Go Down

Topic: Thermal anemometer (Read 6038 times) previous topic - next topic


Apr 27, 2010, 06:16 am Last Edit: Apr 27, 2010, 06:29 am by carl47 Reason: 1
I wanted a non-mechanical anemometer for my outdoor weather station.


We have two choices , thermal or ultrasonic.

The thermal wind sensors have been very well tested by:


Johan Liljencrants has a very good analysis of the thermal properties of these types of sensors.  Its clear from the following graph that the response to airspeed is going to be very non-linear.
This graph basically shows the cooling effect of moving air over the sensor:

The cooling effect goes from a fixed constant value at 0 to a limit of 1 at infinite speed.
Its obvious that the sensor is going to be very sensitive to slow air change and very insensitive to high air flow changes.

Thermal sensors are the most simple so I decided to do some testing myself.
There are many types on the web but I tested  the following:


I ran it for some weeks next to my WMR100 wind sensor.
This confirms the type of response I was expecting from Johan Liljencrants analysis.
Very sensitive at low speeds. You just need to lightly breath on it to get a response.
The response became very inaccurate at the higher speeds

The thermal response of the TO92 transistor is very slow for a wind sensor.
I measured the time constant as 20 seconds.
This is the time required to reach 63 % of its final value when subjected to a step change in wind speed.

There is also a direction version of this circuit:


An extensive search of the web could not find a single commercial anemometer for weather stations that uses a thermal sensor.
There are thermal sensors available but they all seem to be of the type
"Micro Gust Thermal Anemometer" or "Breath Detector".

All the non-mechanical sensors seem to be of the ultra-sonic type.
That seemed to be telling me something so I will go that way for my station.

Here are two good starting points:



All in all there is room for the thermal design in a simple weather station in which you dont need very accurate results. You will also be unable to detect gusts a running average is all that will be available.


As any politition will tell you "there is a third way"

That's to use a strain guage rosette system whereby three strain guages, placed 120 degrees apart around a circular column will respond to both wind speed and wind direction.

This is an idea I dreamed up when I was at university (40 years ago) but have never seen it in practice.

Basically get a (very) thin wall tube of stainless steel at least 100mm diameter by about 1 to 2 metres long.  Mount it vertically with a rigid base.

It would be preferable if the tube bottom was cut away to create three legs but these would have to be identical, otherwise system accuracy would be affected.   Fix the three strain guages near the base at excatly the same height and spaced 120 degrees apart.

The resultant strain figures will represent bending due to wind speed and the direction of bending will represent wind direction.

So, that's the idea, all you have to do is put it into practice



Should have made it clear the tube won't actually bend, but there are bending forces on the tube which are readily measurable by the sensors.


Sounds feasible. If it's practical why no commercial models?

Do you have any links to a working model.

I really like a head start for my projects, it's easier than starting from scratch.

On a lighter note it could also double as an earthquake sensor.


Apr 28, 2010, 07:52 am Last Edit: Apr 28, 2010, 07:53 am by retrolefty Reason: 1
The resultant strain figures will represent bending due to wind speed and the direction of bending will represent wind direction.

I really like that idea. However I assume it would have to be mounted high and in the clear. As strain gauges are low level bridge type devices I assume the op-amp interface would have to be mounted in the instrument. Actually it would probably be better to mount everything in the instrument and run a four conductor cable (+ voltage, ground, and two RS-485 serial data lines) to the display/controller.

Where I would stumble would be the math calculations for the speed/direction from the three raw strain readings. Any hints on what the formulas would look like?



Oct 04, 2010, 07:54 am Last Edit: Oct 04, 2010, 08:02 am by carl47 Reason: 1
I have finally a found solid state commercial anemometer:


I would be very interested in finding some detail of its operation.
All I can find is  that it uses a solid state sensor!

Does anyone know anything about this sensor?

Go Up