Hi Matthieu, You should still think of the calculation as an integral over wavelength. Doing a sum every 5 nm is a kind of "box-car" interpolation to the integral; you're accumulating a series of rectangles of width 5 nm and height equal to the product of the color matching function and your source's spectral irradiance/radiance. So you need to multiply the height (the product) by 5 (the "dlambda"). If you had samples every 10 nm instead and were still doing box-car interpolation, you'd have to multiply by 10. Of course if the end goal is just XYZ in relative units and not illuminance/luminance on an absolute scale, you don't really need to do the multiplication as long as you always do the sum the same way every time. Best, Keith On Mon, Jun 28, 2010 at 1:13 PM, Matthieu DUBAIL <matthieu.dubail@xxxxxxxxx> wrote: > Hi everybody, > I'm looking for an answer I didn't found on my books or internet. > I'm trying to compute the XYZ and luminous flux from spectral data of a > lightsource (taken from argyll and a colormunki). > Since I cannot compute true integrales, I'm using sum products: every > radiant intensity beeing multiplied with color matching functions of the > same wavelenght. Then, all products are summed. > My problem is that the sum product will clearly depends on the wavelenght > interval. I'll get higher XYZ if my wavelenght intervall is 5nm than with > 10nm. > Then how could I get normalized luminance from that? > Sorry if this seamed to be a stupid question, > Thanks by advance, > Matthieu DUBAIL > www.presse-bouton.com >