Page 1 of 1

Thermal Anemometer Design for automotive HVAC - ?

Posted: Sun Dec 09, 2007 11:49 pm
by h3rk
Hi, it's nice to finally post here, as I have spent much time on this site(N&V), running the updates every month prior to the new remodel, I love the site, and look forward to spending some time in here.

Here’s a hobby project that I’m working on, and I’m hoping that I can get some help on this forum.
The project requires an air velocity measurement; sensed in an automotive HVAC system, which by design has no portion long enough to use a differential pressure/pitot tube measurement method. The airflow was originally going to be measured at one speed and interpolated based on fan supply voltage. The system airflow resistance, however, is very dynamic (varies with temperature door position, and mode). Also, some other variables in the control scheme require using an accurate air flow rate to deduce some heat transfer coefficients, used in an overall heat balance control method. Taking all of this into account, I’ve decided measured air velocity is required(fan curves for these blowers are virtually non-existant).
Looking at the normally used anemometers, I have narrowed my choices to turbine, and heated wire/thermistor (or equivalent). I prefer the thermistor solution, as it introduces less flow restriction to the system, has no moving parts that may require periodic replacement/ maintenance.
There are two likely places in the flow-path that this sensor will be placed, either prior to, or after the evaporator. Before the evaporator will provide a more constant air temperature (varies with cabin temperature in recirc mode or with ambient temperature), placement after the evaporator will introduce large, rapid temperature changes as the compressor clutch engages and disengages, but will provide the added functionality of sensing temperature in a location that is not currently measured, and would be a useful parameter to have. I’m not sure if the effect of the evaporator cooling the air during compressor operation will degrade the airflow indication. Will it?

This is a shared project that involves an analog input/digital input and output device, connected via usb to a computer. Control is done with the computer software, which means:

Sensor output has to be 0Vdc-5Vdc.
The output of the anemometer does not need to be calibrated for cross sectional flow area and temperature, this will be done by adjusting constants in the software (part of an overall project directive to allow for use in many types of vehicles). Velocity is all that is required.
Expected air flow rates are somewhat unknown, but the suspected range is 0-500CFM. Cross sectional area range is expected to be 20 insq. to 200 insq.

Questions I do not yet know the answer to:

What exactly is the signal output of the thermal anemometer most proportional to? Mass flow or velocity past the heated sensing thermistor?
Do I need to adjust the output to get velocity, so that it can be used to obtain total airflow as mentioned above?

I’ve seen the circuit referenced at this site: ... cleID=1978
-How are R6 and R7 ‘thermally coupled’?
-Is this PIC design overkill, considering the signal can be adjusted in software?
-Would the standard hot-wire method be better in this case? Where can I find an example?

Would ‘hacking’ an automotive MAF found on ebay (as long as it’s not a vortex pulse counter type) yield acceptable results?
What do I need to do to obtain linear output of the device? Noting that automotive MAF sensors, using the same technology have curves similar to this:

Am I not asking other important questions?
Any help would be much appreciated, I’m sorry for asking such complicated questions about such a simple concept, I just want to be sure I’m doing things right.
How would you do it?

Also with respect to placing the thermistors in a tube? Do they both go in the tube, where only one is exposed (with holes) to direct airflow?

Is there a good schematic for a simple one of these somewhere?

Posted: Mon Dec 10, 2007 12:33 pm
by Externet
A modern mass air flow sensor from a car wreckyard puts out a varying frequency proportional to the flow. Diameter can be up to ~4". Installing the thermistors in a larger duct would alter data and readings would have to be compensated.
An older (70's) sensor from at least Mercedes has a moving gate and puts out a voltage (or resistance) from its rheosthat.
Both are from an inexpensive source to experiment with.

Try what is best for your application. Good luck,

Posted: Mon Dec 10, 2007 1:29 pm
by haklesup
I am no expert on air flow measurement but I can point you to a reletively good reference. Check out the Omega website and search on air flow. They have many instruments and more importantly, they have lots of spec sheets, manuals and application notes to help guide your engineering choices.

