Backscattered ultrasonic microstructural noise can be used to estimate grain size in metals. However for normal-incidence immersion measurements the ring-down of the front-wall echo creates a ”dead zone” where backscattered grain noise cannot be quantified. This poses a problem for near-surface grain sizing efforts. In this paper we explore the use of mode-converted 45-degree shear waves for near-surface grain sizing using a water immersion setup. We discuss how to accurately relate grain noise arrival time with depth of sound penetration in the metal. Then for a set of Ni-alloy specimens having near-equiaxed microstructures we correlate various backscattered noise attributes with grain sizes determined from micrographs. These noise attributes include both time-domain and frequency-domain characteristics. The backscattered grain noise attributes correlate well with grain size, and are relatively insensitive to modest changes in the transducer tilt angle.