LEDs as Spectrally-selective Sensors

Mims (1992) demonstrated that LEDs could be used as spectrally-selective detectors. In a functioning sunlight photometer, Mims demonstrated how LEDs were sensitive to wavelengths shorter than what they emitted. Thus, a blue LED emitting at 470 nm is sensitive to bands shorter than that wavelength (460, 400, 350, etc).

Based on this information, we would expect red LEDs to be the most sensitive visible-light LEDs - all other visible light is a shorter wavelength than red. If a red 5mm LED emits light at 630 nm (a common wavelength for commercially-available red LEDs), then it is sensitive to almost all other wavelengths in the visible spectrum.

Based on what we learned about white LEDs in the "Introduction to LEDs" section, we would imagine that white LEDs would therefore make poor detectors. If the white source is merely a shorter-wavelength LED (blue, violet, or even UV) that is exciting a phosphor compound to create the rest of the spectrum needed for "white" light, then we would imagine that this LED would not be sensitive to high wavelengths. This would not have shown up in previous tests because I have been doing testing with "white" light only - light that is produced by "white" LEDs in the lightbox, which have an effect on other "white" LEDs. A red source, for example, should have very little effect on the voltage readout of the white LEDs, because it emits a wavelength that is much higher than what the blue/violet diode in the sensing LED can detect.

Today I've tried to support this notion with some spectral testing in the lightbox. The 'box comes with a variety of internally-mounted LEDs - red, green, blue (1), blue (2), and UV. I decided to take a look at the red, green, blue (1), and blue (2) spectra today. While I don't know the exact wavelength of the light emitted by these LEDs, I can assure that they are emitting different wavelengths of light because they are visibly quite different. By testing the response of the LEDs to different incoming spectra, I can test the theories above. The graphs of the data are below, separated by LED.

RS1 spectral response:

RS2 spectral response:

RS3 spectral response:

RS4 spectral response:

Complete graphs (with data) may be downloaded below, in .grf format.


What's strange to me here is that I'm getting results from these sensors in very high wavelengths. I expected to see a little variation for lower wavelengths, such as B2 and B1 - that seems to fit with what others have said previously regarding the spectrally selective sensitivity of the LEDs. However, there is very little change for some of the lower readouts, yet very significant changes with the red-spectrum readouts. I'm thinking that the phosphor coating on these LEDs may be making a difference, but I can't quite tell at this point in time - it's difficult to say what exactly is going on here. It looks like it's time to do some deeper research on the phosphor coatings of white LEDs and how they react to light.

New Facts

Today while doing some additional research, I found an article entitled "Spectral and Emission Characteristics of LED and its application to LED-based sun-photometry". While the article itself was not entirely about my line of work, it contained some highly useful graphs and comments regarding LED spectral emissivity and response patterns. What is proved in this article supports limited previous claims that LEDs are sensitive to wavelengths that are shorter than the ones they emit. That, however, is only half of the story. Acharya provides three graphs of particular interest to this project scope - one for a different wavelength of LED. He compares the spectral responses and outputs of red, green, and blue LEDs. All three LEDs show the sort of "peaky" emission spectra that you would normally expect from a wavelength-specific LED. Each spans a total wavelength of no more than 100 nm, with the smallest distribution of emitted wavelength going to the green LED (only 50 nm). What is particularly interesting here is that the LEDs are actually most sensitive to light of a slightly shorter wavelength than what they emit. The good news here is that the emission/response curves overlap for all three LEDs, proving that LEDs of a given wavelength can all be used in the same system without major lighting issues. This may be useful if my next batch of LEDs are poor sensors - I can change to a different wavelength of sensing LEDs. Just as long as I make sure that all of the LEDs are the same.

The .PDF version of the abovementioned article can be found below: