Assessment of the Sources for Disagreement between Two Spectrophotometers

Agreement between different models of spectrophotometers is increasingly being recognized as an issue in the industry. Brand owners are requesting color tolerances that are in some cases too tight to be met when using different make and model spectrophotometers along the supply chain. This often means that printers are being required to purchase a different model of spectrophotometer for each print buyer that they supply. It is a particular problem when printers seek to improve accuracy (and efficiency) by incorporating spectrophotometers into the press. Such inline spectrophotometers are necessarily of a different design than hand-held devices. This paper is part of ongoing work. The ultimate goal of this work is to develop a way to reliably standardize one instrument to another. The assumption is that understanding the physical nature of the differences between instruments will lead the way to a standardization process that provides the best performance without being prone to large errors. In this paper, I look first at how well five spectrophotometers of different make and model agree on what “black” is, and then on how well they agree on what “white” is. Several differences have been determined, including rejection of specular light, calibration of absolute white, aperture size, and goniophotometry. With the exception of calibration of white light, the traditional models for standardizing one spectrophotometer to another do not account for these differences. Thus, it is necessary to proceed with caution when attempting to standardize one instrument to another. Standardization may actually worsen the agreement between two instruments. At the very least, it is necessary to make sure that the physical properties of the standardization set (the set of samples used to correct one instrument’s readings to match another) are similar to the physical properties of the samples to be measured. In particular, translucency and gloss are important. Previous Work toward Improving Agreement There is no dearth of papers that describe techniques whereby measurements of a set of samples can be used to characterize and correct for the disagreement between two instruments. I will use the word “standardize” to refer to this process. See, for example, Robertson (1986), Berns et al. (1997), Rich (2004), Van Aken et al. (2000, and 2006), Chung et al. (2004), and Nussbaum et al. (2011). There is also no dearth of commercially available software for performing this standardization. Software is available from X-Rite (two versions), DataColor, CyberChrome, Color Science Consultancy, HunterLab, and ColorMetrix. See Bibliography section for links to these companies. On the other hand, there were three studies that gave cautionary advice. Butts et al. (2006) tested two commercially available programs and came to the following conclusion: Both profiling programs [Maestro and NetProfiler] were able to improve the inter-instrument agreement on BCRA tiles, but with consistency only for their own instruments... Improvements in BCRA agreement did not produce similar improvements in textile agreement... A significant number of textile samples were adjusted in the wrong direction and are in much worse agreement after profiling. These results were not unexpected. Rich (2004) had this to say: “If the model is built with only glossy tiles, such as the BCRA ceramic tiles, then matte materials, like textiles, are poorly modeled.” Rich recommended a collection of glossy semi-gloss, and matte, chromatic and achromatic (near neutral) samples for the characterization. Seymour (2013) came to similar conclusions when analyzing BCRA tiles to determine the characterization: [S]tandardization with a seemingly reasonable set of samples and a seemingly reasonable underlying mathematical model can be a worthless endeavor, and can often significantly worsen intra-model agreement... it was found that the BCRA series II set of tiles is lacking in that it does not provide for reliable differentiation between errors in nonlinearity and wavelength shift... This suggests that the addition of a few well chosen tiles to the BCRA set may improve their ability to standardize one instrument to another.