[werk_ag]

Hi, Zdenek

I tested voltage output and analyzed program. WeatherDuino calculates UV Index by this formula (WeatherDuino_RX_v1.4_b034.ino, table RX_TX, raw 212):

IndiceUV = ((ADC1 * 16.5) / 1024.0)

But this is not correct. It includes the impact of acrylic cover, but there is a question, if that impact is linear or not. And measurement in area of low UV indexes has to be fail, because the curve of UV Index isn't here linear:

First of all, are you testing your sensor in horizontal position? Maybe I'm wrong, but even in very bright sunny days, I never been able to read a UV index of 11, without any correction in the reading formula. I reach to this formula, by comparing readings (after assembling) with other nearby weather stations.

Said that, yes, the used formula includes some correction.
Not taking in account the nonlinearity of the readings between UV index 0 and UV Index 1 (less significant part of the readings) , theoretically the formula should be: IndiceUV = ((ADC1 * 12.0) / 1023.0)

Each one may need to slightly adjust this formula according to characteristics of the material (thickness, UV transmission, etc) used to cover the sensor (it must be covered by something with good UV transmission, to protect it from natural elements).
Each person can build the Solar Radiation / UV sensor in a different manner with different materials, because of this, some final software / hardware adjustments always be required.
Nearby weather stations and Wunderground are of great help at this stage.

Your changes in the code are an interesting approach, and don't seem wrong to me. If they work better for your, why not use them?
I only see one drawback: The results are always integer numbers. I know that UV index is defined as a scale of integer numbers, but personally I like have them with one decimal place (it's better for the graphs too).

Below is real time graph of my UV readings