Page:Tristimulus specification of the Munsell book of color from spectrophotometric measurements (IA jresv31n1p55).pdf/26

 between 480 and 600 mμ is not considered significant. It is probably caused by nonuniformity of the sample. Differences of this magnitude can be obtained when a sample is rerun with the same backing if the sample and backing have been removed and reinserted between runs.)

The effects of such spectrophotometric differences on the computed values of Y, x, and y are shown in table 3. Since these samples probably illustrate the maximum effects to be expected from the two backings, it is apparent that the differences in color caused by measurement with white or with black backing are mostly unimportant.

The use of calibration curves on each record sheet-those enabling corrections of wavelength errors, 100-percent and zero curve deviations, and aging of the MgO comparison surface, as used at the National Bureau of Standards—enables spectrophotometric data to be obtained with much less care and worry regarding certain details of operation than if these calibration curves were omitted. Omission of the curves makes it necessary for the operator to take great care, for example, in the insertion of the graph sheet in the instrument, in continually checking the wavelength calibration of the instrument, and in controlling or watching the graph paper for expansion or shrinkage with change of humidity. A new MgO comparison surface must be prepared each day, and the question of reproducibility of such surfaces thus enters. The possibility of erratic differences in results between the two investigations is thus present, but since different actual samples were measmoed, no further conclusions can be reached regarding the erratic differences between the Glenn-Killian and the National Bureau of Standards data.

TABLE 3.—Effect of backing on colors of 1I1unsell samples Munsell sample

[Values are computed from the spectrophotometric curves shown in figure 10.]

With respect to the differences between values of X, Y, Z, x, y, and z resulting from differences in computational procedure-3 0 selected ordinates, as against weighted ordinates at every 10mμ it has been shown [21] that such differences are small for samples such as those considered here, much less than some of the differences shown. Only small and unimportant errors are therefore to be expected from this difference in computational procedure.

Detailed comparison of the values of x and y obtained by Glenn and Killian with those obtained at the National Bureau of Standards may be made by the inspection of figures 2 to 8 or by study of the published data. Only two additional points will be noted here. 1. Certain consistent differences in the respective chromaticities are apparent when the (x, y)-data for certain groups of samples having the