This article studies the positioning problem for wireless networks when TDOA measures are used and the reference anchor node is not previously known. We carried out various experiments to show the impact on accuracy when a poor selection of this reference is achieved. Furthermore, we study the use of SNR at receivers as a mean to proper select the closest node as the reference anchor, previously to perform mobile positioning. An appropriate mechanism to perform this selection is provided within a simulation platform built to study network-based positioning using space-time diversity in realistic conditions. This approach shows that the use of a signal that measures or estimates the received power is a viable mechanism for the proper selection of the anchor node even in a shadowed environment, avoiding the severe degradation that involves a poor selection.