Reproducibility comparison between two multi-gonio-spectrophotometers

Abstract

New instruments named multi-gonio-spectrophotometers have appeared to measure and characterize the goniochromatism of special materials like metallic, interference, luster and pearlescent samples. These devices are instruments what can measure the spectrum of the sample with different illumination and observation angles, these angles usually agree with some requirements marked in ASTM and DIN standards related to the gonio colour appearance characterization. On the other hand, the inter-comparison between spectrophotometers, both at repeatability and reproducibility levels, has usually appeared in many contributions in last years. The main purpose of this study is to compare at reproducibility level two multi-gonio-spectrophotometers according to the ASTM E2214-08 guidelines: one of them, with 5 measurement (directional) geometries, highly recommended by the automotive industry, while the other one is a bench-top instrument, with 10 measurement geometries. Therefore, they only have 5 common measurement geometries: 45ºx120º, 45ºx110º, 45ºx90º, 45ºx60º, and 45ºx25º. For the reproducibility comparison we will also make some statistical studies that include a Hotelling’s test and a statistical test of inter-comparison to know the confidence interval of the partial colour differences ΔL*, Δa*, Δb* and the total colour difference ΔEab. It is done using as database a collection of 91 metallic, interference, luster and pearlescent samples, which were measured 20 times without replacement for both instruments. The final findings show that except the 45ºx120º geometry, which is the nearest to specular direction, the reproducibility differences between both instruments are statistically significant. This means that these differences are due to systematic or bias errors (angle tolerances for each geometry, photometric scales, white standards, etc), but not exclusively to random errors. However, both statistical tests used here are not valid to discriminate and quantify the detected bias errors in this inter-instrument comparison.