Abstract
Near-infrared spectroscopy (NIRS) is an efficient and non-destructive method for quantitatively analyzing ingredients in a material. However, this method requires an expensive and time-consuming process to establish a calibration model. As a consequence, once the calibration model is established, it usually is used for many other spectrometers. Then, more or less a disparity in the predicted ingredient content of the material is observed. This phenomenon is called the instrumental difference and gives a difficulty in the application of NIRS. The reason is that waveform of near-infrared (NIR) spectrum is influenced by the characteristics of the spectrometer. We examined the process of measurement and found the spectral waveform was distorted by the variation of spectral response within wavelength resolution of the spectrometer. Therefore, the large spectral response variation and low wavelength resolution causes the big waveform distortion and then the big difference of predicted value of contents. Using seven linear array spectrometers, we analyzed this distortion and evaluated the difference of the predicted value. The results coincided with the analysis. By understanding the mechanisms of instrumental difference, we can design a spectrometer with minimal instrumental difference. This result could be applied to any other spectroscopy that acquires spectral waveform by dividing the sample waveform by the reference waveform.
