ing-ink specimen are found again to a close approximation (compare 0.630, 0.704, and 0.145 from table 7 with 0.631, 0.704, and 0.145, respectively, from table 5c.
.jpg)
Ten selected ordinates sometimes give significant information (see table 7); thirty selected ordinates are often sufficient (as above); and one hundred selected ordinates are sufficient for all but a few very irregular spectral distributions (such as produced by gaseous discharge tubes). These wavelengths are available for many sources in other publications [13, 21, 49]. Nickerson [114] and De Kerf [27] have published studies of the reliability of the selected-ordinate method of computation.
Analog and digital techniques have been developed for use with automatic computing devices to abbreviate the labor of computation required by spectrophotometric colorimetry. In the application of the analog technique [25] data are sent directly from the spectrophotometer to the computer without wavelength-scale and photometric-scale corrections. In one application of the digital technique [12, 95] spectral data are punched directly on cards and fed into the computer, while in another [75] the spectral data are corrected and then punched on cards and fed into the computer.
Tristimulus values, may be obtained by direct comparison of the unknown light with an optical mixture of three primary lights in a divided photometric field. Since the primaries of the CIE standard colorimetric coordinate system are imaginery, such a tristimulus colorimeter cannot be made to read directly. It must be calibrated by measurements of four known stimuli, and then may yield tristimulus values, ,,, by a transformation the reverse of that indicated in eq. (3). Since the color matches set up in a tristimulus colorimeter designed to cover any substantial part of all possible colors with a single set of primaries exhibit serious metamerism, the field has to be relatively small, subtrending about 2 deg at the observer's eye. This restriction to a small
14
