Constant Current Hot-Wire Anemometer

joeyd999

Joined Jun 6, 2011
5,236
Here is my solution based on your numbers. Here are the assumptions:

1. Rwire = 6.3 Ohms at 250C
2. Iwire in still air = 100mA
3. Pwire in still air = 63mW
4. Power range: 10 to 1. Max Pwire is set at 650mW

The power dissipated by the wire (which will be a function of ambient temperature and air speed) can be computed as illustrated on the schematic.

The wire will be maintained at 250C until the max current is exceeded (by Q2).

The Vo in still air will be about 1.630V. The Vo at 650mW will be about 5.155V.

The current into the wire will be clamped (by Q2) at about 700mA.

U1 *must* be a very low input offset op-amp, or you will get *large* errors. I recommended a chopper stabilized part (though it is surface mount only). The common-mode range should include ground, and the output needs to swing from about 1.5 to 7.5 volts. Max output current is no more than 2 ma.

You'll get best accuracy by using 0.1% resistors where indicated.

Choose the Q1 darlington for at least 5W power dissipation, and a current gain of at least 1000.

Please note high power resistors R1 and R5!

Keep lead lengths short to your hot wire...they will add errors!

The circuit will be unstable without the R6/C1 filter network. This is because there is a thermal lag between current and temperature in the wire. I guessed at the proper values. Your mileage will vary.

The circuit is *fully* protected from faults...you shouldn't have any problems. Good luck!


EDIT:

Also, you can confirm the operating temperature of your wire by putting it in an oven with a thermometer and raise the temperature. Plot P vs. T. The operating temperature is where P is projected to cross 0 mW.

EDIT 2:

The Op-Amp I specified achieves low offset errors by using internal digital switches and an oscillator. You will need an 0.1uF cap across the power supply close to the chip to keep noise from being injected into the signal.
 

Attachments

Last edited:

joeyd999

Joined Jun 6, 2011
5,236
Correction: when I said the circuit is fully protected against faults, what I meant is the wire is protected from faults. It shouldn't fry under any conditions.

On the other hand, R1 will dissipate up to 5W if the wire is short circuited. If you need short circuit protection, you need to increase the power handling capability of R1 to 5W.

Apologies!
 

Thread Starter

lfgrdwill

Joined Dec 9, 2011
27
joeyd999 - Your circuit is more complex than i was hoping for, but with that comes better measurements. I may or may not resort to something like what your drew up. I was really hoping to just throw a single resistor in a circuit with a constant current source and measure the voltage across that resistor (which is actually a hot-wire) in different flow speeds.

MrChips - I am pretty familiar with bridge set ups now. Thanks anyways for offering to upload a diagram. What i don't get is, why are they superior to a more simple circuit containing a resistor and a constant current source?
 

Thread Starter

lfgrdwill

Joined Dec 9, 2011
27
I can't even imagine. Actually i've seen some ridiculous stuff on this campus. That's why i'm not EE. I took a mechatronics class and it was one of the worst experiences of my life.
 

Adjuster

Joined Dec 26, 2010
2,148
In open-loop operation, (driven by a constant current or voltage) the bridge circuit provides an output which is dominantly an indication of change, standardised to zero at a particular condition. This may be more convenient than dealing with the basic sensor voltage. In particular, this method partially rejects effects of fluctuations in the "constant" current supply. The raw sensor voltage is almost directly proportional to the drive current.

A much greater freedom from concern about supply voltages is obtainable with the method where the bridge is kept balanced in a closed-loop system, as explained earlier.

Perhaps given your limited requirements for a qualitative indication of turbulence, a very simple approach will give the results you want. In the end, the only way to be sure about this is to try it.
 
Last edited:

joeyd999

Joined Jun 6, 2011
5,236
I recognize that I am beating a dead horse, but I realized in retrospect that I could make tuning the loop filter easier if I could separate the heavy load of the wire from the filter itself. See attached.

I added R7 which makes selection of reasonable values of R6 and C1 far easier.

I am pretty sure that the chosen values for the filter are adequate for the task, if not maybe a little slower than it has to be. But without knowing the thermal characteristics of the wire, it'd be impossible to determine the fastest possible filter without potentially introducing instability.

