Images: DANIELS & DANIELS AMPLIFIER CIRCUIT converts current from photodiode into a high-or-low voltage output. It runs off two 9-volt batteries.

When not flying as an airline pilot, Monty Robson works to involve young people in science. Under his leadership, the Western Connecticut Chapter of the Society for Amateur Scientists has raised $50,000 toward a professional-quality telescope for a high school in New Milford. The telescope will be fully automated and connected to the Internet so that students far and wide can access everything the heavens have to offer.

Because any telescope can be ruined by rain and excessive wind, it is important to keep constant tabs on the weather. I'm designing the weather station for the automated observatory and thought I'd share with you my ideas for a cloud detector.

We humans can see clouds at night because they reflect light. But it is tough to build a weather sensor that relies on variations in the brightness of visible light, because the moon's brightness also fluctuates. A cloudless sky with a full moon can be brighter than a cloudy sky with no moon. Fortunately, the moon and the clouds differ in one crucial respect: the moon delivers little infrared light to the earth's surface, whereas clouds scatter infrared strongly. At night, as the air temperature drops, the ground vents some of its excess heat as infrared radiation with a wavelength of around 9 microns. If the skies are clear, most of this energy escapes into outer space. But a fluffy layer of condensed water vapor returns thermal energy back to the ground. That is why, all else being equal, a clear night is chillier than a cloudy one.

The reflected radiation can be observed with an infrared radiometer, which produces a voltage that increases with the intensity of the infrared light striking it. But there is a problem. Photons emitted by the earth's surface have such a low energy that inexpensive silicon sensors, such as photodiodes, must be chilled with liquid nitrogen, lest the signal be swamped by the thermal jostling of atoms within the detector itself.

One simple way to avoid the cryogenics might be to take advantage of the telescope's urban setting. Incandescent lights, including house lights and streetlights, are much hotter than the ground and so produce more energetic radiation. If enough photons from these sources scatter off the clouds, they should overcome the noise within a photodiode operating at room temperature. So I decided to try an unchilled radiometer. The early data look promising, especially for low, thick clouds, but I hope you will conduct your own experiments where you live and send me your results.

The infrared photodiodes at your local electronics suppliers are most sensitive to wavelengths around 0.9 micron, or 900 nanometers. But they also pick up some visible light--wavelengths between about 400 and 700 nanometers--so you must screen this light out. Some photodiodes, such as the NTE3033 that I purchased at Fry's Electronics for $4, are encased in an opaque plastic that blocks visible light but not infrared. Others, such as the SD3421 from Honeywell Micro Switch (call 800-367-6786 or 815-235-6838), will require an external filter. Edmund Scientific sells a circular filter 1 inch in diameter for $5 (call 609-573-6250 and ask for part no. H43948)

As used in this radiometer, the photodiode transforms the photon intensity into a very weak electric current, which must then be converted to a voltage and amplified. For the circuit, I chose the AD795JN operational amplifier, manufactured by Analog Devices (800-262-5643 or 617-329-4700 to find the nearest distributor), in part because it produces scarcely a whisper of electronic noise. You might experiment with lower-grade op-amps such as the TL082, available at Radio Shack. But old standbys such as the 741 op-amp are far too noisy.