Topic: Precision of WN34 Soil/Water Temp Sensor

[hiljo]

Hello,

I have a WN34L temperature sensor in my pond for some time. Never really looked at the data. Only at the graphs and value how cold/warm the pond is. Today i noticed however that the sensor's precision is not by 0.1 degrees, but 0.5/0.6 or even 0.7 degrees. I'd expect that it would be 0.1, but cannot find this in the technical documentation somewhere.

My pond temperatures today went from 16.8 to 19.8 in the following steps: 16.8 - 17.3 - 17.9 - 18.5 - 19.2 - 19.8. Also see the graph:
 [ You are not allowed to view attachments ]

Question: is this normal, or would an accuracy of 0.1 degrees be more normal (like the WH31 sensors for example)?

Thanks!

[Gyvate]

I suggest to ask Ecowitt ...
but to keep the terms clear
precision, accuracy is probably +/- 1°C
resolution would be the steps in which the readings are shown, usually 0.1 °C.
So you are not talking about precision or accuracy but about (observed) resolution

Before you come to a conclusion from your sensor readings in the Ecowitt cloud, you have to take into account that
- the sensor transmits only every 77 seconds, the data collection (loop) may be more often - Ecowitt to ask what it transmits: max, average of that interval
- ecowitt.net displays only every 5 minutes smallest time resolution, again how the data loop at ecowitt.net works will define the numbers displayed (max, avg) - again Ecowitt to be asked

also, in your question in the OP you assume that the temperature changes in (linear) steps of 0.1 degree within a five minute interval (within which may not be the case) - neither the 0.1 nor the linearity.
And even if the temperature changes let's say in steps of 0.2° C, this doesn't mean the the sensor cannot measure in steps of 0.1 °C ...
But getting more info from Ecowitt could at least make things clearer.

question to Ecowitt would then be:
a) resolution
b) accuracy
c) data collection within the 77 second transmission interval (max, avg ?)
d)data collection in the ecowitt.net loop for the 5 minute interval (max, avg ?)

[mcrossley]

Looks like you may be converting integer Fahrenheit values to Celsius

[hiljo]

but to keep the terms clear
precision, accuracy is probably +/- 1°C
resolution would be the steps in which the readings are shown, usually 0.1 °C.
So you are not talking about precision or accuracy but about (observed) resolution

Correct. Couldn't find the right word in English. We're talking about resolution here.

Looks like you may be converting integer Fahrenheit values to Celsius

Yes that happens (also on Ecowitt.net). But how does that impact the resolution?

[Mandrake]

I suspect this might be just the way ecowitt.net is capturing the data.
Looking at my data in Home Assistant I can see that the data is recorded to two decimal places (xx.xx degrees C) and no evidence that it steps in 0.5C

[hiljo]

I've contacted Ecowitt. They have passed it on tho their software team for confirmation. So I'll wait for their explanation.

[Gyvate]

once Ecowitt reply we can see further
but still, possibly all three or four factors are involved
(1) the internal capturing loop of the sensor with its output (max or avg) every 77 seconds
(2) the real change of temperature during the posting cycle
(3) the processing of the Ecowitt cloud loop for the (usually) minute based postings of (1)
(4) a °F/°C conversion rounding effect

@hiljolodewijk
as for (4) - switch your Ecowitt cloud dashboard display to °F and have a look how the resolution looks then

[hiljo]

@hiljolodewijk
as for (4) - switch your Ecowitt cloud dashboard display to °F and have a look how the resolution looks then

I have done that now. It almost seems that is uses a resolution of 1°F (but not exactly). See https://www.ecowitt.net/home/index?id=62923 - last graph on the bottom of the page.

The Fahrenheit values are always the same resolution:
62.2
63.1
64.2
65.3
66.6
67.6

No other values between these temperatures are reported (also checked 3 more previous days). So no 62.3, or 67.0 for example.

I also use a custom upload to a database. In here there are 22.500 measurements of the WN34 sensor. So I did a 'DISTINCT' list of all the unique temperature values, and in those 22.500 records, there are only 64 unique temperature values

[Gyvate]

The Fahrenheit values are always the same resolution:
62.2
63.1
64.2
65.3
66.6
67.6
No other values between these temperatures are reported (also checked 3 more previous days). So no 62.3, or 67.0 for example.

