significantly reduce rounding error when doing framepos_t + BBT math