My home has a crappy roof. It’s badly insulated and overheats easily in the summer. The goal is to insulate it, but due to a multitude of concurrent projects we won’t be able to do it right now.
I had the idea of spraying my tiles with a mixture of lime diluted in water, thus creating a really cheap, ecological, and rather reflective (or rather high albedo) coating on the roof tiles. However, how could I ensure that this works? Maybe I would have the placebo perception that it does, whereas it does nothing in reality.
To measure the effect of the white tiles, I should add temperature sensors inside the home, but we’ll also need to compare them to the sun irradiance.
Thankfully, on the first company I created, we had a pyranometer that we used for energy efficiency benchmarks. A pyranometer is a device that measures the solar irradiance. There are different types (thermopile, photovoltaic, ..), but they all basically map the solar irradiance ($W/m^2$) to a voltage output ($V$).
I have the chance of having a thermopile pyranometer. More specifically, a Kipp&Zonen CM10. It was calibrated by the University of Geneva so I know it measures what it should measure. The output voltage of such pyranometers is rather small.
The calibration constant of mine is $4.94 ±0.02 mV/kW/m^2$. That means that when the sun is shining full brightness, it will output around $\approx 5mV$. This is small! As I’d like to have a resolution of around $10 W/m^2$ I will need to be able to measure a voltage resolution of $50 \mu V$.
Raspberry pi to the rescure
I love Raspberry Pis. They are were rather cheap, and their GPIO allows them quick interfacing with other ICs.
I chose the MCP3421, a 18-bit ADC that talks I2C. It as a programmable gain amplifier of up to x8, allowing to boost the input signal from the max $5 mV$ to $40 mV$.
The voltage reference is $2.048V$, so 18 bits of resolution gives us steps of $\frac{2.048V}{2^{18}} = 7.8125 \mu V$. That means that, noise aside, the internal resolution is $\approx 7.8 \mu V$, which translates to an input resolution of $\frac{7.8 \mu V}{8} \approx 1 \mu V$. Given the voltage output of the pyranometer, $ 1 \mu V$ equals around $0.2\frac{W}{m^{2}}$.
The MCP3421 was then soldered to a (fugly) board made at home:
Some python code later (visible here) and the Pi started sending data to the OpenTSDB instance:
Interestingly, we can use the pyranometer to check that the production of the solar panels is correct by graphing the irradiance and the power production of the solar panels:
