaks in the power spectrum. A restricted analytical formula is given along with a more general semi-analytical procedure relying on a low-complexity Monte-Carlo estimation. Based on the developed tools, it is shown that the chaos-based frequency modulation of the timebase of the PWM signal is able to cancel the asymptotically periodic trend of the correlation thus reducing the impulsive part of the power spectrum.