I checked that with the manufacturer and he said it only operates reliably and accurately above - 20C below that they do not think it will break but they don’t guarantee that you get accurate values.
For the upper limit it works reliably up to 60C and could break above 80C.
Thanks for the quick reply!
I can probably just create a detachable mounting point and bring it in if the cold becomes too much 
.
Thanks as well for building a great product that leverages open hardware and software!
Love the work you folks do 
.
I might be able to test below zero temperatures. I don’t have an outdoors kit yet, but I plant to get one and can report back here this winter. 
Yeah same, I think -20 might be good enough as it is sheltered on my balcony, and I am building a box for it (to add weather protection) would most likely disrupt the readings for the sensors .
So uh. I have the kit outdoors now, and there’s certainly some issues. Maybe I haven’t installed it correctly, but the temperatures are way off from the official weather numbers.
Here’s a screenshot from the last week, hourly:
There you can see it’s mostly above zero all week, and always above zero for the past 24h. The weather is actually very different. According to the official report, temperature has not gone above -3C in the last 24h, and it has barely gone above freezing for the past week.
More specifically, the max in my airgradient dashboard is 7.9C but the max according to the weather office is 1.9C! A 6C difference… I would show more graphs, but Discourse thinks I’m a new users and i can’t :p.
So there’s something wrong with the sensor… I wonder if the LED might be heating up the enclosure, it’s so bright!
Looks like they are aware of it and working on a compensation algorithm, but so far, it hasn’t made it into the Arduino code
A Simple Linear Correction Algorithm for Temperature (airgradient.com)
Well that is certainly interesting! That post mentions:
Important: Before we come up with globally recommended temperature correction algorithms we need more data across broad temperature ranges, various locations and all seasons of the years. We will keep you informed as we introduce these more general correction algorithms.
If you have some data that you think could help us developing these, please feel free to share with us!
I wonder: what kind of data? How do you get the control value? And how do we share?
We are collecting data at some university sites that co locate our monitor and have also precise temperature instruments.
We still want to collect a bit more data sub zero and then calculate and publish recommended algorithms.
Interesting! Well I’m not at university and I haven’t been for a while, so I can probably not provide that level of precision. But I can provide the cold weather! And if you’re interested, I might be able to find people here with reliable sensors you could ship an AG to for comparison. I do have good contacts in the university world around here… I have also contact at environment canada for what that’s worth…
Let me know!
Great. Could you please get in touch with me through the support form on our website.
It has been consistently below freezing where I live, and our outdoor unit has not registered much below freezing at all. Is there a time frame when the temperature correction algorithm might be implemented or anything I can do to help with the data collection? I also have an indoor unit I can take outside.
I discussed it with our science team and we would like to have something in the next two weeks.
To understand your data better, can you please let me know if your AirGradient permanently in the shadow?
I’m getting about 10 F higher on my OpenAir compared to weather.com for my area. Unit is in the shadow under a deck awning.
Mostly yes. It’s on the first floor, and there’s a balcony above, with little light in the backyard.
Adding my 2 cents. I added a bme280 to the unit so I have some reference. The bme280 also has a slight selfheating issue but that doesnt really matter for this example. I can confirm that the PMS sensors are higher than ambient but can go below 0. Also you can see what the sun does on a clear day while temp stayed just below 0. (terras_temperatuur is a cheap weather sensor but it is hanging at the same place)
So I would say the housing traps heat giving all sensors slightly higher temperatures. And possibly also the PMS generates a bit of heat. The PMS is running on a 3 min interval.
Did you add the bme280 inside the outdoor enclosure or is it an external probe?
I was just wondering how much is just the PMS5003T being inaccurate, vs being in the AirGradient enclosure. If it was inherently inaccurate, I would have expected to see other people using some compensation algorithms for it, but if it was specific to this enclosure, maybe not.
Anyone interested/willing to open the enclosure and leave the PMS sensor out in the open for a bit to see if temp becomes more in line with reference/reliable sources?
Its inside on the free i2c spot. So just above the cutout in the blue underside.
Both have an effect but when we tested it, the pms sensors were less impacted by the enclosure heat up than a separate sensor inside the enclosure. However this was at much higher temperatures than what we are looking at right now.
Our science team is currently analyzing a lot of data and I hope I can soon share some of these results.
