Analyses of absorption from disk winds and atmospheres in accreting compact objects typically treat the central emitting regions in these systems as point sources relative to the absorber. This assumption breaks down if the absorbing gas is located within $few\times 1000\cdot GM/c2$, in which case a small component of the absorber's Keplerian motion contributes to the velocity-width of absorption lines. Here, we demonstrate how this velocity-broadening effect can be used to constrain the sizes of central engines in accreting compact objects via a simple geometric relationship, and develop a method for modeling this effect. We apply this method on the Chandra/HETG spectra of three ultra-compact and short period neutron star X-ray binaries in which evidence of gravitationally redshifted absorption, owing to an inner-disk atmosphere, has recently been reported. The significance of the redshift is above $5\sigma$ for XTE J1710$-$281(thiswork)and4U1916$-$053,andisinconsistentwithvariousestimatesoftherelativeradialvelocityofeachbinary.Forourmostsensitivespectrum(XTEJ1710$-$281),weobtaina1$\sigma$ upper bound of 310 $km$ $s-1$ on the magnitude of this geometric effect and a central engine of size (or, at the $3\sigma$ level). These initial constraints compare favorably to those obtained via microlensing in quasars and approach the sensitivity of constraints via relativistic reflection in neutron stars. This sensitivity will increase with further exposures, as well as the launch of future microcalorimeter and grating missions.