Toward a more accurate and extensible colorimetry. Part I. Introduction. the visual colorimeter‐spectroradiometer. Experimental results

A bright ten‐degree split‐field colorimeter is used to study pairs of visually‐matching lights. This colorimeter is coupled by quartz light pipe to a spectroradiometer, and yields accurate absolute spectral power distributions of every viewed light. Six normal human observers use three disparate sets of spectral primaries. In Part I of the article thorough descriptions of the instrument, its calibration, and experimental results are presented. (1) By the Maxwell method, sets of 28 three‐band lights, each light matching a broadband reference‐white light, are obtained. From these sets are found strong and unexpected dependence of perceived brightness per watt on spectral content, and large systematic errors in computed chromaticity (1964 CIE Observer). (2) From sets of white two‐band lights, each matching the broadband reference light, large differences in wavelength of the visually complementary spectral lights, from those indicated by the 1964 CIE chromaticity diagram are found. (3) Sets of Maxwell‐method and maximum‐saturation‐method color matching functions are obtained for each primary‐set and for each observer. (4) Study, at the colorimeter, of perceived brightness of spectral lights in isolation yields three‐component perceived‐brightness‐per‐watt curves like those of Stiles and Crawford of 1933. (5) Perceived brightness of 3‐band white mixtures always increased upon removal of any one of the three spectral components, in agreement with MacAdam's findings of 1950. (6) Visual tests were made of Grassmann's additivity assumption, and of Maxwell spot phenomena. Part II includes discussion of chromaticity and brightness errors and of other aspects of colorimetry problems‐transformation of primaries, structure and shape of the chromaticity diagram, and normalization of color matching functions. Part III gives avenues of possible improvement and general comment.