This work describes the implementation of a wireless sensor network (WSN) to locate an acoustic source in two dimensions. The network is made up of three sensor nodes located at the corners of a right triangle; a fourth node in the vicinity of the array acts as a base station providing connection to a notebook; the acoustic source is assumed to be within the rectangular area defined by the three sensors. To detect the sound wave, an analogue circuit activated by sound intensity was designed to cause a hardware interrupt to the nodes. The localization uses the method of the Time Difference of Arrivals (TDoA) of the sound wave to each node pair to build an equation system with three hyperbolic equations, and whose solution is the estimated location of the acoustic source. The non-linear system is solved using an optimization-based numerical method. Although the time synchronization error of the implemented WSN is below 10 microseconds, which makes TDoA suitable, other uncertainties, such as in the actual sound speed and the presence of breeze, does systematically influence the results. The localization errors found in experiments in open environments were below 4%. Specifically, for a 10m×10m area, the worst case deviations are below 44cm for the 90% of the measurements. Finally, the WSN uses IRIS nodes (2.4GHz) which were programmed with TinyOS, and the localization algorithm was fully implemented in JAVA.