Hotwire Airflow Meter

In my study of automotive engine management technology I've always been fascinated with the implementation of various sensors and actuators typically used in modern cars. Some of these devices have very stringent operating conditions and come with accordingly high price tags!

A most important sensor technology is the airflow meter, as seen in many modern cars with fuel injection since the late eighties. These devices have come a long way since electronic fuel injection was getting off to a start and for that I'm glad; some of the old airflow meters are crude and fussy to maintain. Today we have airflow sensors with unprecented reliability that even have the capability to clean themselves during normal operation.

Early airflow meters used a flap or vane in the path of intake air which was deflected by the air moving by. The deflection would typically be converted to an electronic signal by the attachment of a precise potentiometer to the shaft of the vane.

The modern variant has no moving parts and offers very little restriction; this is done by using a wire placed in the path of the air flow. The wire is heated until its resistance reaches a pre-defined value. By measuring how much heating is required one can establish how much air is moving past the wire as it will cool it down.

Out of curiosity I wanted to detemine if I could make such a device and so I set about doing some research. It turns out that it's really quite simple: place the wire in a Wheatstone bridge which is then placed inside a feedback loop which maintains the bridge's balance. One might deduce that if the resistance of the wire were to change, the circuit could automatically change the current through the bridge to heat or cool the wire and thus return its resistance to a value matching that of a fixed resistor on the other side of the bridge.

It might seem quite strange at first to measure the resistance of a wire and change it by directly changing current through the wire all at the same time. Once you think about it though, it's quite straight-forward and works beautifully.

If you take a look at the example circuit below you'll see what I've done.

The "hotwire" in my circuit is a No.44 lamp with the glass removed. It can be tricky to do this without destroying the filament but it's certainly possible if you take your time.

The LF411 op-amp determines whether the bridge is balanced and makes the TIP31 apply changes in current to maintain said balance. The reason this works is because the resistance of the "hotwire" depends on how hot it is and temperature is related to power input. By changing the power through the wire we indirectly affect its resistance (there is a time lag because of the thermal mass of the wire).

In the most simple terms: the circuit is trying to force the hotwire to have a resistance of 15 ohms as defined by the resistor on the left leg of the bridge. If this resistor were substituted for a different value, the circuit would then maintain the filament at that resistance if it could.

A few more deductions can be made: the 15 ohm resistor always dissipates the same amount of power as the hot wire (this is a problem if you want a really big hot wire). Increasing the value of the 15 ohm resistor will increase the operating temperature of the hot wire. Of course, these changes only occur within the ability of the circuit to compensate for them. If you put a 1k resistor in place of the 15 ohm, the circuit would try to make the wire hot enough to have a resistance of 1k. Unfortunately, this will burn out a No.44 bulb's filament and the circuit would run out of voltage trying to do so.

The 2N3904 and meter on the right side of the schematic are there to provide a visual indication of how much voltage is being dropped across the sensing bridge. As more air flows by the hot wire, the meter deflects further. This output of course could be used to provide a signal for a microcontroller, an analog circuit, or anything you desire. The meter can be calibrated to 0g/s (zero grams per second air mass flow) by setting it to zero when there is no airflow past the sensor. As indicated in my schematic, the 15 ohm resistor can be trimmed for zero-setting but the amount of trimming required should be very small.

I should talk about 7662 IC I've used: you don't see these very often but they are quite useful. This is known as a charge-pump converter and is used for generating a negative supply rail when only a positive one is available. The 7662 charges the 10µF capacitor  to whatever voltage is available (5V in this case) and then switches said capacitor into the negative rail. It switches this capacitor back and forth very quickly so that it can store up energy each cycle which is used to provide energy for the negative voltage rail. I chose the 7662 simply because I have some in my parts collection and I wanted to have a split power supply for the LF411 while only using an external 12V positive supply.

The mounting of the "hot wire" is critical to the performance of this airflow sensor and I by no means purport to have anything near as good as what an automotive fuel injection system uses. Laminar (non-turbulent) airflow over the filament for all operating conditions is absolutely essential as any air that passes the filament must only pass it once. If the air were turbulent and passed around the filament multiple times or stagnated around it, the measurement would not faithfully indicate how much air was passing through the sensor tube. In my test setup I used putty to hold the No.44 lamp inside the wall of a paper tube so that the filament is in the path of air flow through the tube while the leads for the lamp are open on the outside.

As you can see in the picture below, the system is very simple to assemble and its uses are fairly obvious. One could use this as an anemometer for measuring windspeed, as a measure of one's breathing, as a measurement of the performance of fans or air filters, etc. Of course the original idea came from air-flow measurement for fuel injection systems but some improvements would be necessary to make my airflow meter suitably accurate for such purposes.


Previous page: RC Trigger Control