The invention is an instrument for sensing the state of polarization (SOP), and for transforming the SOP of a beam of light from an incident continuously-varying arbitrary SOP to a desired exit SOP, using a polarization compensator under feedback control. A polarization sensor uses two or three samples of a beam to sense the Poincare sphere latitude and longitude error in SOP. A polarization controller adjusts the SOP of light, which is then sensed by the polarization sensor, which develops signals to drive the polarization compensator using feedback methods. Unlike prior-art systems, the feedback seeks a mid-point rather than an extremum in the sensed signals, so there is no sign ambiguity in the feedback control. Further, the sensor signals indicate orthogonal displacements in SOP that correspond to specific elements in the polarization controller, so there is no ambiguity as to which element needs adjustment in order to correct a given error in SOP. The result is a high-performance instrument with high-speed response to changes in the SOP of the incident beam, and very low error in the exit SOP.