Hot-Wire Wind Sensor

[hschir]

The gurus suggested I start a new topic for this as I included it in my post in sequences etc on multiple ftp uploads to a website - which I'm still trying to resolve.

Anyhow I found this nifty hot-wire wind sensor that sells for about $17.00 US and I had a couple sent through the mail to me here in Australia.

After a solid re-evalution I have revised my conversion formula for the modern device hot-wire wind sensor (http://shop.modernde...cts/wind-sensor). There's a schematic on that site and some general information. I'm running it on the 5V supply on a LabJack analog channel.

So far I have two of these devices, but I intend to get a few more and place them at various heights to obtain a wind profile.

While a calibration would have been better against a certified calibrated anemometer (preferrably hot-wire) - I only have a Lutron LM-8000 hand held unit with 6 shaft propellor type vane and STILN windwatch - neither of which are particularly precise, but between the two I was able plot wind speed versus voltage output of the sensor and apply Kings law to what I am confident is as good as I can get it - say within +/- 0.15 m/s.

The formula I am using in converting the voltage output to m/s is:

((V^2 - 1.4641)/4.9442)^1.7746

These values were obtained from the regression line of the plot of V versus m/s readings I took over 0 to 2.0 m/s

[hschir]

The gurus suggested I start a new topic for this as I included it in my post in sequences etc on multiple ftp uploads to a website - which I'm still trying to resolve.

Anyhow I found this nifty hot-wire wind sensor that sells for about $17.00 US and I had a couple sent through the mail to me here in Australia.

After a solid re-evalution I have revised my conversion formula for the modern device hot-wire wind sensor (http://shop.modernde...cts/wind-sensor). There's a schematic on that site and some general information. I'm running it on the 5V supply on a LabJack analog channel.

So far I have two of these devices, but I intend to get a few more and place them at various heights to obtain a wind profile.

While a calibration would have been better against a certified calibrated anemometer (preferrably hot-wire) - I only have a Lutron LM-8000 hand held unit with 6 shaft propellor type vane and STILN windwatch - neither of which are particularly precise, but between the two I was able plot wind speed versus voltage output of the sensor and apply Kings law to what I am confident is as good as I can get it - say within +/- 0.15 m/s.

The formula I am using in converting the voltage output to m/s is:

((V^2 - 1.4641)/4.9442)^1.7746

These values were obtained from the regression line of the plot of V versus m/s readings I took over 0 to 2.0 m/s

I have to say that I have my reservations of this "evaluation" - basicaly because the circuit uses an NTC Thermistor without the specifications disclosed so that is my next avenue of investigation to arrive at a defintive conversion formula - bear with me ...

[hschir]

The gurus suggested I start a new topic for this as I included it in my post in sequences etc on multiple ftp uploads to a website - which I'm still trying to resolve.

Anyhow I found this nifty hot-wire wind sensor that sells for about $17.00 US and I had a couple sent through the mail to me here in Australia.

After a solid re-evalution I have revised my conversion formula for the modern device hot-wire wind sensor (http://shop.modernde...cts/wind-sensor). There's a schematic on that site and some general information. I'm running it on the 5V supply on a LabJack analog channel.

So far I have two of these devices, but I intend to get a few more and place them at various heights to obtain a wind profile.

While a calibration would have been better against a certified calibrated anemometer (preferrably hot-wire) - I only have a Lutron LM-8000 hand held unit with 6 shaft propellor type vane and STILN windwatch - neither of which are particularly precise, but between the two I was able plot wind speed versus voltage output of the sensor and apply Kings law to what I am confident is as good as I can get it - say within +/- 0.15 m/s.

The formula I am using in converting the voltage output to m/s is:

((V^2 - 1.4641)/4.9442)^1.7746

These values were obtained from the regression line of the plot of V versus m/s readings I took over 0 to 2.0 m/s

Please ignore the conversion formula - I have discovered it is incorrect, although the basis of Kings Law is not. The issue is that unlike a normal hot wire anemometer that uses a platinum wire with a positive resistance to temperature coefficient - the hot wire sensor described uses an NTC Thermistor which is totally different ball game - I'm working on it ...

[hschir]

After weeks of attempting to calibrate the “hot-wire wind sensor” and attempt to resolve a conversion formula between voltage versus wind to a high degree of confidence, I can confirm that I’ve reached that outcome to my satisfaction.

Based on pages 150 – 151 of the ‘weather-station-instruments’ (http://www.syscompde...instruments.pdf ) and Okamoto (http://ir.lib.shizuo...1/090526015.pdf) papers and by experimental analysis using my homemade wind-tunnel (see attached photo’s), I was able to adjust the value for k in the equation on page 150 and apply an offset that gave me true readings at 0 m/s and 5 m/s.

