Like it, thank you.
Gary
Just remember that the 4.23°C given in the above example is 39.614°F, but only having three significant figures means that only 39.6°F can be used so you are back to just tenths for the precision when using the Fahrenheit scale.
And depending on which sensor they use, they’re going to be in the range of 0.1 degC to 0.5 degC accuracy for the underlying sensor (ie, 0.2 to 1.0 degF)
now if the temperature rises, do you get ten different values going from 4.23 to 4.24, 4.25 … up to 4.33 or is the next value 4.33 with no intermediate results? The specs tell us it goes in steps of 0.1C (in fahrenheit the steps are even 1.8 times bigger).
My suspicion is that it goes in steps of 0.1, but as the tempest has some algorithm to compensate for the sun heating the instrument or wind cooling it, they report more digits, so the conversion to fahrenheit can be rounded correctly (as it doesn’t make sense to round the intermediate results). So you can get results like 4.24 but that would be due to change of sun/wind conditions,
So the answer would be still no, it doesn’t make sense to report more than one decimal after the decimal point.
And for weather conditions itself it never makes sense in the first place.
1 Like
Unintentionally, I still have the listener running since my first check, so I have about 12 hours of data logged to console. Here are the last 10 minutes
Air Temp: -10.4
Air Temp: -10.44
Air Temp: -10.46
Air Temp: -10.45
Air Temp: -10.46
Air Temp: -10.46
Air Temp: -10.44
Air Temp: -10.45
Air Temp: -10.48
Air Temp: -10.48
great. and I love these kind of experiments. It is of course unclear if this is due to change in wind/solar or just temperature. A good test would be to put in for example in a freezer, let it get used to it for 5 minutes and then check how the values drop. There is no light and no meaningful wind. If it still goes in 0.01 degree steps, than the specs are wrong.
Taking it down and putting it into the freezer is going beyond what I’m willing to contribute to this thread. But should I remember this the next time it is night time and the wind speed is calm, I’ll chime back in.
With that said, I don’t think the specs would be “wrong” in stating .1 deg since that is the reported value for most practical purposes. I don’t think they would want to spec it at .01 resolution, even if it does do it, when they are only displaying .1 as that would almost certainly lead to people hounding them “Why is my device only showing 1 decimal?” Even more so if Tempest is only confident in an accuracy of .1 despite its precision being more fine.
the stated accuracy for tempest is +/- 0.2 degree celsius under controlled conditions. If that is still the case for changing solar/wind is a bit unclear.
Excellent - that’s exactly the page I hoping to find somewhere. Thanks !
Just for comparison - here are the Davis specs for the VP2 and Vue models, via the links on the Davis website.
for temperature those specs are slightly worse than the tempest
sigh - you do like to argue don’t you…
At a similar price point all vendors are going to use similar or identical sensors so under the hood the precision can’t be ‘that’ different at the hardware level.
- ds18b20 is +/- 0.5C
- dht22 is +/- 0.5C
- max30208 is +/- 0.15C
- tmp117 is +/- 0.1C
Pick a chip you like. They’re all in the same range. Nobody is going to read better than 0.1 degC precision (0.18 degF)
just to make it clear there is a difference between accuracy and precision.
For the tempest the stated accuracy is +/- 0.2 C and the precision is 0.1C.
well, you were the one that wrote to do the experiment and listen to what the hardware tells us. Currently the hardware does change the output in 0.01 degree steps. I tried to come up with an explanation why that could be the case with a sensor that goes in 0.1 degree steps and a little experiment that could possibly confirm that.
I definitely like doing experiments.
Right now it is completely dark, temperature is near freezing, and my wind speed average is 0.5, with gusts of 2mph.
The reported temps are coming in at 2 decimal places.
1.89
1.88
1.88
1.87
well, that looks as if the unit indeed does deliver 0.01 steps! great.
Is it useful for weather info? I don’t think so, but it is good to know.
Is that something that is Unique to that particular Tempest device?? I’ve never seen temperatures on any of my devices deliver more that 1/10th of a whole degree. i.e right now. . .34.5° F. T’would be nice if he gave his StationID number. . .I’d like to examine that a little more closely.
I imagine this is the raw data coming from the API, and it will be the same for all Tempest devices. It is rounded in the app/webpage to a tenth of a degree, as the sensor does not have the accuracy to report to a hundredth of a degree
2 Likes
I suspect it’s an average of multiple more frequent internal-only readings over a 1-minute reporting period…
2 Likes
This is the raw data that is coming from the sensor. It sends that data to the hub and the hub broadcasts it on the local wifi network at the same time it sends it to WeatherFlow. I’m using a listener program running on another device on my network to get the data.
The data you see in your app is what the hub sent to WeatherFlow whom formatted it and sent it to your mobile app. I’m sure they do this because, as mentioned by the other commenter, that level of precision doesn’t have a whole lot of use cases. For most users, the difference between 56.7° and 56.69° means nothing, and it’s also probably outside the margin of error for the devices sensor making it irrelevant anyway.
1 Like