This figure shows an example of peak position drift caused by temperature drift of the spectrometer at different temperatures. The spectral peak is not symmetrical because the number of pixels that make up the spectral peak is very small, and under the influence of temperature drift, the highest pixel of the spectral peak moves one pixel from left to right.
At this point, we have two methods to evaluate the drift of this peak position. 1. Check the position of the highest pixel point. Obviously, the peak position has shifted by one pixel, which means approximately 0.15 nm. 2. Fit the peak values before and after drift, and find the fitted peak point between two pixels. The drift of the fitted peak point is about 0.07nm So which of these two methods is correct? We believe the second one is correct, reflecting the true drift of the peak position. The reasons are as follows. 1.High resolution results in a low pixel sampling rate for spectral peaks, and directly observing the shape of a pixel does not reflect the actual shape of the peak. 2.The spectral peak is physically closer to a symmetrical distribution, and the shift in peak position will not cause a change in peak shape. 3.By fitting all the pixels that make up the spectral peak, the peak shape obtained is closer to the physical reality of the peak.
Therefore, using fitted peak shapes to determine peak positions and evaluate spectrometer drift is the correct method.
Is it necessary to do so in practice? It depends on the actual situation. If 1.Low spectral resolution and high sampling rate of spectral peaks require more pixels to form a peak, and these pixels themselves are connected in a smoother and more symmetrical peak shape. Then, the position of the highest pixel or between two pixels can be directly found, and the peak position can be confirmed through visual evaluation.
2.With high spectral resolution and the need to accurately know peak values, the above fitting method is required. Fitting can be achieved through spline algorithm or Gaussian or Lorentz fitting, and one can choose based on the spectral properties of the tested sample. If you don't know how to choose, please use spline curves. |