Newton-Raphson reciprocal approximation , 1/x and 1/sqrt(x)

The Newton-Raphson approximation of the reciprocal, i.e. 1/x and square root reciprocal is provided along with accuracy testing jigs.

You are now following this Submission

The reciprocal calculation in a real-time controller is quite computationally expensive. The Newton-Raphson approximation of the reciprocal saves on the computational cost at the expense of accuracy. If the reciprocal of an input signal is calculated, with the signal itself having a limited change from one execution cycle to the next, then the previously calculated Newton-Raphson reciprocal serves as an initial "guess" to the new the iteration. This results in a large improvement in accuracy.

The number of iterations can be modified according to the required accuracy.

An analysis can be found in:
Fowler, D. L., and James E. Smith. "An accurate, high speed implementation of division by reciprocal approximation." Proceedings of 9th Symposium on Computer Arithmetic. IEEE, 1989.

Cite As

Gert Kruger (2026). Newton-Raphson reciprocal approximation , 1/x and 1/sqrt(x) (https://www.mathworks.com/matlabcentral/fileexchange/71675-newton-raphson-reciprocal-approximation-1-x-and-1-sqrt-x), MATLAB Central File Exchange. Retrieved .

General Information

MATLAB Release Compatibility

  • Compatible with any release

Platform Compatibility

  • Windows
  • macOS
  • Linux
Version Published Release Notes Action
1.0.1

Removed an unnecessary variable from the reciprocal calculation which was used for performance analysis. The functions are now suitable for use (copy and paste).

1.0.0