I think you conclusion is right. It also shows that the Sky unit is measuring something unaffected by your filters and uses that to calculate the UV value as it clearly doesn’t measure important part of the spectrum that would effect the UV value. But if one knows the physics behind it, one could perhaps deduce the UV value from measuring a different part of the curve or even a single point. (a point that isn’t influenced by your filter)
UV filters block a small percentage (between 0.1 and 5%) of the light that passes through them. Because of how the light interacts with your filter , this reduces the sharpness and contrast of your images very slightly.
Interesting experiments, @iladyman. For reference, the UV sensor in the SKY is the “off-the-shelf” APDS-9200 sensor from Broadcom and it’s response peaks around wavelengths of 320nm. Here’s the spectral response of the sensor, pulled from the sensor’s spec sheet:
3 Likes
I have a feeling that the typical UV filters are cutting of but not high enough compared to the UV sensor in the SKY
https://www.lenstip.com/120.5-article-UV_filters_test_-_supplement_Hoya_HD_UV_67_mm.html
Here you see a high quality Hoya and it is near flat till 340 nm and from there steep up.
This is with a quality filter. Down the page you have links to other pages with other brands and some really cut off way less so the UV sensor picks up UV as it would not be filtered at all.
4 Likes
that kind of curve combined with a typical uv photo filter curve should have a significant influence on the measurements. Perhaps the coating on the sensor isn’t working and way more of the visible light makes it trough to the sensor.
Has anyone noted that the Broadcom APDS-9200 UV and Ambient Light Sensor was discontinued on June 8, 2018 with a last time buy of December 31, 2018?
Well, we definitely noticed that! And so we made a pretty large last-time buy back then in order to give us plenty of time to find a replacement. And the good news is: we found one well before running out of stock. Liteon makes a nearly identical part, the LTR-390UV (they likely bought the design from Broadcom) and all indications are it will be supported for the foreseeable future.
5 Likes
it seems to be rather imprecise with regards of UV index (can be off by 1 full index number)
Yes, but that’s no different than APDS-9200. A nominal accuracy of 5% FS is typical for UV sensors, which also drift over time. Our CL process improves upon that tremendously, pulling it down closer to 2% FS, and keeping any drift over time in check.
4 Likes
I’m sure you already looked at this one, but I’m using a VELM6075 in my DIY weather station, on the off chance you haven’t!
Thanks. Yes, we have played with the Vishay UV sensor. And we like it, though it doesn’t seem to be any more accurate than the APDS-9200 or the LTR-390UV. Also, we would need an additional sensor for solar radiation since the 6075 is UV only. That’s possible, but more complicated and more expensive. Vishay is supposedly working on a combo sensor, but nothing is available yet. Maybe in the future!
3 Likes
Ah yes, I’m using the TSL2591 for light (IR, Visible and Fullspectrum) - I believe there are some calculations to get solar radiation from that but I haven’t looked into it yet (using it for Sky Quality Meter calculations at the moment).
2 Likes
but there is a graph with a few measured points vs a calibrated machine, and there it appears to be off by at least 10-15% for some points.
There was a paper published in the Journal of Measurements in Engineering, Vol 8, Issue 4 (Dec 2020) that provides a corroboration of Solar Radiation = ROUND(Illuminance/120, 0): “A conversion guide: solar irradiance and lux illuminance” https://doi.org/10.21595/jme.2020.21667
1 Like
yeah, for Solar irradiance and illuminance such a relation is useful. the UV value is however a different story.
I’ll be “that guy” to dig up an old thread…but I’m really glad I found this thread… I’ve been debugging a graphing program for 4 hours. My lux value simply will not display, and I couldn’t figure out why. That was until I happened to notice a very tiny piece of the Lux color just barely below the Solar Radiation line. The reason I couldn’t get the Lux value to display is because it’s directly beneath the Solar Radiation value.
Ah the life of a programmer… 
1 Like
I’ll come back to this topic because I think I have something new to add. While it’s been noted here that “Brightness” and “Solar Radiation” are the same thing in different units (factor of 120), “UV” and Solar Radiation have some variability in the relationship. I just started experimenting and, in my data, this is a function of seasonality. If you look at a scatter plot of UV vs. Solar Radiation something weird is going on. Another variable is needed and it turns out to be season. Let’s color the dots by month.
We can take the average UV/SolarRadiation ratio by month
month ratio
May 0.00775
Jun 0.00824
Jul 0.00909
Aug 0.00933
Sep 0.00844
Oct 0.00736
So UV is stronger in the summer for any given level of brightness.
We can develop a conversion formula with a regression using months as dummy variables.
term coefficient
SolarRadiation 0.00875
monthMay -0.230
monthJun -0.0687
monthJul 0.130
monthAug 0.158
monthSep 0.00400
monthOct -0.150
You could get rid of the dummies by using days away from either solstice as the season variable.
Why does this happen? I am in at about 40 deg. North. I an guessing that when the sun is lower in the sky it has to pass through more of the atmosphere an this selectively filters more UV.
Given how much the sun moves in a single day, you might think hours-from-noon might make a big difference. Surprisingly the effect is is only pronounced near sunrise/sunset. Sorry for the crappy table. As a noob I can only post one plot.
UV-
Solar
Ratio
Hours from Noon x10000 Solar
-7 68 7
-6 82 46
-5 85 130
-4 86 267
-3 86 413
-2 86 531
-1 86 617
0 87 645
1 87 626
2 87 572
3 87 469
4 86 330
5 86 213
6 85 103
7 84 34
8 67 6
NOTE: I don’t have 12 months of data because the IFFT link broke over the winter and I didn’t notice.
2 Likes
a really nice data analysis. Love it!
In the end it might be dependent on things like variations in the ozon layer, so what might look like some correlation this year, might not be the same correlation next year.
2 Likes