IASIM 2018 Advertisement

Matching axes in calibration transfers

ianm's picture

17. Matching axes in calibration transfers.

I have a question to address to your e-mail group.

The question is:

What should the scales be for the X and Y be when matching a slave instrument to a master instrument ? The Y-axis could be in transmittance or absorbance, and the X-axis could be in wavenumbers or wavelength. For both we have nonlinear relationships. Is there a best choice for matching instruments and calibration transfer ? Dr. Siesler presented a paper at Pittcon in which he compared a dispersive instrument with an FT-NIR, and showed a wavenumber difference of 1.5 cm-1 in the 4200 cm-1 region and a 13cm-1 difference in the 9000 cm -1 region. This looks like a big nonlinear effect but it really is less than 1.5 nm difference for the respective regions on a wavelength scale. This does not prove that the wavelength scale is better, but it does illustrate the problem.


Karl Norris

From: Howard Mark

Bruce - my comments for what they're worth:

Re the absorbance vs transmittance question: the conversion between them is straightforward, so going from one domain to the other is almost trivial, and so putting both instruemnts into the same domain is also trivial. For good and valid reasons, the absorbance domain is preferable for quantitative work. My intuition tells me that the same reasons would argue in favor of that domain for the transfer process, but I don't know that this question has ever actually been addressed.

The question of wavelength vs wavenumber (frequency) conversion is much trickier. The problem there is not the errors per se (those could be corrected); the problem is that the inherent nature of gratings is to disperse linearly with wavelength, while the inherent nature of FTIR instruments is to "disperse" linearly with wavenumber. Thus, while you could convert one unit to the equivalent value of the other, the actual measured data will still not be the same, because they are measured at different equivalent bandwidths, which equivalency furthermore changes across the spectrum. On top of this is the difference in the shape of the bandpass, upon conversion this shape would also change. Theoretically you could compute the necessary corrections if you had accurate enough information, but in practice I think that you cannot

A) know them

B) accurately enough

even in principle, because in both directions you need accurate information at less than the resolution of one or both instruments, at at least one end of the spectrum.