Application of weighted least squares to OSL vector error correction

The use of least squares has been beneficial in providing more accurate one-port calibrations. Commercially available ECAL units have taken advantage of this method. A more generalized application of a least squares method, the weighted least squares method provides benefit when the models for the calibration standards are not all trusted equally. The use of the weighted least squares provides a method of discounting the effect of calibration standards as the model accuracy degrades instead of abruptly dropping the use of the standard outside of a specified frequency range avoiding discontinuities in subsequent measurements. A detailed view a 1.85 μm calibration kit demonstrates the improved results possible using a weighted least squares method. The paper includes the description of a proximity function that enhances the results when the response of two or more standards begin to cluster. Additional enhancements due to databased models over traditional polynomial models is also presented.