Total downlink interference is a function of the arrival rate, call duration and bitrate of the users in the network coverage area and hence is a function of user traffic characteristics. Traffic modelling for voice-only DS-CDMA wireless communication is based on Poisson process. For data and multimedia traffics the Poisson process is not an appropriate model. Therefore, the problem of characterizing the total interference in a heterogeneous wireless network is different from that of voice-only networks. The effect of non-Poisson traffic on the performance of wireless network was studied  and the requirements for self-similarity and long-range dependence (LRD) of traffic were examined. It has been shown that the capacity of a DS-CDMA network, designed for Poisson traffic, is substantially reduced when the input traffic has long-range dependence.

In this research, we derive general conditions under which total interference is asymptotically self-similar. These conditions depend on call durations, bit-rate variations, and channel characteristics. We assume a slow-fading channel; it is possible to show that for longer time scales, fast fading is averaged out and the long term characteristics are governed by slow fading. We show that these conditions are more general than the earlier results. Furthermore, we generalize the asymptotical self-similarity of interference, which was conned to data-centric (single service) networks, to heterogeneous (multi-service) networks.