Classic Doppler equation can only provide the axial velocity of blood flow. To acquire the complete flow vector, estimation of the non-axial flow velocity is essential. For Doppler-bandwidth-based transverse estimation, however, accuracy is limited because of the complex dependence of the Doppler bandwidth on the geometry and the location of the sample volume in the vessel. Specifically, the Doppler bandwidth tends to be overestimated because it is conventionally decided from the difference between maximum Doppler frequency and Doppler shift frequency. The maximum Doppler frequency only depends on the peak flow velocity within the vessel and can be used as a stable parameter in flow estimation. However, the Doppler shift frequency is susceptible to the position of the sample volume and it decreases when the sample volume is not centered within the vessel. The distance between the center of the sample volume and the central line of the vessel is referred to as the position offset of the sample volume. Based on the stable nature of maximum Doppler frequency, a novel method utilizing the differential maximum Doppler frequencies from two parallel beams with different beam widths is proposed to improve the accuracy of transverse estimation. In-vitro experiments were performed to validate the proposed method and results were compared with the conventional method. In this study, a steady flow condition was considered and two 5-MHz pistons were used to generate the two beams with different widths. For the conventional method, it is demonstrated that the Doppler bandwidth is severely overestimated when the position offset is present. For the proposed method, however, the differential maximum Doppler frequency is relatively stable even in the presence of the position offset as long as the sample volume is sufficient in length. Hence, both accuracy and stability of the transverse estimation can be significantly improved by taking advantage of the differential maximum Doppler frequency.