Explanation of the spectra converter, 3 explanation of the spectra converter – BUCHI NIRCal User Manual

Page 139

Advertising
background image

Chemometrics

NIRCal 5.5 Manual, Version A

139



NOTE
Calibration wavelength: deselect wavelength range, which were not measured with the
original spectra. By several instrument spectra in the same project use the smallest
measured range.

NIRLab spectra have not integer wavenumber; the wave selection is after conversion
rounded to integer.

In case the calibration was calculated with a special wave selection, it is suggested to make
several trials to find the best wavelength selection.

Apply the pretreatments as used in the existing calibration.

NOTE
Pretreatment: use "Absorbance" (log) as additional first data pretreatment by NIRLab N-
200 spectra, in case not "Absorbance inverse" (ilg) was used as first data pretreatment.
This can be used for each calculation with the calibration wizard activating "Keep
preselected Pretreatments for all calibrations".

NIRFlex N-400 spectra have 3 times more data points after the conversion. Using the
same data pretreatment give other result. In case e.g. "1st Derivative BCAP" was used in
NIRCal 4, after the spectra conversion "1st Derivative BCAP Gap2" will nearly give the
same result in NIRCal 5. The same with other derivatives and smoothing.

10

Perform the first calculation (Overview) and adjust the number of primary and secondary
PCs according the existing calibration.

NOTE
Compare the results of the calibrations. The Q-Value should be different for quantitative
calibration, while the calculation equation is improved that leads to slightly different Q-
Values.

11

Increase the robustness of the calibration with spectra recently measured with NIRFlex N-
500.

12

Optimize the calibration and save the best according the Lifecycle.

13

Close and save the project. The created calibration is now available in the NIRWare
Application Designer
.

3.15.3 Explanation of the Spectra Converter

Cubic spline function is used for the spectra conversion. The cubic function will be fitted on the data
stepwise with 3rd degree polynom:

y(x) = a * x

3

+ b * x

2

+ c * x + d


The Y-value for the desired X-values (wavenumber) are extracted from the resulting function curve.

The result is: not a linear connection of the points

Advertising