Wavelength Accuracy / Reproducibility
How exactly can the wavelength be determined, e.g. of line of monochromatic light?
If something is stable and thus reproducible it can be calibrated accurately. Furthermore, accuracy is of course limited by the pixel dispersion, i.e. the less bandwidth one pixel covers, the more accurate the wavelength info can be represented. The design of the spectrometer broadens even a physical very narrow line, so multiple pixels are illuminated. This allows Peak wavelength fitting, which is limited to about 1/10 of the bandwidth per pixel (pixel dispersion) (see Sub-pixel resolution).
While the pixel dispersion is a design parameter (what spectral rage is covered by the whole detector array) stability is a matter of design quality, material selection and mounting techniques. A thermal drift scales with the overall spectral range covered, therefore, only relative drift values present decent quality criteria.
The stability of spectrometer setup is challenged by by mechanical instabilities and stress, thermal expansion, material deterioration. To minimize effects, proper materials with low thermal expansion coefficient have to be selected (e.g. ceramics, titanium, Dispal), superfluous components have to be avoided (e.g. imaging gratings to avoid additional lenses / mirrors), and stable mounting (e.g. cementing) to be applied.
In case a spectrometer body is made very stable, a good wavelength calibration leads to a high wavelength accuracy. Such a calibration is performed by using line sources, with a reasonable number of spectral lines over the desired spectral range. HgAr sources are a kind of standard for the UV-VIS range, sometimes in combination with Holmium filters. From the measured and known lines, the center positions of the individual pixels can be calculated. Good results are achieved by using a polynomial fit procedure of third or fourth order. A stable spectrometer is permanently calibrated, while a good calibration is an art by itself. Back to top▲
The calibration is typically not the limiting factor for long-term absolute accuracy but overall stability of the unit. Back to top▲
A calibration requires that the optical axis of the spectrometer is well defined. In case of a classical slit with removable light guide this is not always given. By reattaching the fiber-optic part, small mechanical tolerances may lead to an alteration of the optical axis and thus an error within the wavelength calibration. A permanently attached fiber-optic input overcomes this issue.
The calibration should be done with the grating fully illuminated, for in principle all grooves are involved in the generated of a spectrum. At significant lower NA of incoming light, a grating area too small might see light, the generated spectrum may differ in position and resolution from a fully illuminated one; therefore, the calibration is most likely not. Back to top▲