Instead of using the Out signal from the op amp of the circuit (attached), I used the raw voltage output from the Wheatstone bridge circuit as I see no need for a buffer amp for impedance matching purposes to the LabJack input and also, as it gives greater precision in the lower region velocities of interest.

I then compared readings between the vane anemometer and the “hot-wire” anemometer for various wind speeds over the 0 – 5 m/s the result being a deviation by the “hot wire” device and the vane anemometer between negative 0.1 to 0.2 m/s. IMHO that’s as tight a precision as can be expected from a device of this type and given the cost of the device ($17 US).

With k = 2.23 and an applied offset of -0.4, the final conversion formula arrived at is: U = (V/2.25)^[(8.3^-1)^-1]-0.4, where U is the wind speed and V is the voltage across the thermistor.

So now that I have a formula – I can calibrate each individual sensor according to this formula, zeroing them using the offset value as required.

I have to revisit the calibration of water level sensor’s using a Wheatstone bridge rather than a simple voltage divider as I presently have them configured because a change in resistance of the level sensor corresponds to a change in water level and a Wheatstone bridge will improve the precision of these measurements in a similar fashion as in the wind sensor. I have a couple left that I haven’t permanently installed yet to refine the calibration of these.

One final point. The wind sensor is directional, that is it mostly senses wind velocity in one plane, which is fine in an air conditioning duct, but if used outdoors you would probably need at least 2 to compensate for changes in wind direction and chosing the highest value reading as representing true wind velocity.

WindSensorSchematicRevC1.pdf

[randi402]

I have built my own sensor. I have yet to calibrate it, but was more interested in relative flow. I appreciate your research here. The formula will be useful. I have used a NTC thermistor of 470 ohms at T25C. It is driven by an LM317 in a constant current configuration with a 1W 25 ohm resistor to produce 0.052mA. The source voltage is 12VDC 500mA wall transformer type power supply. I divided the voltage across the thermistor over a 10K to 2k divider to reduce the voltage below the U3's 2.44 VDC limit. I compare the flow stream thermistor with an identical circuit in a non-air flow shielded cage. This compensates for the air temperature. It seems quiet sensitive to very low flows. I wasn't smart enough to find a $17 sensor like you did. I have multiple circuits and will cost about $7 each through DigiKey.

[randi402]

I have tried to calibrate the wind sensor I made with terrible results. I tried using my wireless weather station to calibrate but the readings are transmitted at intervals and very so widely it was useless. I then did a calibration with th sensor arrays out the window of the car. In the end I concluded that each sensor acts so differently based on the amount of contact it has with the structure that the curves don't fit. My application is inside a biologic test chamber where it is very moist. I have to seal up the leads to prevent electrolysis, shut current paths, corrosion, etc. so the sealant effects the readings. I found that my thermistor was running about 150 degrees F in dead air room temperature. This is well above my environment. It turns out that at that that in the range of operating temperatures my thermistor will heat from 150F to almost 220 F. In that range the thermistor is running in the linear potion of the curve. I can calibrate it at a few points on dead air. This gives me a slope and offset to subtract from the flow sensor reading based on the temperature. I haven't done the calibration yet but have confidence it will work. I will post the results when I have them. I really appreciate the efforts of the others who have posted in this thread. I have bought a couple of the above mentioned sensors to compare them.

[randi402]

After weeks of attempting to calibrate the “hot-wire wind sensor” and attempt to resolve a conversion formula between voltage versus wind to a high degree of confidence, I can confirm that I’ve reached that outcome to my satisfaction.

Based on pages 150 – 151 of the ‘weather-station-instruments’ (http://www.syscompde...instruments.pdf ) and Okamoto (http://ir.lib.shizuo...1/090526015.pdf) papers and by experimental analysis using my homemade wind-tunnel (see attached photo’s), I was able to adjust the value for k in the equation on page 150 and apply an offset that gave me true readings at 0 m/s and 5 m/s.

Instead of using the Out signal from the op amp of the circuit (attached), I used the raw voltage output from the Wheatstone bridge circuit as I see no need for a buffer amp for impedance matching purposes to the LabJack input and also, as it gives greater precision in the lower region velocities of interest.

I then compared readings between the vane anemometer and the “hot-wire” anemometer for various wind speeds over the 0 – 5 m/s the result being a deviation by the “hot wire” device and the vane anemometer between negative 0.1 to 0.2 m/s. IMHO that’s as tight a precision as can be expected from a device of this type and given the cost of the device ($17 US).

