Supplementary material for the article: m-EDI Measurement Using Low-Cost Spectrometric Sensors Based on Photodiode Arrays with Narrowband Color Filters: An Exploration of Alternative Calibration Methods

This directory contains 10 spreadsheets with the calibration data of 10 AS7343 13-channel color sensors from AMS-Osram. Each sheet uses two calibration methods, referred to as: Multi-variable Linear Regression (MLR) and Selective Channel Elimination (SCE). Both methods are applied to obtain the coefficients that would allow the measurement of the mEDI parameter through the weighted sum of the readings from each sensor’s channels. The results obtained have been used as supplementary material for the experimental validation of the methodology proposed in the article titled: “m-EDI Measurement Using Low-Cost Spectrometric Sensors Based on Photodiode Arrays with Narrowband Color Filters: An Exploration of Alternative Calibration Methods.” The input data used in each spreadsheet include the spectral distribution of 21 light sources along with the readings obtained by the sensors to be calibrated for each of these sources. Of the 21 illuminants used, 14 were used for sensor calibration and 7 for verifying the resulting fit. The test bench used to obtain the input data allowed simultaneous application of the same light level and spectrum to both the sensor being calibrated and an Ocean Optics HR2000 spectrometer used as a reference device. All spreadsheets share the same structure and functionality. Below is a description of the contents and purpose of each tab in the spreadsheets: Sensor_readings At the top of this tab, direct readings from each sensor channel (COUNTS) for the 21 illuminants are included. At the bottom, the spectral distribution of each light source is provided, measured in mW/nm/m², ranging from 380 to 900 nm in 1 nm increments. Based on the direct sensor readings (COUNTS), normalized readings (BASIC COUNTS) are calculated according to the sensor manufacturer's instructions to minimize the influence of integration time and gain settings. To balance the error weight for each illuminant when applying the MLR or SCE algorithms, and thus reduce the impact of level differences between illuminants, a scaling factor is applied to both the BASIC COUNTS and the spectral power distribution. This scaling factor ensures that the readings for all illuminants have the same root sum of squares value across channels. SPD_illuminants This tab displays the illuminant spectra after applying the scaling factor mentioned above. Linear_regression At the top of this tab, the data to which the MLR calibration method is applied are shown. The coefficients obtained using this method are displayed at the bottom, along with the estimation error of the m-EDI for both the calibration and test illuminants. SSR_quasi_normalization This tab shows the procedure for calculating the coefficients to be applied to each channel’s spectral response (QN-coef) to ensure that the maximum value of the response is similar across all channels. This procedure is referred to as Sensor-spectral-response quasi-normalization. Selective_channel_elimination This tab contains the calculations related to the application of the SCE method. On the left side, the results of each iteration are shown as new channels are added to the coefficient calculation. Below each iteration’s results, a cell indicates “Accept” if the newly added channel did not result in anomalous coefficient values, and “Discard” otherwise. On the right side of the tab, different combinations of channel addition order are included, along with a table comparing the coefficients obtained using the MLR and SCE methods, and a comparison of results for the 21 illuminants used in the tests.