This should provide more stable stepping, preventing floating-point errors from accumulating on each step/sample.