We present a novel experimental apparatus for the capture and replay of physical controls (mechanical knobs), as well as a set of acquired models and a design discussion related to the characterization approach taken here. Our work extends existing research by addressing problems surrounding identification of physical controls, including sensor gripping techniques for arbitrary target devices; and improved hardware and algorithm combinations for finer capture resolutions. Models were acquired from 5 real knobs, based on 2nd order model fits to torque and kinematic results of swept-sine excitations.