Explicit airflow detection is essential in many applications. High power-density electronics are liable to overheat and self-destruct when cooling-fan failures go unnoticed. Heating and air-conditioning systems often incorporate multipoint monitoring of ventilation-duct flow. Clean-room air-handling systems with undetected dirty, blocked air filters can ruin process yield. Laboratory fume hoods can contain volatile solvents or toxic reagents, making adequate air turn-over critical to safety.
In these and similar scenarios, the consequences of undetected airflow interruption can range from the merely expensive to the frankly dangerous. Therefore, it becomes necessary to use some reliable means for airflow detection. Usually, either a mechanical pressure-actuated vane switch or one of the various types of heat-transfer-based airflow sensors is employed.
An advantage of thermal sensors is that they contain no moving parts. But they often require several watts of heating input to run hot enough to overcome ambient temperature variations. The detector described here is a power-thrifty member of the thermal genre. It employs an ambient-compensated airflow-detection scheme based on differential heating of a series-connected transistor pair (Fig. 1).
In operation, 200-mV reference regulator A1 maintains a constant Q1/Q2 current drive equal to 40 mA (i.e., 200 mV/R1). Since the two transistors pass the same current, their relative power dissipations are determined solely by their respective VCE voltages. For the circuit constants shown, these power levels work out to 4 V × 40 mA = 160 mW for Q1 and 0.75 V × 40 mA = 30 mW for Q2. The 130-mW heat-flow difference leads to a temperature difference determined by the heat-dissipation-versus-airspeed characteristics of the 2N4401's plastic TO-92 package. The TO-92's thermal-impedance-versus-airspeed characteristic is well approximated by the simple equation shown in Figure 2:
ZT = ZJ + 1/(SC + KT√AF )
where:
ZT = "total immersion" junction-to-case thermal impedance = 44°C/W
SC = still-air case-to-ambient conductivity = 6.4 mW/°C
KT = "King's Law" thermal diffusion constant = 750 µW/°C−√fpm
AF = airspeed in ft./min.
Therefore, the Q1/Q2 temperature differential ranges from 130 mW × 200°C/W = 26°C at 0 fpm (zero flow), to 130 mW × 75°C/W = 10°C at 1200 fpm (the 14-mph breeze found at the output face of a typical 100-cfm cooling fan). This flow-dependent temperature differential gives rise to a flow-dependent VBE differential via the 2N4401's typical-transistor VBE temperature coefficient of −2 mV/°C. Comparator A2 matches this Q1VBE/Q2VBE ratio to R2/R3. Under high airflow, Q1 is cool and Q1VBE/Q2VBE > R2/R3, which makes A2's output high (i.e., "flow OK"). With a stagnant airflow (as might connote fan failure, flue fouling, or filter fill-up) Q1 is allowed to heat up, driving Q1VBE/Q2VBE < R2/R3. This causes A2's output to slew low, asserting the low-flow fault-alarm condition.
For these circuit constants, the no-flow alarm threshold is 100 fpm (Fig. 1, again). But this "line in the sand" can be easily adjusted. Raising Q1's power dissipation by boosting collector current increases the threshold. Setting R1 = 4 Ω, for instance, would bump Q1's power input to 200 mW and quadruple the low-airspeed setpoint to 400 fpm. Increasing R1 allows the setpoint to be moved the other way (toward a lower flow level). For example, R1 = 6.4 Ω would cool Q1 to a tepid 125 mW and thereby quarter the no-go flow criteria to 25 fpm.
Besides being adaptable to different flow rates, the circuit also can accept different supply voltages. In these cases, R1 must be multiplied by (V+ − 1)/4 to hold Q1's IC × VC heating level constant. Note that a different supply rail also will change the minimum voltage rating needed for A1/A2. LM10Ls are rated for 7 V maximum, while LT1635s are rated for as high as 14 V. The LM10 can tolerate up to 45 V.