Carbonyl sulfide (OCS), the most abundant sulfur gas in the Earth's atmosphere, is a greenhouse gas, a precursor to stratospheric sulfate aerosol, and a proxy for terrestrial CO2 uptake. Estimates of important OCS sources and sinks still have significant uncertainties and the global budget is not considered closed. One particularly uncertain source term, the OCS production during the atmospheric oxidation of dimethyl sulfide (DMS) emitted by the oceans, is addressed by a series of experiments in the atmospheric simulation chamber SAPHIR in conditions comparable to the remote marine atmosphere. DMS oxidation was initiated with OH and/or Cl radicals and DMS, OCS, and several oxidation products and intermediates were measured, including hydroperoxymethyl thioformate (HPMTF), which was recently found to play a key role in DMS oxidation in the marine atmosphere. One important finding is that the onset of HPMTF and OCS formation occurred faster than expected from the current chemical mechanisms. In agreement with other recent studies, OCS yields between 9 % and 12 % were observed in our experiments. Such yields are substantially higher than the 0.7 % yield measured in laboratory experiments in the 1990s, which is generally used to estimate the indirect OCS source from DMS in global budget estimates. However, we do not expect the higher yields found in our experiments to directly translate into a substantially higher OCS source from DMS oxidation in the real atmosphere, where conditions are highly variable, and, as pointed out in recent work, heterogeneous HPMTF loss is expected to effectively limit OCS production via this pathway. Together with other experimental studies, our results will be helpful to further elucidate the DMS oxidation chemical mechanism and in particular the paths leading to OCS formation.