Images: DANIELS & DANIELS
The first stage of the circuit yields 10 millivolts for each nanoampere generated by the photodiode. The second stage boosts the signal again, but it also magnifies the circuit-generated noise. Conveniently, the signal I am looking for is very low frequency, because the night sky is of nearly constant brightness. Therefore, the circuit can cut noise by using a low-pass filter--consisting of a resistor and a capacitor--without affecting the signal. The filter blocks frequencies above 10 hertz, which account for two thirds of the noise generated by the AD795JN in this circuit. (Technically, the bypass capacitors in the first stage also fulfill this function.) Overall the second stage boosts the output of the first stage 100-fold while keeping the noise output to only a few tenths of a millivolt.
In complete darkness, stage two of my prototype gave a signal of 3 millivolts with random fluctuations of about 0.3 millivolt. When I placed the device inside a dark and windowless bathroom and pointed my TV remote under the door, the output jumped 300 millivolts.
The third and final stage uses a chip called a comparator to check the output against a reference voltage that mimics a cloudless sky. The LM339, available from Radio Shack (part no. 276-1712) for about $1, has four comparators on a single chip, only one of which is needed here. The comparator turns the analog signal from the second stage into a two-stage output to indicate cloudy or clear.
For use as a cloud detector, encase the circuit in a grounded and weatherproof metal box. Cement aluminum foil inside a large plastic funnel and mount the photodiode near the bottom [see illustration]. This reflective horn guides skylight onto the sensor and blocks radiation from the ground. Mount the horn so that it points straight up. Under a starry sky, the second stage of my unit put out about 0.5 volt. When clouds rolled in, it increased to a little over 1 volt. It did not respond to moonlight.
To calibrate the instrument, point the horn straight up on a clear night and adjust the potentiometer R1 until the voltage at the comparator's negative input is 0.2 volt greater than the signal registered at its positive input. Then the output of the third stage should be approximately 0 volts. When you tip the reflective horn toward a light, the reading should jump to almost 5 volts. Test the detector on the next cloudy night. The circuit should generate about 0 volts when the horn is covered and about 5 volts when exposed to the cloudy sky. Readjust R1 if necessary. City dwellers and suburbanites should both be able to find a setting that reliably distinguishes between clear and cloudy skies. Given a suitable interface, this signal could be fed into a computer.
With minor changes, you can create other useful instruments. For example, if you read the output of the second stage directly with a digital voltmeter, you have an extremely sensitive near-infrared light meter. Because an object passing by will change the amount of infrared energy that reaches the sensor, the device can also be used as a motion detector. Replacing the infrared photodiode with one that is more sensitive to visible light makes a visible-light radiometer, which can do such things as measure the light pollution in the night sky and the energy output of bioluminescent organisms.
I gratefully acknowledge informative conversations with George Schmermund and Russell Wallace.