The European EAR-IT project addresses “real-life” experimentations of intelligent acoustic for supporting high societal value applications in a large-scale smart environment. For instance a city emergency center can request on-demand acoustic data samples for surveillance purposes and management of emergencies. In this paper, we will present experimentations on streaming encoded acoustic on the SmartSantander large scale test-bed. We will present the various audio hardware that were developed to meet the constraints of audio capture and transmission with low-resources devices. We will highlight the main sources of delays and will show how multi-hop streaming of acoustic data can be achieved by carefully taking into account these performance limitations with appropriate audio aggregation techniques.