With k = 2.23 and an applied offset of -0.4, the final conversion formula arrived at is: U = (V/2.25)^[(8.3^-1)^-1]-0.4, where U is the wind speed and V is the voltage across the thermistor.

So now that I have a formula – I can calibrate each individual sensor according to this formula, zeroing them using the offset value as required.

I have to revisit the calibration of water level sensor’s using a Wheatstone bridge rather than a simple voltage divider as I presently have them configured because a change in resistance of the level sensor corresponds to a change in water level and a Wheatstone bridge will improve the precision of these measurements in a similar fashion as in the wind sensor. I have a couple left that I haven’t permanently installed yet to refine the calibration of these.

One final point. The wind sensor is directional, that is it mostly senses wind velocity in one plane, which is fine in an air conditioning duct, but if used outdoors you would probably need at least 2 to compensate for changes in wind direction and chosing the highest value reading as representing true wind velocity.

Wow I love what you are doing. My daughter went to UNSW for her Masters. One of my other daughters refers to me as Dr. Brown from the movie Back to the Future. I think it's the hair. Seeing your apparatus I had some comments, It looks like you are controlling wind with a baffle made of cardboard. I did a lot of work with flow meter calibrations and this can present several problems. The air flow in a tube is a profile, dead at the tube surface and fastest in the center. Putting a bluff body in the tube not only interrupts the flow profile it creates vortexes. There is a whole system of flow meters that counts the number of pressure changes per second behind a bluff body. It might work pretty well for you. I helped design on for helicopter wind speed. My point though is that will cause pulses of air speed that may average out, but also may give you noise in you signal. You generally want ten diameters of straight line pipe ahead and five below the meter for the profile to stabilize. That said I may have a better solution for you. I bought this fabulous blower from this guy on ebay. http://www.ebay.com/itm/330850667577?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1423.l2648 $18 US lus shipping. The thing converts 120 VAC to DC and has a speed control in it. It can create 400 Watts of wind. But the great thing is you can use the DAC channel with an OpAmp or a potentiometer to control your flow rate. The auction is ended but I think he has two left, I'm trying to get him to re-list them. I want at least one but you may want one too. They are used, but they cost $780 US new. they are the Windjammer 116630 without the input flange. If I can attach the PDF I will but you can jump through some hoops and get it here. http://www.ametektip.com/index.php?option=com_content&task=view&id=173&Itemid=2

[hschir]

Thanks Dr. Brown

You make some good points, particularly wrt to the wind tunnel. I will be revisiting the wind sensor scenario shortly, as I have been tied up with my PhD studies.

The 'wind-jammer' looks like a good way to go, an alternative available locally which would be easier for me to obtain would be from:

http://www.jaycar.com.au/productView.asp?ID=YX2532&form=CAT2&SUBCATID=981#1 and flow rate could be controlled using

http://www.jaycar.com.au/productView.asp?ID=MP3209&form=CAT2&SUBCATID=1002#3

Having said that - I would be very interested in what you finally come up with ...

[randi402]

Hschir: You are probably right about the motors. Though I think you could control speed with just a variable voltage regulator that can handle two amps for less money. I have a completely different problem than you. You are creating a wind tunnel to cal something going on a pole, you have dry air. All you need is the right speed at about 3/8 of the way to the center (profile correction for average). The fact is you can achieve that with your current apparatus if you just put 10 diameters of straight pipe ahead and five behind. I have a continuous process where I need to lift one meter of wet bio-mass and pull humid air through it. The motor you selected can lift 2.5 cm of water column. I need one that can do a meter of water column. My motor is 400 Watts and can blow or suck 125 cm of water. I just love the thing cause I got it $0.02 on the dollar. You really don't need it unless you want a high speed wind tunnel.

I have been thinking about your orientation problem. Have you considered mounting the devices on weather vanes with three slip rings. You could make slip rings out of copper pipe slipped over a wooden dowel. Either use graphite motor brushes or little spring wires as contacts. It's not like this thing is going to be spinning at high speed it's just aiming at the wind. Put a capacitor between your ground and positive on the vane side to make up for contact power noise and one on the signal output on the tower side for signal noise coming back. You could put a spray paint cap on the shaft over the slip rings to protect it from the weather. You are pretty cleaver I'm sure you can put something like that together.

I really appreciate all the work you have done on the algorithmic. It will really simplify my work. Good luck on your PhD. What are you getting it in, if I might ask? My wife and I both wanted to be meteorologists. Years ago she did the amazing feet of winning the West Virginia State Science Fair all four years in a row. If we can help let us know.