Review of the manual for this now discontinued model may be useful to filling in your knowledge gap

What I really don't know is how well this method reacts to rapid changes in temp and humidity of th airstream.

As for a thermally coupled resistor; ideally the two resistors would be in the same package (3 or 4 pin SIP) but 51/1K is probably not going to show up even on Vishay. To make this, simply place both resistors in a piece of shrink tubing with a bit of heat sink greese and you will have them thermally coupled. Make sure the leads don't short of course.

If the sensor is self powered, the output voltage range is in the spec (And prabably less than 5V). If it is vehicle powered then it is likely in the 0V to 12V range. In either case, level shifting can usually be done with an op amp or a resistor divider (shift down only). A single wire output WRT chassis ground has sufficient noise margin at 12V and analog output but if it is a 5V digital signal, you may want to use a differential signal or isolate it using a 1:1 isolation transformer at each end. Watch out for ground noise similar to what you get when you use an amplifier and head unit without isolation and chassis ground.

Posted: Mon Dec 10, 2007 4:24 pm
by h3rk
Thanks for the advice!!!


I just pruchased a 20$ ford MAF on ebay, I'm aware of that voltage-freq converter possibility, and I'm hoping I can bypass it somehow if it exists in this one. It's cheap so if not, no biggie, I'll move on to your suggested models.

Yeah I stumbled on to Omega in my googling, I found a lot of guidance there, but not as much as I found in that PDF you linked to, thanks.
One thing I found interesting was the non-linearity of the RTD output, and how they said it is good for allowing for a wide range.

How is this accomplished? Do they use 2 (or more) slope constants that shift based on where in the range the reading lies? is that why some instruments have range switches: To put into use slopes that approximate only that span to which it best fits? I'm probably way out to lunch.

Yes the sensor connection I have to work with on the IO is 5V 3-wire.

Thanks for the information on components, methods to extract the signal, and the heads up on the noise, a lot to digest - exactly what I was hoping for.

Posted: Tue Dec 11, 2007 1:29 pm
by terri
I broke open an 1157 (12 volt) bulb and ran a constant
2 volts across it and measured its resistance by the
current flow. More air, less temperature, lower

I had to fiddle with externals a little to get it
in the flow range I wanted.

Had I wanted to get more sophisticated, I could have
stuck it in a bridge circuit, but all I wanted to do
was get a rough relative idea of air velocity through
household heating ducts (furnace itself shut off.)

I had always wanted to calibrate it, just for the heck
of it, by sticking it up on a pole and driving around
with it. Never did.

But hey, it worked!

Posted: Tue Dec 11, 2007 4:52 pm
by Robert Reed
"Had I wanted to get more sophisticated, I could have
stuck it in a bridge circuit, but all I wanted to do
was get a rough relative idea of air velocity through
household heating ducts (furnace itself shut off.) "

That sounds interesting. How had you planned to keep the other side of the bridge (1157) out of the moving air mass.

"I had always wanted to calibrate it, just for the heck
of it, by sticking it up on a pole and driving around
with it. Never did."

I did that years ago to calibrate an annemometer. Had my wife drive while I recorded data. Really woked neat, but Oohh - the weird looks I got from passing drivers :grin:

Posted: Tue Dec 11, 2007 9:32 pm
by terri
Robert Reed queried:

"That sounds interesting. How had you planned to keep
the other side of the bridge (1157) out of the moving
air mass."

Just by wires, with the remainder of the bridge out
of the air flow. The additional resistance of the
lead-out wires could be balanced out by adjusting one
of the birdge arm's resistance. Standard Wheatstone
Bridge-balancing technique. Obviously, I did not get
that far, the problem having been solved with just
the one 1157 bulb's filament changing its resistance
with the air flow. (At the time, all my e-stuff except
for my ham tranceiver was in storage. I had just moved
into the place after a divorce. In February. The
previous owner had remarked about the poor heat
distribution in the place.)

