Recent studies have shown that assisted atomization depends on the development of two successive instabilities. The mean drop diameter can be predicted as a function of the characteristic lengths of these instabilities. In the first part predictions of the KH model are compared with experiments. A discrepancy is found for low dynamic pressure ratios and/or high liquid velocities. In the second part we investigate a method of passive control of the transverse instability, with metal needles placed on the splitter plate. This method is expected to be relevant at low gas velocity, when atomization is incomplete. Preliminary results show that for a transverse spacing of the obstacles corresponding to the most amplified wavelength, the atomization process is improved in the case of low gas velocities. If however the spacing is larger than the transverse instability wavelength, the performances of the atomization mechanism are degraded.