Oh, that's how it works...

I was reading about Forrest Mims employing LEDs as light detectors for building a simple sun photometer. I wondered how straight forward such an implementation could be. I'm pretty familiar with the idea that most light emitters also generate some kind of response when they are illuminated (unfortunately, flashlights still don't run backwards in a meaningful way, Calvin). Picking off this response out of the noise is a little challenging, but commonly done. For instance, laser diodes have been used in matched emitter/detector pairs. Intuitively, one would guess that a plain LED would respond similarly to a photodiode, but with different, and likely lower, spectral response.

As far as using plain LEDs as light detectors:
It turns out the process is very simple. They act just like their photodiode brethren. Both current and voltage scale with light input. I took a blue LED, attached it to the µA input of my multi-meter, and pointed it at the desk lamp.

Ding! There's signal here.

Switching to mV, it pegged the meter. What a good sign! Going between the shade of my hand and the desklamp scaled across ~1.5V.

Ding! Ding!

That was easy.

On the principle that different LED colors respond to different wavelengths, I went as weird as I could imagine to see what would happen.

I grabbed one of the mystery black LEDs and attached it the the µA input, and pointed at my hand, the computer screen, and the desk lamp. There was some variation, but not as much. Similiar results were obtained with the voltage input.

At times, the voltage response seemed inverted... it looked liked the voltage was dropping as I moved between light sources. This was very confusing for a moment, until I realized I was resting it near my cup of coffee when I went to fiddle with the meter.

That totally broke the enigma.

That little black LED didn't care much for light I could see, but it sure did love a hot cup of coffee.

Ding! Ding! Ding!