Even if OP doesn't want this, maybe someone else would like to cobble it together and tell me if it works. I don't have the time!
 

Attachments

Thread Starter

lfgrdwill

Joined Dec 9, 2011
27
Adjuster - Good info regarding the bridge.

Joeyd999 - thanks again

How much more difficult is building a constant temperature setup with a servo amplifier feeding current back into the whetstone bridge? In case my project ends up requiring this. I am familiar with the basic symbolized diagram, but i have a hunch the generic diagrams i have seen are over simplified.
 

joeyd999

Joined Jun 6, 2011
5,236
Adjuster - Good info regarding the bridge.

Joeyd999 - thanks again

How much more difficult is building a constant temperature setup with a servo amplifier feeding current back into the whetstone bridge? In case my project ends up requiring this. I am familiar with the basic symbolized diagram, but i have a hunch the generic diagrams i have seen are over simplified.
Ummm...the circuit I gave you is a servo feeding a bridge...

Actually, if you can find a DIP amp with the low Vos (in the 10 uV range) and the in/out characteristics I specified, the circuit can be built quite fast on a perf board. I could do it in an hour, if I had the desire.

Pretty much everything you need is on the schematic, except for specific part numbers for the Darlington (you will also need a heat sink for the Darlington!) and the resistors/caps. And it should plug-and-play, assuming the numbers you gave me were correct. (Don't forget to add a 0.1uF cap from +9V to ground near the amp!).

You will also need a 9V 1A power supply. Don't use over 9V or you'll be dissipating more power than I designed for!

BTW, one of the assumptions is that you wanted to operate at 250C (because that is the data point where you gave me an actual resistance). To save power, you could run it at, say 100C. It is easy to measure the resistance at that point: just get a good ohm meter, apply it across your wire, and submerge it in boiling water (the wire, not the ohm meter! :)).

If you want to do this, let me know (and tell me the 100C resistance), and I will recalculate the values for you. This will probably give you better than 10:1 power range which may not be enough at 110MPH.

IMPORTANT: I have not tested this circuit with your wire. If it doesn't work, don't blame me, please!
 

GetDeviceInfo

Joined Jun 7, 2009
2,192
Question;

Wouldn't implementing a MAS sensor from an automobile be considerable.

You mentioned 'turbulent' air flow. Can you expand on that. In the old days, we measured pressure drop across an orifice, but turbulence was your enemy.
 
Last edited:

joeyd999

Joined Jun 6, 2011
5,236
Question;

Wouldn't implementing a MAS sensor from an automobile be considerable.
Don't think so...he is looking for turbulent conditions at various locations in the air flow (I think). Therefore, the sensor must be as small and unobtrusive as possible (so the sensor does not contribute its own distorting effects on the airflow). I think he also wants to pin-point the locations of turbulence.

A MAS is going to be huge for this app. In addition, the response time is going to be slower than I think he wants (he mentioned 300Hz).

In the end, the circuit I provided is a (custom) MAS, without the temperature compensation*, using his wire probe, and could achieve 300Hz bandwidth with the right loop filter component selection.

*He will also discover it makes a *great* thermometer, albeit overly complicated and high-power!
 

Thread Starter

lfgrdwill

Joined Dec 9, 2011
27
Don't think so...he is looking for turbulent conditions at various locations in the air flow (I think). Therefore, the sensor must be as small and unobtrusive as possible (so the sensor does not contribute its own distorting effects on the airflow). I think he also wants to pin-point the locations of turbulence.
Yes! This is exactly the goal.

I got carried away being home for the holidays. Thank you for all your suggestions.

Right now i have solved for the necessary current to heat a 3.5 ohm wire to 250 degrees at 5 m/s. As the flow speeds up to 50 m/s, the temperature decreases to about 100 deg C. With a lower temperature in a constant current circuit the time constant will decrease so that is good, but the sensitivity according to one of the books i have (very respectable author but i am new to this subject) will drop by a magnitude of 20.

I appreciate the "complex" circuit you drew up for me but my professor is going to want to try constant current first. I forgot to mention that we have limited power available...a couple hundred milliamp hours. The constant current setup would use less power.
 
Top