A rainproof Cloudsensor


Other Projects

!!! Important Notice !!!

The earliest ideas for this project go `way back' to the apparation of comet Hyakutake in 1996. For some time it rose in the (very) wee hours of the morning and it took considerable effort to rise out of bed; just to have dense clouds grinning back at me through the window ...

Some time later I stumbled upon the web site of the Bradford Robotic Telescope. I found the weather sensors especially interesting, as they included a cloud sensor that could be easily copied with homebrew equipment.


The above image shows the working principle of the cloud sensor:

  • (Left) If it's cloudy, the clouds will reradiate almost as much heat back down to the ground -- or in this case to the top of the Peltier-element -- as the ground radiates upwards. There's no chance for a big temperature difference to build up between the two sides of the Peltier-element, hence only a very weak current can flow, or to be more exact: the thermo-voltage is small (it all depends on what you're measuring).

  • (Right) If the sky's clear not much of the heat-radiation from the ground gets returned by the atmosphere. A (still) small (but sufficient) temperature difference can be sustained. A larger current emerges, the thermo-voltage being higher.

    For two years I had my cloud sensor in use, with a sensitive analog instrument as it's display. It was built according to the BRT description, i.e. the Peltier-element was sandwiched between two metal plates separated by a piece of foamboard, of which one was turned to the sky while the other one only `saw' the roof. During this time I gained more than enough experience with the behaviour of the sensor and decided to improve it.

    The Peltier-element I employ has a size of 40 mm squared. IMHO it shouldn't be of a significantly smaller size as this would also diminish the signal returned from the sensor. I considered it extremely annoying that the sensor wasn't rainproof, i.e. a few drops of rain would cool the upper plate and simulate clear sky conditions. A nuiscance of similar extent was the formation of dew on the sensors top, effectively making it blind to clouds. I also wanted to have a display I could read without having to turn on the lights.

    The rain problem could only be solved by a cover for the sensor. I knew there were materials that were transparent to heat-radiation, but I didn't have the slightest idea where and how to get them or if there might be every-day substances with the desired characteristics. After some time of experimenting I got a hint from Georg Dittie, telling me that plastics consisting only of Hydrogen and Carbon are often transparent in the far IR. Thus I was pointed at a very abundant source of material:

    Common food wrap (at least here in Germany its made of polyethylene)

    The immediately apparent problem of low stability was solved with a substructure of very thin brass rods over which I placed a layer of fly-screen. The food-wrap fastened to this has braved hailstones and gusty wind for over a year now.



    The relatively narrow pyramid is neccessary to keep raindrops from clinging to the food wrap, otherwise they would efficiently mask a break in the clouds when the shower is over. To keep spray from entering the pyramid from below (Peltier-elements aren't very fond of that!) the base also has a cover with a few venting holes sealed with fly-screen. The Peltier-element is suspended from the brass rods by thin nylon thread. The sensitivity of the sensor was of course reduced by the cover, but I've never had a `false alarm' since.

    At first I had substituted the analog instrument with a circuit that compared the sensor's output with two adjustable voltages. When the signal surpassed the lowest voltage a green LED lit up; when the second voltage was reached a red LED was turned off. Thus I had a signal similar to that of a traffic light:

  • `Red': It's cloudy.

  • `Yellow' (red and green): There are breaks in the cloudcover.

  • `Green': No clouds !

    Unfortunately this circuit spat a lot of noise into the AM ranges of my radio, so I stopped using it.

    Thanks to some lucky circumstances a broken mirror-gavanometer fell into my hands. After some arduous work with (almost) invisible wires I managed to revive it. I exchanged the indicator bulb with a red LED and placed two green LEDs behind the read-out window -- one near the `0', the other one near the `100' mark. In a very clear night (as good as they get here ) I selected a suitable range and adjusted a(n added) variable resistor until the index lay square on the `100' mark.
    Now I can tell if it's cloudy outside without having to leave the bed or putting on my glasses ...


    For a start you will probably be using a simple analog instrument. Here are some hints for this:
    (be sure to also read this)

  • If possible try out several different meters because the sensitivity rating says nothing(!!!) about the output you can expect under a clear sky. I had a 3 mA (full scale) meter deliver a much stronger signal than a 0.1 mA meter (again: full scale)!!

  • If the sun can shine on you sensor you'd better disconnect the meter. The sun can cause a signal which easily exceeds the ``clear sky'' output by a factor of a hundred and is of the opposite polarity. I had nothing bad happen yet, but you just could damage the movement of you meter.

    Important Notice:

    In the meantime I had reconnected my cloud sensor to the old analog display and several other meters as well. There I noticed that the output of my sensor with a very clear sky always lay in the range of 2.1 - 2.2 mV -- independant of the connected meter (at least with the ones I've got). Unfortunately this also means that the internal resistance of the meter plays a just as important role as its sensitivity.

    With sheer luck my first analog meter had a very low internal resistance (about 11 Ohm) and a bearable sensitivity (1 mA, full scale). Under ideal conditions I got a signal of 0.2 mA, or 20% of the full scale. Assuming you want to use a similar cloud sensor a derivative of the ever popular `URI'-formula can show if it makes sense to use a given meter.
    An example: 

                  U = 2.2 mV   (= very clear sky)
              R = 1 kOhm   (= internal resistance of the meter)

with:
              U / R = I

in combination with the above values we get:

              I = 2.2 (micro)A
    The full scale of the instrument in the example given should not significantly exceed 10 microamps!!
    (unless you can do with less than 20% of full scale as your largest signal)

    The display suited best for our `needs' seems to be the ``traffic light'' type mentioned above. If I manage to still find the circuit (in working order) in the bottomless pit of my electronics chest I'll post it here.




    Finally here's an (admittedly: bad) image of my cloud sensor in action. If this picture seems familiar to you it may be because you already saw it as a part of this one. The weird color cast is due to the aurora of 6./7. April 2000.


    Questions ?!?