Topic: WS-2902D reading half wind speed

[DCH3416]

I purchased an Ambient Weather WS-2902D a few months back and have been fairly pleased with it so far. I have logging set up, which is sending data to my local server over wifi.

Examining the previous months of records after some recent storms. I found the wind readings to consistently be exactly half of the nearby ASOS. Which is strangely specific, even after accounting for being in a neighborhood.

I came across this post searching around the internet: https://www.wxforum.net/index.php?topic=28713.msg278935#msg278935

First I looked at the "average windspeed" sensor.  Looking at the raw bytes, my initial guess was that it is:  the quantity of revolutions of the anemometer in the 16 second interval.  However, I was somewhat shocked to find that its actually and only:  the quantity of revolutions in the last 8 seconds of the 16 second interval; and all revolutions in the first 8 seconds are ignored.  I tested this quite thoroughly, doing the below a number of times using different rates of spin of the anemometer:

If I hand-spun the anemometer *only* in the first 8 seconds of the 16 second interval, the sensor always reported 0 revolutions aka 0 windspeed no matter how many clicks of the reed switch I heard.

If I hand-spun the anemometer after first letting 8 second elapse with no spin, the sensor reported as revolutions the exact quantity of clicks I heard from the reed switch.

If I hand-spun the anemometer briefly in the first 8 seconds while counting the clicks, then again briefly in the second 8 seconds while counting again the clicks, the reported quantity of revolutions from the sensor was ONLY the clicks from the final 8 seconds.

Given that apparently the anemometer is apparently sampled for just 8 seconds out of the 16 seconds between transmission intervals. I recon the display/wifi thing that come with the sensors is not accounting for that. And is assuming that it's receiving 16 seconds worth of wind readings. Hence why I'm seeing half.

Is anybody else out there with a WS-2902 or similar observing similar behavior?

[Gyvate]

you seem to be comparing apples with pears
if you read the post you are referring to closely, you can see, that the array referred to is the WH24 (the "boat"), an old array which is no longer produced. What you have got with your WS-2902 is the WH65 array which behaves differently
Ambient write on their website regarding wind readings of the WH65 array:
"The WS-2902 weather station sensor array wirelessly updates wind speed and direction to the console every 16 seconds.
The sensor array updates wind speed and direction once per second and then calculates the 16 sample average (wind speed) and the peak or maximum wind speed in the 16 second update period (wind gust) and passes this to the display console each transmission period.
The resolution of the sensor is about 1.1 mph. Thus, the wind gust has a resolution of about 1.1 mph.
The wind speed, since it is a 16 sample average, has a resolution of 0.1 mph.
The wind direction is transmitted real time every 16 seconds, and is not processed like the wind speed is.
The wind gust displayed on the console, unless viewing Min/Max, is the Wind Gust for the last 16 second period not for the day"

Even if your array for some reason discarded the first 8 seconds of the 16 second interval, half would not make sense as wind speed is not a cumulative observation (except for windrun). You would only get the average of the last sampled 8 seconds and the highest value here as gust. And then  some higher or lower values could escape the detection for forming the average or the gust value (= max wind speed in the sampling interval). That this would systematically be 50% is very unlikely.
Comparing wind speed with neighboring stations is not something which is accurate as
- wind like rain is not a homogeneous phenomenon
- the location of the sensor has to be at the same height above ground
- potential obstacles responsible for turbulences have to be the same
- the environment influencing wind behavior has to be identical
all things which are unlikely to be the same

Beyond that you can of course check if your anemometer (the cups) turns without resistance - maybe some cleaning is needed or something creates an obstruction to free turning ...

[DCH3416]

Even if your array for some reason discarded the first 8 seconds of the 16 second interval, half would not make sense as wind speed is not a cumulative observation (except for windrun). You would only get the average of the last sampled 8 seconds and the highest value here as gust. And then  some higher or lower values could escape the detection for forming the average or the gust value (= max wind speed in the sampling interval). That this would systematically be 50% is very unlikely.

The theory I am using is the sensor is sampling the cup rotation 8 times for 8 seconds. And then zeros for the remainder of the 16 second period.

Which might look something like this:
{1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0}

Where 1 is a one second reading of the anemometer sensor, how fast it's spinning.

That's bundled up and transmitted to the receiver which I believe is taking that array, totaling up and dividing by 16. So effectively half of the averaged data would be zeros. I haven't verified that in terms of what's being sent over the air. Just thinking of how companies tend to behave with firmwares and such, where they have quirks like that across models.

I was also looking at formulas to calculate physically (by video taping) how fast the wind cups are spinning and correlating that against the data I have logged. What I might do is take the sensor down on a calm day and just drive it down the highway and see if the wind speed matches up to how fast I'm going.

I'm reasonably confident it's not a mechanical issue though. If there's an issue at all. The sensor is only a few months old and I probably would've felt that during installation.


