The emissivity of a body is equal to its absorbance at the same temperature.

G.R. Kirchhoff, Kirchhoff's Law of Radiation.

According to "Meteorology: The Atmosphere and the Science of Weather, Fourth Edition", J Moran & M Morgan (isbn:0023833416), Meteorologists use a scale from 1-10 to describe the sky conditions. 1 being completely clear and 10 being completely overcast. Using a simple integrated analog to digital convertor, it should be possible to directly interface the sensor to a computer serial port.

## So, you want to build a radiometer

### Goals

1. Not too crude
2. Output variable voltage
3. Ridiculously simple
4. Detect infrared, visible, and/or ultraviolet wavelengths

## We have two options

2. An array of \$0.39 photo-emissive diodes (LED)

## Now that we have chosen LEDs

At minimum, an array of three super bright LEDs per wavelength. I choose three sensors because we want to measure the radio flux over a wide surface area. The typical surface area of a large LED is 211.6 sq millimeters (460 x 460 micrometers). Four sensors would be a waste and two sensors would probably not provide an accurate area measurement. I don't think that the spacing would be overly important as long as each LED is evenly spaced at more than 1 centimeter from each other. Each LED would represent a vertex on an equilateral triangle.

The LED should ideally have a perfectly flat lens instead of concave or convex. The LED array should be flush mounted on a cold, with respect to the desired wavelengths, surface. For visible wavelengths we can use either a white LED (RGB composite) or just a yellow (~550 nm) LED. Yellow-Green is the color most sensitive to humans as that is the color of the Sun, so that would be the obvious choice for a daylight 'luminosity' meter. Infrared is simple, an Infrared LED designed for a TV remote. Ultraviolet should be simple. UV LEDs are often used as the exciter in UV LASERs (i.e. photocopiers), so they should be abundant.

With a single 3-LED array, we should have a radio flux of 10 mW per square meter. That isn't much. (A July noon produces about 50 mW per sq cm.) But, it is enough to provide a reasonably accurate measurement of solar radiation. It would not be accurate enough to measure much less than diffused solar radiation as sky conditions are typically recorded in 1/10 increments.

## Proof of Concept

For a simple test, grab a volt meter and a visible light emitting diode (I chose a Red 650nm LED for the test because I had one with a big lens ). Connect the LED to the volt meter and expose the LED to various light sources. I get a variance of 0 mV to 1150 mV (dark to light) per LED.

### Test Measurements

Reading light, measured from the page of a book
50 mV
Completely overcast sky (09:30 Local) through a triple-pane window
120 mV
50 W desk lamp
350 mV
100 W desk lamp
650 mV
150 W desk lamp
850 mV
Completely clear sky, Direct (15:00 Local)
1150 mV
Completely clear sky, Diffuse (15:20 Local)
450 mV
Dark hallway
5 mV