Color by numbers

Colorimeters, which measure color, are currently used in a number of product-development and control environments, as well as for product acceptance testing and a variety of other applications.

Colorimetry uses a set of three spectral response curves, standardized by the Commission Internationale de l'Eclairage (CIE), that approximate the human visual response. These functions are called the CIE colorimetric standard observers (see figure). Instruments designed to measure color, known as tristimulus colorimeters, consist of three or four channels with filters designed to match these functions.

To properly use an instrument to measure color, you should know the instrument's measurement uncertainty. Only then can you tell whether your colorimeter can do the job you need it to do. If colorimetric measurements are inaccurate, a user may accept an inadequate product or reject a perfectly good one.

The accuracies of colorimeters can vary significantly, depending on the spectra of the colors measured. In the case of tristimulus colorimeters, the mismatch of the relative spectral responsivities of the individual channels to the CIE color-matching functions can be a source of error when the spectral distribution of the source under test differs from the calibration source.

Colorimeters are typically calibrated by their manufacturers. If you look at an instrument's data sheet, you will see uncertainty values on the order of ±0.002 in chromaticity (x, y) and 2% in luminance (Y). Be careful, however--these uncertainties are often for measurements of an incandescent source with a color temperature of approximately 2856 K. These sources approximate the spectral power distribution of CIE Standard Illuminant A, a common standard source used in photometry and colorimetry. If we measure an artifact with the same spectral distribution as the calibration source, measurement errors tend to be canceled out, and the uncertainty will be small. Consequently, if we were to limit measurements with the calibrated colorimeters to sources having spectral power distributions similar to that of Illuminant A, there would be little need for improved calibrations.

If, however, we want to measure the color of artifacts having very different spectral distributions from Illuminant A--for example, displays, light-emitting diodes (LEDs), discharge lamps, etc.--then the error analyses become complex, requiring knowledge of both the source spectral distribution and the spectral response of each of the colorimeter filter channels. In most cases, this information is not available to the user, and the uncertainties in color measurements are often not well known. So even though we can calibrate an instrument very accurately against an incandescent source, we don't know how well the instrument can subsequently measure the chromaticity of a particular artifact.

For example, if we take a number of colorimeters and measure the chromaticity of various colors of a display, we find inter-instrument variations in the measurements as large as 0.01 in chromaticity and 10% in luminance, depending on the color and type of display being measured. These variations are an order of magnitude larger than the manufacturer's uncertainty specification, which is based on the measurement of a CIE Illuminant A type source. Similar problems occur in colorimetric and photometric measurements of LEDs. In this case, laboratory measurements can disagree by 20% or more, depending on the spectral distribution of the LED measured.

Uncertainties in color measurements on this order are too large for many current commercial, industrial, and military applications. For these applications, additional care must be taken to understand and reduce the measurement uncertainty. The National Institute of Standards and Technology (NIST) has established calibration services for colorimeters and spectroradiometers to be used in display measurements, as well as for photometric and colorimetric measurements of standard LEDs. These artifacts should assist laboratories in analyzing and reducing colorimetric uncertainties. oe

For more information on NIST colorimetric calibration services, see http://physics.nist.gov/photometry_calibration.