It was a real quick-and-dirty alternative to just
"feeling" the air coming out of each duct, or trying
to read a piece of tissue paper taped to the ducts.

Obivously, I kept the furnace itself off to avoid
confounding heat loss in the 1157 filament with varying
temperatures coming out of the ducts.

(I have to backtrack on the 2 volts. I used a
slightly-used "D" alkaline, as I recall now,
twenty years later, and a 10-ohm 5W resistor to
limit the current to the 1157 bulb.)

Addendum re 1157 bulb

Posted: Tue Dec 11, 2007 9:48 pm
by terri
I just realized that you may have thought that an
"1157" may have been some kind of integrated bridge

No. It was just a standard 1157 automotive lamp with the glass broken away.

Posted: Wed Dec 12, 2007 7:37 am
by MrAl
Hi there,

One way to measure air speed is to use two temperature measuring
devices, heating one of them to some fixed level above ambient
temperature, and measuring the current (or power) required to
keep the second unit the same fixed number of degrees above
the ambient. For example, with a 30 deg C ambient the first unit
temp reading will be 30 and the second might be 31, but say we
want a 10 degree difference, so we pump more current through
unit #2 until it's temperature goes up to 40 deg C. We then measure
the current it took to do this and probably calculate power and go
from there. In still air it will take less current to heat #2 up to
10 degrees above the ambient than when the air is moving, because
the moving air will take more heat away with it.
Of course there will be some settling time and it will most likely
have to be calibrated. The key idea though is to keep sensor #2
as close as possible to the required temperature above ambient at
all times (like 10 deg C for our example).

I know this works with temperature transducers, but i have to wonder
now if this would work with ordinary thermistors. Would be interesting
to try. One thermistor would be kept out of the air flow while the other
right in the middle of all the action, and heated to maybe 10 deg C
above the other using a feedback system to maintain that 10 deg C
always. Thermistors self-heat with increasing current anyway,
so i guess it would just be a matter of measuring their resistances
using a uC chip to provide feedback and measuring the current
required to keep the 10 deg C rise in the second unit (uC again)
and providing a conversion chart in EEPROM for converting from
current to air speed.
I like the idea of using the car to calibrate the home brewed unit,
that would be cool as long as there was no wind that day. The only
other way i guess would be to have an already calibrated air speed
indicator and a good blower fan and compare readings and produce
the lookup chart that way.

Posted: Wed Dec 12, 2007 12:44 pm
by h3rk
The two element description seems right on with what I'm reading about certain MAF sensors, especially toyota and early 90's ford sensors.


I've ordered one much like this one

I converted the kg/h to SCFM, and also further determined velocity by removing dividing by the circular area of the MAF tube opening. Then I applied the the velocity to the range of expected cross sectional areas that it will be used in, to see if I will have a good useable range of voltage. I'm getting about 2/3 of the range, if I put it right at the outlet of the fan, where the cross sectional area is about 18 insq.

All of this I did in excel, where I made a graph and overlayed a 3rd order polynomial trendline with origin through zero, on a XY graph of velocity versus voltage, to aid with the software interpretation. I used velocity, so that area can be a user variable.

I plan to test this with a handheld anemometer and a fan.

If this doesn't work, I'm going to try the bulb idea out; and then I'll work on a 'from scratch' design that I can set to use the full 5 volts from 0-500 SCFM.

Posted: Thu Dec 13, 2007 6:42 am
by MrAl
Hi again,

Actually in my last post i was saying that one thermistor would be
kept out of the air flow but really both thermistors would be placed
in the air flow, just one of them heated and the other not heated.

The thing about resistive type elements is that you really need two
of them to get any decent results with air streams that can vary
in temperature. The resistance can easily vary by 50 percent
with change of temperature only, and this would give a grossly
incorrect reading about the air speed.

The thing about bulb filaments is that they are not that rugged either.
I am wondering if there would be a better choice for a set of roll your own
sensor elements.