[randi402]

hschir: Am I missing something? You say the k = 2.23 in your formula. I see the 0.4 offset and a denominator of 2.25 but no 2.23. Are there some calculations I'm not seeing or is that a slight typo?

I'm also curious about how you dealt with two other problems. The makers are saying the Raw Velocity signal won't go below 1.8V. So:

A) How are you dealing with the 2.44 Volt input limit on the LabJack Analog inputs or do you have the High voltage type LabJack?

Is this output inverse, i.e. it is low at high speed and high at low speed or is it positive going?

[hschir]

Randi402:

1. I'm doing a PhD in applied physics - thanks for your 'good luck' sentiment ...

2. You are right, the manufacturer does say that V_Raw won't go below 1.8V - refer to attached cct. diagram and document I wrote describing how (I think) the circuit works ...

Hope that helps !

The Modern Devices Wind Speed Sensor.pdf

[randi402]

HSCHIR: You might want to consider a simpler approach. After reading all the links you suggested above it appeared your formula (U = (V/2.25)^[(8.3^-1)^-1]-0.4) might be reduced. By using the Velocity output and zeroing it I could ignore the offsets. The result reduces to something of the form V = (E/k)^x

Where V equals Velocity, E equals the voltage at the Velocity output connection, k is an amplifier quality that sets the line slope, and x is the exponent or the equivalent of the ^1/8.3 term in your equation. x is used to flatten the curve of the data.

This description gets complicated. I have four flow meters in my system essentially two redundant sets. One in each pair is a thermistor flow meter of my own design, using 12 VDC and an LM317 to create a constant current of 100 mA through the 470 Ohm thermistor. I run the voltage across the thermistor through a voltage divider and feed it into a LabJack input. My device is quite temperature sensitive, but pretty linear over the ambient temperature range of operation. I take the local temperature and use a standard slope offset correction to zero/correct the thermistor reading for temperature. The Modern Devices board is pretty immune to temperature in my testing so far. In each pair there is one flow device of each type. I can make the flow operate in a closed loop.

The blower is incredible. Very linear and operates with a control voltage of 0-10 volts, but will not turn below 2.5 VDC on the control input. It's almost 1 HP on a 3 cm tube. If you have seen the movie Space Balls, when Dark Helmet throws the spaceship throttle all the way forward and they pass light speed into Plaid. This thing does plaid. The reason I mention this is that current heated Thermistors flow graphs have an arc that goes up rapidly and then asymptotes to a flat line. When the velocity gets this high, they have all kinds of thermal dynamic effects that cause the flow reading to actually peak and go back down.

That said here is what I did. I made a closed loop with the devices in their actual operating locations. I ran the loop at slow speed to settle the temperatures. I stopped the flow and I zeroed my four sensors. I then started logging data. I took twenty plus data (once a second) at each flow rate. I raised the blower to 3 VDC, and let it settle for a few seconds. Then I took 20 plus data points, Increase to 4 VDC, settle, take data and so on to 10 VDC. I averaged the 20 points for each of the four Velocity devices and temperatures. This left me with missing data points for 1 and 2 VDC respectively. I just entered 1 and 2 on the averaged Excel data table. I made a formula to calculate the corrected velocity for each point with the denominator and exponent as data above the respective columns. That formula took the average data for a voltage setting, divided it by the denominator, and raised it to the power of the exponent. This gave me a rough corrected flow rate with two fake data points. Those fake data points worked well, because zero is zero, and those two points established a desired slope line.

Below are two links where you can see the spread sheet summary and graphs. The graph at the right of the table is one made from uncorrected data on ModernDevice1. You will note a straight line up to 2, a hook, and then a curve. At this point I play with the Exponent to get a straight line up to 6 VDC (above this point the data goes down instead of up). I don't need to calibrate above 6 VDC because it causes the air to get hot and I need it to cool the process, I'll never operate at that flow. Then I adjust the denominator term to get the two lines to line up, removing the hook. After a few iterations of the two terms the line crosses at 0, 1, 2, 3, 6 and is pretty close at 4 and 5.

​The numbers vary because my sensors are coated to protect them from saturated humidity. That caused the thermal inertia to vary with each probe. Below the table you can see the corrected graphs for each sensor. Basically when I put in 5 VDC to the blower, all four flows read 5. This could easily be done with meters per second instead of control voltage.

http://randicuster.blogspot.com/2013/05/blog-post.html?token=4QpOkT4BAAA.DfKcjms8K_ycmTVjW1kkQg.arbUWhR85iuiJsNGXZNUBg&postId=3066212422232060224&type=POST