The present invention relates to an instrument for measuring in relatively short periods of time concentrations of constituents in optically dense materials using the spectra near infrared radiation transmitted through thick samples of the material while the material is either stationary or flowing. The invention uses a broadband incandescent intensity stabilized light source combined with collimating optics to transmit a parallel beam of light through the material under test. The light transmitted through the material is then collected by a focusing lens and imaged onto a rectangular entrance slit of a special purpose spectrometer. This spectrometer has no moving parts and employs a fixed diffraction grating to physically spread the image of the entrance slit into a continuous range of wavelengths. A portion of the diffracted slit images covering the selected portion of the near infrared range is focused onto an array of individual rectangular photodiodes. By using relatively large area photodiodes and a relatively small number of photodiodes, high sensitivity is achieved and low intensity radiation levels can be measured quickly. By using a relatively narrow spectral range, medium resolution can be achieved. The outputs of each photodiode, or the outputs of a selected number of the photodiodes, are fed into current to voltage converters: either resistive (instantaneous) or preferred capacitive (integrating). Thus the outputs of all photodiodes are measured in parallel, which reduces the time to acquire the spectra. The gain of these current to voltage converters is programmable so that both high intensity and low intensity near infrared radiation levels can be measured without reducing the intensity of the radiation incident on the material under test and thereby eliminating the need for moving parts in the spectrometer. After the spectra are acquired, they are operated on by models developed to predict the percentages of various constituents in the material. These models are pre-calibrated using spectra obtained from materials of known concentrations and developed using chemometric, neural net, and/or genetic algorithms.