We bought several sets of a supposed canned blackbody lab from Pasco. My idea was to use them for PhySci 119 (stellar astrophysics). The gear toook ages to come. When we finally tried it it turns out the manual is a piece of junk. It uses refractive index data between 400 and 1000 nm but extrapolates this in the plotted curves out to 4500 nm. Needless to say the curves do not have the proper Planck shape. A guy called Jon Hanks (hanks@pasco.com) at Pasco came up with some some more points on RI curve out to 2300 nm (from Schott catalog). Using these I have suceeded in getting reasonable looking Planck curves over a limited wavelength range.
Following two plots show: Left - the RI versus wavelength data and the spine interpolation I have used here. Right - translation from sensor angle to RI (comes from geometry - see manual for formula).
The following plot shows the combination of the above two to convert sensor angle into wavelength of light.
Before we can plot blackbody curves here there is another step which is missed in the Pasco manual. Since the sensor slit width is a constant angle we need to correct the readout voltage such that it is intensity per unit wavelength, rather than per unit angle. So we need to calculate the slit width in units of wavelength. The variation is considerable as seen in the plot below. The turn over is probably not real - unfortunatley the RI versus wavelength data is too sparse to determine the derivative accurately in the 1000 to 2300 nm region and the spline interpolation introduces artifacts. I played around but could not do better. A good analytic fitting function might help. Pat Palmer mentioned such a thing may exist. The fit provided by Jon Hanks is much poorer.
Dividing the sensor output voltage by the slit width we can now plot spectral intensity curves. Shown in blue are the data curves for lamp voltages from 1V to 10V. I have fit the Planck function between 600 and 2300 nm - shown as red lines, with extrpolation outside of fit range shown dashed. Note that the sensor does not respond in below 600 nm - this portion of the the spectrum certainly makes it through the optics as you can see it by eye! Also there is evidence that the optics and/or sensor are cutting out below above 1700 nm as the data no longer fit well. The prism is apparently good up to 2300 nm, and the lense appears to be 80% at 2000 nm (according to catalog pages faxed to me by Jon).
In the above fits I allowed the temperature to be a free parameter for each curve, and a single normalization parameter for all curves (representing the unknown efficiency of the optical system). The plot below shows the resulting temperature versus lamp voltage plot which looks pretty sensible.