Hi all. Couple of newbie questions potentially. I've had a google and come up blank on this, so putting it out there for discussion. 2 questions.
Background
I built a thingy with the intent that I could determine an approximation of the UV index at a given location and time. Most UV index warnings are for a broad area and assume clear sky. What if you're in the shade, or even indoors? It'd be interesting to know how much UV exposure gets through the windows (shouldn't be much). As I live in NZ UV exposure is stonking in the summer.
Being very new, I ended up buying and using two sensors, a GYML8511 and a SI1145.
The GYML8511 measures uv in the 280-390nm range, and I output the UV intensity (of those wavelengths) in mw/m3^2. The UV energy intensity is not directly related to the UV Index scale (being based on a number of factors such as the energy of specific UV wavelengths, cloudy cover, etc), and I couldn't find a formula that I could use to approximate the scale using it.
The SI1145 on the other hand does not directly measure UV, but instead visible and IR light. It then uses an algorithm to calculate an approximate UV index.
The problems
First, the SI1145 provides index values much much higher than the forecast index. EG, at 12:00 the forecast might be a UVI of 4, and I have a value of around12! The only times it's come close to the forecasted UVI is when it's early morning or late afternoon and the value should be 1 or less...
Second, I notice the orientation of the sensors to the sun is critical in getting the maximum result. Directly facing the sun obviously gets the highest result, but even slightly repositioning can drop values significantly. IE, just tilting by one or two degrees will drop the index value potentially by 1.
Solution to problem 1?
I looked into ways to calibrate the sensors to provide something more meaningful, but I was in over my head when I saw reference to address registers etc. Instead I've come up with a plan to record the UV values of my thingymajig, and note the forecasted UVI. Over time I can then plot the data against each other and generate an equation. I think the idea has merit as you can see in the picture below that there is a fairly good fit so far between the UV intensity measured by the ML8511, and the forecasted UVI on a sunny day:
I need a lot more data, and will need to keep updating the calibration equation over the summer as we're currently only hitting a UVI of around 5 or 6 in spring, I expect it to hit 13+ in the summer. I've tried to factor in data on cloudy days but this is far too chaotic, so plan on sticking to clear days. I am assuming if I calibrate in clear skies, values I get on cloudy days could still be valid as it will be based on the UV getting through the clouds.
Solution to problem 2?
The sensors on my thingy are mounted fully exposed on the surface of a jiffy box. I didn't want to put them behind something as I thought I'd lose some of the light I was trying to measure. However, I now wonder if I should use a dome diffuser of some sort to try and broaden the angle of measurement somewhat. The problem is I am struggling to find something available out there. PIR fresnel lens domes are UV stabilised and would be no good for the ML8511 I assume. I do however wonder if it might work well for the SI1145? I know there are materials that transmit UV light well (EG, Boroscillate glass) but I have yet to find a boroscillate fresnel dome?