Recently, rack-mounted devices such as smart-phones were proposed to be installed on vehicles as the sensing source to detect road surface conditions by measuring the vibrations of the vehicles. However, due to lack of models, most proposed solutions need individual sensing system tuning or training based on extensive experiments to capture overall system characteristics and usually the outcomes can't be applied to other systems due to various factors, e.g., the suspension systems of vehicles, the mounting racks or adhering methods, the weight of the onboard device, and even the position on the vehicle. It becomes an obstacle for pervasive deployment of such systems. In this work, we intend to verify the feasibility of modeling vibrations measured by onboard devices by underdamped oscillation model (UOM) through time and frequency domain analyses. If the vibrations of the phone can be modeled by UOM, the measurement uncertainties caused by various factors that affect the sensed vibrations may potentially be eliminated. The vibrations caused by road anomalies could be measured in a relative objective way by the ratio of the vibration caused by road anomalies and the background vibrations. Our experimental results suggest that it is acceptabl to model the vibrations of onboard devices by the UOM.