Hi all:
Interactive discussion,.....this is good.

I have looked at interpolation algorithms on lower performance processors that have limited math capability. The various algorithms such as polynomial fitting and spline curve fitting and so forth may give better results on tables of data that represent significant curves, but the computational requirements can bring an 8051 to its knees.

Linear interpolation works better when your data table represents a much more gentle type of curve.

A nice technigue is to plug the data table that you have into a spreadsheet and try the linear interpolation out and see if you get satisfactory results as compared to a 3 or 4 point curve fit algorithm of some type. (MathCad is also a nice tool to use for this analysis).

In the past I have had situations where the sensor data table really needed to have a better scheme than linear interpolation becasue the simpler scheme let to too much error. So what I did was to expand the size of the table to 4x the original by using a polynomial curve fit in MathCad to define new data points. I was then able to plug the new data table into the microcontroller application and use linear interpolation for the sensor and achieve the results I desired.

Michael Karas