I also use a custom upload to a database. In here there are 22.500 measurements of the WN34 sensor. So I did a 'DISTINCT' list of all the unique temperature values, and in those 22.500 records, there are only 64 unique temperature values

so the resolution is 0.1 - the displayed value differences between adjacent reported readings are between 0.5°C and 0.7 °C (16.78, 17.28, 17.89, 18.5, 19.22, 19.78) or between 1.1 and 1.3 °F

It would be interesting to have a WN30 probe put next to the WN34 probe and compare its readings ....

Also, how the WN34 behaves in a "laboratory" environment, where you can raise the temperature by 0.1°C and see its performance - then you can exclude effects coming from the capturing loop

[hiljo]

Also, how the WN34 behaves in a "laboratory" environment, where you can raise the temperature by 0.1°C and see its performance - then you can exclude effects coming from the capturing loop

Yes, but as i said, i made an extraction of all the unique temperature values between now and 3 months ago (every 5 minute a record), and in those 22.500 records, there are only 64 unique temperature values. So I'm sure that a laboratory environment would be more precise, but in this case the sample size (n) in the population is more then enough for a validation.

Still curious what Ecowitt has to say, and what the experiences of other users are. Is it my sensor, or is this common?
Haven't found an Ecowitt.net dashboard with a WN34 also yet.. 

[DelChard]

I Have a WN34 Soil (electrically the same as the water).
My Ecowitt IDs are  72153 72159 & 72284.
The resolution is showing 0.1C even at the 5 minute Ecowitt interval.
I've attached a CumulusMX screen grab showing the 1 minute interval values.

I'm surprised your pond is showing such an increase over a 5 min interval.
I wouldn't expect  either of my ponds to vary so significantly.
 [ You are not allowed to view attachments ]

[Mandrake]

likewise I have both WN34 types

My Ecowitt data:
https://www.ecowitt.net/home/share?authorize=02APX6

[olicat]

Hi!

Same here - 3 WN34S and 1 WN34L.
I can clearly see temperature changes of 0.1°C (or even °F).

Oliver

[box]

Same here with my WN30

0.1 degrees

I suppose the size/shape/depth of the pond and the positioning of the probe may be a factor?

My pond is about 4 feet deep where the sensor is placed, but it varies in depth from 6 feet to 18 inches and has a surface area of around 12ft by 7ft

The main reason for monitoring temp is to determine when/if I need to heat it to protect our hibernating turtle in the winter, but it is interesting to see how the temperature changes with outside temperature/sunlight levels etc

[hiljo]

I'm surprised your pond is showing such an increase over a 5 min interval.
I wouldn't expect  either of my ponds to vary so significantly.

Well, that would point in the direction of a faulty sensor, looking at all your reactions and your experience DelChart. Waiting for a response from Ecowitt. Maybe they can confirm this.

[Rover1822]

I guess it also depends on your "pond" its volume, sunlight, depth of water where the probe is,  you have a waterfall etc.

[DelChard]

I guess it also depends on your "pond" its volume, sunlight, depth of water where the probe is,  you have a waterfall etc.

Every point you make is correct . You missed temperature differential (air to water)
A core subject of engineering is thermodynamics, and unless hiljolodewijk's sensor is located just under the surface, I can't see that sort of temperature shift happening, particularly in the Netherlands.
Check Box's Ecowitt Id, the max shift he is seeing is 0.2C over 5 min.
I've not bothered with a pond sensor, basically because of the thermal inertia involved.

I think hiljolodewijk has a sensor problem.

[hiljo]

I guess it also depends on your "pond" its volume, sunlight, depth of water where the probe is,  you have a waterfall etc.

> 1500 liters, and about 1m deep. So no, it is not a fish bowl 

[hiljo]

I think hiljolodewijk has a sensor problem.

After some e-mail contact with Ecowitt, they have shipped now a new WN34 sensor. Hopefully that will solve the problem. I'll let you guys know if this is the solution (but it has to be it). Thanks for your help and comparison with your WN34 sensors 

[hiljo]

Issue solved. Yesterday I received a new WN34S sensor from Ecowitt. This sensor nicely measures every tenth degree. So it was indeed a sensor problem  .
Not what I expected in the first place, but thanks to you all we figured this out. Thanks again! 

[davidefa]

what can I say... maybe some quality check problems in ecowitt, but certainly a super user support!