Our problem isn't our timebase, it's ORDER. Our current DR algorithm is second-order (using position, velocity, and acceleration) AFAIK. Going to a third-order algorithm (using jerk in addition to acceleration, velocity, and position) will yield better DR behavior...