The expanded uncertainty of the NIST primary color-temperature lamp-standards is 8 K (k=2). However, both the long-term stability and the spectral distribution of the lamps can significantly change with burning time. By performing the tristimulus colorimeter calibrations against detector standards, the uncertainty of the color measurement uncertainty and the long-term changes in the measurement uncertainty can be significantly decreased.
The first detector-based tristimulus colorimeters were designed with an irradiance-measuring trap detector, where a temperature controlled filter-wheel was inserted and moved between the input aperture and the silicon tunnel-trap detector. However, because the windowless photodiodes were exposed to the ambient air, the illuminance responsivity decreased by 0.2 % from 2003 to 2007 and by another 0.1 % to 0.15 % to 2009. Also, the baffling inside of the trap-photometer was not efficient enough, allowing a small portion of incident light to bypass the filter as a result of internal reflections in the photometer. To improve these problems, a second generation tristimulus colorimeter was developed.
In the new-generation colorimeter design, the detector was closed with a wedge window. The 0.5° wedge was needed to avoid interference fringes in the output signal of the photodiode when calibrated at the tunable-laser based SIRCUS facility. The spaces on the two sides of the temperature controlled filter-wheel are small resulting in five orders of magnitude blocking in the ultraviolet and infrared wavelength ranges even when the input aperature is overfilled by the incident light. Also, the thickness of the filter packages is at least 4.5 mm to minimize fringes.
A picture of the second generation colorimeter is shown in the upper right column, where the front cover, including a 5-mm diameter aperture, was removed for better illustration. A preamplifier with eight decade gain selection is attached to the side of the colorimeter. The filter combinations were individually optimized to obtain the smallest possible spectral mismatch to the CIE standard color matching functions. The measured spectral responsivity functions of the colorimeter channels are shown on a logarithmic scale at the right. The relative expanded uncertainty of the spectral responsivity measurements was 0.1% (k=2). Measurements of this reference colorimeter show no responsivity degradation between 2007 and 2009.
The illuminance responsivity scale was transferred to the reference tristimulus colorimeter through calibrations at the SIRCUS facility and the Spectral Comparator Facility (SCF). Measurements at SIRCUS provided the responsivity scale between 360 nm and 960 nm, while measurements at the SCF extended the scale to 1300 nm. The combined uncertainty of the illuminance responsivity scale provided by these facilities is 0.2 % (k=2), a factor of two improvement over the existing scale published in 1996.
The scale for color temperatures can be determined either from spectral irradiance responsivity measurements or from spectral irradiance measurements at the Facility for Automated Spectral Calibrations (FASCAL). The expanded relative uncertainty of source-based, color-temperature calibrations is 2 K (k=2) at 2856 K. The color temperature uncertainty increases to 8 K (k=2) at 2856 K when propagated from the FASCAL-calibrated lamp standard to the reference colorimeter on the photometer bench. In contrast, the expanded uncertainty of the new detector-based color temperature scale is 5 K (k=2) between 2200 K and 3200 K.