Some of these spectrographs have noise approaching 10% of the maximum intensity value, which is a little concerning.
The 10-reading was ambient sunlight from the windows and it was very much indirect--- and resulted in a nominal value of 10 all across the board (no spikes). All artificial lights were turned off. The 10 was consistent across the board, which we verified about 30 times (glancing at it in between LEDs) and never once saw a spike.
The 10 ambient noise is quite consistent in every graph. How is this concerning? I admit a 0 would be perfect but I don't see how it would be necessary since it was consistent across the board.
An interesting bit of experimental design info would involve how the light was collimated from the LED module into the spectrometer. This is a fiber-optic input (600um if I recall correctly) and would need to be lined up perfectly (machined fixture perfectly) if an external collimator was not used. Normalizing this data would make it a bit more useful, as it would become instrument-independent.
Once my finals are over for the semester in a few weeks I can upload a pic of the wooden jig I built to hold these steady. While I can't claim any sort of extreme accuracy, the wooden jigs held the LED's quite tight in their arrangement.
The wooden jig was placed under the probe to fit on all 4 sides along a marker-line that was drawn, so there would have been no more than 1/8" difference in the placement of the LEDs from one test to the next. At a 12" distance I see no reason to suspect this would corrupt the data enough to worry about, for our purposes.
Different currents/distances and PAR measurements would yield some interesting results, however this assumes the LED performs in a linear manner, which many LEDs do not.
Agreed. They definitely drop in efficiency as current increases above their "sweet spot", which seems to be around 350mA for most of the 3-w types.
But taking a couple reference points would help make it predictable enough, I suspect. I did a spectral graph of the XML at 3 different currents and was glad to see how consistent the spectral output stayed.