GE Analytical Instruments ©2016
111 of 270
DLM 68100-09 EN Rev. A
Chapter 5: Calibration and Verification
Certain recommendations can be made about the use of two-point calibrations:
•
The concentrations of the points should bracket the expected range of concentrations of the samples.
The accuracy of the calibration will be maximized when the concentrations of the points are close to
the concentrations of the samples.
•
Because Auto Dilution contributes some uncertainty to the calibration, the calibration will be most
accurate if the calibration at point 2 is not done using Auto Dilution.
•
It is recommended that the 5,000 ppm range and higher ranges be calibrated using at least two points.
•
If you plan to use the Auto Range feature, calibration of the 50,000 ppm range is important because the
preliminary measurement used to determine the appropriate analysis range is performed in this range.
For best performance of the Auto Range feature, it is recommended that the 50,000 ppm range be
calibrated using a six-point, point-to-point calibration (see
). The recommended points are:
Reagent Water, 100 ppm, 1,000 ppm, 5,000 ppm, 20,000 ppm and 50,000 ppm.
Calibration with More than Two Points
When the calibration is performed with three to six points (multi-point calibration), the user has the choice of
having the Analyzer calculate the calibration constants from a linear fit of the points, or from a point-to-point fit.
Linear Fit Multi-Point Calibration
The advantage of selecting the linear fit is that the calibration is affected less by uncertainties in the
measurement, or errors in the concentration, of any one point.
depicts a linear fit calibration with 6 points. The point having the lowest concentration is measured
first, and the points must increase in concentration at each step. Point 1 can either be reagent water or a higher,
known concentration.
In either case, the Analyzer performs a linear regression to calculate the intercept, a, and slope, b, of the
calibration:
C = a + b*R
A correlation coefficient, R
2
, is also calculated. The correlation coefficient is a statistical measure of agreement
between the measured values (R) and standard concentration (C).