While there are obstructions in the area. It has a good 100+ feet radius around it give or take a tree trunk. And a clear westwardly view. I just have a hard time rationalizing the wind losing that much energy. Especially compared to nearby users, one of which has a WS-5000 and that seems to report more reasonable figures.

[Rover1822]

While there are obstructions in the area. It has a good 100+ feet radius around it give or take a tree trunk. And a clear westwardly view. I just have a hard time rationalizing the wind losing that much energy. Especially compared to nearby users, one of which has a WS-5000 and that seems to report more reasonable figures.

You would be amazed how fast wind speed drops with a little turbulence. I have a wind turbine and have encountered it. A difference of 20' (feet) in height is massive, the closer you are to ground oh it is just horrible on how much you lose. a 100' clear radius is not even close to what you need for accuracy, although, you can go higher to mitigate.

And there is no guarantee that whom you are comparing to, is installed correctly, as you can also get false highs, not really false as that is what the sensor reads, but induced higher, such as between buildings or geographic  features that cause a funnel effect.

[Gyvate]

@Rover1822DCH3416:
regarding the sampling example, that's what a WH24 would do and its console being aware of that would multiply by 2.
The modern consoles expect a 1:1 from the anemometer (16 samples over 16 seconds for the WH65/WS69/WS68) and know about the sampling and transmission intervals of the WS80s and WS90s - that's why you find the "boat" option to be tagged in the console device settings if you are still using a WH24 with a modern console e.g. a GW1100.
But you will most likely not have a WH24 but a WH65 => either the sensor location is responsible for low(er) measurements or it's a hardware issue:
either physical obstruction of the cups rotation mechanism or something else in the electronics - e.g. a no longer fully operational reed switch etc.

[DCH3416]

You would be amazed how fast wind speed drops with a little turbulence. I have a wind turbine and have encountered it. A difference of 20' (feet) in height is massive, the closer you are to ground oh it is just horrible on how much you lose. a 100' clear radius is not even close to what you need for accuracy, although, you can go higher to mitigate.

I installed high enough to clear nearby buildings but not so high that trees would start to be a problem. I guess unless I were to stick it onto fairly large structure. Of course then are you getting real world information or just going to a spot that has more wind.

And there is no guarantee that whom you are comparing to, is installed correctly, as you can also get false highs, not really false as that is what the sensor reads, but induced higher, such as between buildings or geographic  features that cause a funnel effect.

Yeah one other user appears to have the wind direction orientation off by 90 degrees. Would be nice if these units had a compass or similar mechanism built in.

Just wish I would've tested the sensor further before installing. So I could have some degree of knowledge that what it's reading is correct. Almost need to run some sort of fluid dynamics simulation just to see how the air sloshes over the houses.

[Rover1822]

@Rover1822:
regarding the sampling example, that's what a WH24 would do and its console being aware of that would multiply by 2.
The modern consoles expect a 1:1 from the anemometer (16 samples over 16 seconds for the WH65/WS69/WS68) and know about the sampling and transmission intervals of the WS80s and WS90s - that's why you find the "boat" option to be tagged in the console device settings if you are still using a WH24 with a modern console e.g. a GW1100.
But the OP will most likely not have a WH24 but a WH65 => either the sensor location is responsible for low(er) measurements or it's a hardware issue:
either physical obstruction of the cups rotation mechanism or something else in the electronics - e.g. a no longer fully operational the reed switch etc.

I was just going with the comparing to another location, not the device differences. Just the hazards in comparing to another station

[Gyvate]

@Rover1822:
regarding the sampling example, that's what a WH24 would do and its console being aware of that would multiply by 2.
The modern consoles expect a 1:1 from the anemometer (16 samples over 16 seconds for the WH65/WS69/WS68) and know about the sampling and transmission intervals of the WS80s and WS90s - that's why you find the "boat" option to be tagged in the console device settings if you are still using a WH24 with a modern console e.g. a GW1100.
But the OP will most likely not have a WH24 but a WH65 => either the sensor location is responsible for low(er) measurements or it's a hardware issue:
either physical obstruction of the cups rotation mechanism or something else in the electronics - e.g. a no longer fully operational the reed switch etc.

I was just going with the comparing to another location, not the device differences. Just the hazards in comparing to another station

No worries - I confused the addressee - corrected now 

[ASOSWX]

Keep in mind if you're comparing to the ASOS Metar winds that is an averaged reading.  You would need to look at the 5 sec or 5 min data which you may not have access to. 

[DCH3416]

Keep in mind if you're comparing to the ASOS Metar winds that is an averaged reading.  You would need to look at the 5 sec or 5 min data which you may not have access to.

Yes, I accounted for that. I actually compared peak wind gusts.

It makes sense, the ASOS is on a giant field where as I'm reading below the tree line. I just wish I had a reference to compare against to validate the data I'm seeing.