impactoz
Full Member
Posts: 142
Raspberry Pi: Yes
|
Post by impactoz on Nov 2, 2020 23:20:56 GMT -8
My mlx90614 sensor arrived to play with.... (its an IR Temperature sensor - determines the temperature IR in the direction you point it - point it to the sky!!!) Straight into the skyweather station it goes.... Its a cheap sensor under $10.... A little soldering to add a grove connector... and away we go! Had a few problems getting it to work - Adafruit libraries only installed for Python 3, and wouldnt work with SkyWeather - tried unsuccessfully in getting something to work, found a hack by not using the libraries on an import and including it on the file with the code - so got it working... I got interested in trying this out from reading olegkutkov.me/2017/08/10/mlx90614-raspberry/The mlx90614 reads at least 2 temperatures - ambient and object.. Which is interesting.. I was only looking at the object temperature - what its pointing at... SO after a day of running - the theory looks like it might actually work ! The current temperatures reported by the mlx90614 are; Ambient: 29.81 Object 6.53 A difference of more than 20C and guess what - visually its cloudy !!! Early in the day it was clear - and the difference was probably only 4C So I am going to monitor it for a few days - and see what happens.... If I can confidently determine cloudy / overcast / clear days by the temperature differences... I will move this code permanently into my skyweather system so it can report this information as well. I will share my results and code if it all pans out.... Fun times! (running out of sensor ideas now....) |
|
impactoz
Full Member
Posts: 142
Raspberry Pi: Yes
|
Post by impactoz on Nov 3, 2020 11:43:41 GMT -8
This definitely looks like its going to work... In the last 24 hours I have noticed
A difference of 20C or higher - absolute clear skies, not a single cloud
A difference of 19C or less - some light cloud around
A difference of 8C or less - full cloud coverage of the sky.
It appears to work during daytime and night time - without any changes....
So very interesting - its now only a matter of multiple days to determine the cut off points. Currently I am suggesting that I might use
>20C difference - Clear
12C > 20C - Cloudy
< 12C - Overcast
All subjective though - but there appears to be some truth in this madness!
|
|
|
|
Post by SDL on Nov 4, 2020 8:25:49 GMT -8
Wow! This is pretty interesting. Do you have an idea of how much of the "sky" you are seeing with the sensor?
Is it under a quartz window? What material are you looking through?
BP
|
|
impactoz
Full Member
Posts: 142
Raspberry Pi: Yes
|
Post by impactoz on Nov 4, 2020 11:57:26 GMT -8
Spec says 6 degrees field of view - so assume its a narrow path... but at 6 degrees looking up - still hits a large amount of space...
Currently I just have the sensor in the open - wanted to experiment to see how well it worked before I looked at how I would mount it - ie. does it matter or change the results if its under glass - I am unsure at this stage. Could not find any reference to this - so just do not know....
I have seen some sensors where it looks like a telescopic lens coming of it - probably to help with the range of vision - certainly that would fill up with water and dust. My cheap one is just a surface mounted IR with a bubble over it - in theory I might be able to drill a hole that size, ensuring the sensor was above the surface and completely silicon around it....
Early results are showing promise...
I had it pointing out on an angle to the sky - to avoid the weather vane sensors... made no difference. Already thinking do I want to mount it like the lighning sensor - so it does not get disturbed by the weather vanes - but testing does not seem to show any impact of the wind speed / wind direction vanes causing any interference.
Do I want to mount at the top of the pole - so many choices...
Still experimental
|
|
|
|
Post by SDL on Nov 6, 2020 10:36:04 GMT -8
6 degree view? Hmm. That's a pretty narrow view of the sky, but not insignificant.
I think this is really interesting project.
BP
|
|
impactoz
Full Member
Posts: 142
Raspberry Pi: Yes
|
Post by impactoz on Nov 6, 2020 17:14:48 GMT -8
Its still early days in the analysis... and I thought the same thing 6 degrees does not cover much sky...
Playing with it still - I have it pointed straight up, and then I tried on a 10 degree angle outwards from straight up - does not make a difference (not sure where I would mount it - and was concerned the weather vanes might cast a shadow - does not seem to)...
Even though its only 6 degrees field of view. The temperature differences are still significant. If all I am trying to determine is clear sky / overcast (where the sky is full of cloud) / cloudy (partly blue, partly cloudy) - then the 6 degrees appears to be enough...
I have seen clear blue sky - and thats an easy one... I have seen complete cloud cover and the difference in temp is remarkable... but on a cloudy day - where the sensor points - if its cloudy or clear... there must be enough water in the atmosphere - as it is not showing one or the other, it shows in between...
What I need to work out yet - is on a cold day... say 10 degrees, does that mean 20 degrees difference means I get -10 or what. Is it linear or not ?
Further testing - but its fun!
|
|
impactoz
Full Member
Posts: 142
Raspberry Pi: Yes
|
Post by impactoz on Nov 8, 2020 12:40:30 GMT -8
My experimentation continues... Typical Melbourne weather extremes!
I now have the IR reading of the sky being logged in the database along with all the other data - like outside temperature, so I can see my samples for the whole night now, and also look at the pictures were taken.
I had clear skies all night and all day....
During the day it was easy - for clear skies there was always a difference of 20 degrees or more.
However over night the difference dropped to as low as 16.5 (IR was -6 and outside temp was 10.5)
I need to keep watching this to determine what is more accurate...
a) Night time has a different formula to day time
b) The difference to determine clear sky needs to be lower than 20 - maybe 15 ?
c) Do I work on percentages - Greater than 50% is clear sky...
A few more days / weeks data is required!
|
|
|
|
Post by SDL on Nov 9, 2020 10:53:44 GMT -8
Have you looked at the correlation between humidity and your data? That could show some interesting data that could work into your equations.
BP
|
|
