Re-evaluation of daylight spectral distributions

The spectral power distribution of colorimetric daylight illuminants was determined based on measurements [1] done some 40 years ago. The data were averaged and standardised at every full 10 nm. Later, to follow colorimetric practice, these data have been interpolated (linearly) to every 5 nm, but the factors to calculate the values have not been changed. We show in this paper that this is not correct, and leads to some discrepancies, especially if data with even smaller step size are required. We suggest that when correcting this discrepancy one should at the same time go from the linear interpolation technique, a rather crude method for a spectral distribution with many minor minima and maxima in the spectrum, to a non-linear one. We show that a third order spline interpolation can provide smooth functions resembling the original data to a high degree. 1. First statement There are inconsistencies in the calculation method of Daylight spectral distributions. These should be corrected. State of the art In the 60s the spectral distribution of daylight was standardized between 300 nm and 830 nm at every 10 nm, based on detailed measurements performed in many places of the globe[1]. From the set of measurements the mean distribution (mostly referred to as S0(λ)), and two characteristic distributions (S1(λ) and S2(λ)) were calculated. S(λ) = S0(λ) + M1S1(λ) + M2S2(λ) (1) Factors M1 and M2 are functions of the colour temperature: D D D D y x y x M ⋅ − ⋅ + ⋅ + ⋅ − − = 7341 .