Considering traffic dynamics greatly improves noise estimation in urban area. This can be achieved by coupling a dynamic traffic model with both noise emission laws and sound propagation calculation. Determining the relevant noise source and traffic representations to estimate classical noise descriptors (LAeq and statistical descriptors) near traffic signals has been recently studied. This research topic is extended in this paper to more specific descriptors which are able to capture noise dynamics at the traffic signal scale, for usual urban traffic situations (upstream, in front of, and downstream a traffic signal) and different distances from the road (5.5, 10 and 15 m). It appears that 14m-line sources ensure an estimation of all descriptors with errors below 2 dB(A) if traffic dynamics is precisely described. Macroscopic and microscopic car-following models are both relevant to highlight noise dynamics triggered by the traffic signal, but some differences between those traffic representations are observed.