Snowmelt processes on sea ice are the key drivers determining the seasonal sea-ice energy and mass budgets. Around Antarctica, snowmelt on pack ice is weak and very different than in the Arctic, with most snow surviving the summer. It is therefore important to understand the mechanisms that drive snowmelt, both at different times of the year and in different regions around Antarctica. Doing so, we compile time series of snowmelt onset dates on perennial Antarctic sea ice from 1992 to 2014 using active microwave observations from European Remote Sensing Satellite (ERS-1/2), Quick Scatterometer (QSCAT) and Advanced Scatterometer (ASCAT) radar scatterometers. Describing snow melt processes, we define two transition stages: A weak backscatter rise indicating the initial warming and metamorphism of the snowpack (pre-melt), followed by a rapid rise indicating the onset of thaw-freeze cycles in the interior snowpack (snowmelt). We compare these with pan-Antarctic temporary snowmelt onset dates in the uppermost snowpack retrieved from diurnal variations in the brightness temperatures from passive microwave (PMW) observations. Results show that QSCAT Ku-band (13.4 GHz signal frequency) derived pre-melt and snowmelt onset dates are earlier by 25 and 11 days, respectively, than ERS and ASCAT C-band (5.6 GHz) derived dates. Snowmelt onset dates from the shortwave PMW observations (37 GHz) are later by 13 and 5 days than those from the scatterometers, respectively. Based on the observed successive timing of melt events retrieved from different sensors and microwave bands, we developed a conceptual model of the temporal evolution of snow temperature and metamorphism and their effect on different microwave wavelengths during the spring/summer transition. These results suggest that future multi-frequency microwave satellite missions could be used to resolve melt processes throughout the vertical snow column. Overall, results show that the magnitude and timing of seasonal and diurnal variations in Antarctic snow on sea ice are highly dependent on latitude, with earlier and more frequent snowmelt in the north. All retrieved melt onset dates show large interannual variability but no significant decadal trends.