In this circuit, the 5V and 12V
supply voltages are used as
references, so the temperature control
will vary with supply voltage. More
precise control would require more
precise voltage references.
U1B sets the minimum fan
voltage at 5V. If you use a fan with
a higher minimum voltage, you would
need to apply that voltage to pin
6 through a voltage divider. For
example, a fan that needs at least 7V
to operate would make the equations
look like this:
12V x R1 ( 7 - 5V) x R2
R1 + 7K R2 + 10K
12V x R1 ( 12 - 5V) x R2
R1 + 4K R2 + 10K
■ FIGURE 2. The thermistor for this
project was a bead-type thermistor
glued into the barrel of a ring terminal.
Conductive epoxy is the best way to do
this. You can also get thermistors already
attached to tabs, drilled for mounting.
If you solve these (see “Working
the Equations”), you get R1 = 6.25K
and R2 = 4.95K. These values would
produce 7V to the fan at 40°C, and
12V to the fan at 60°C. In some cases,
when you solve equations like this,
you get a negative resistance value.
In those cases, you have to change
a resistor or reference voltage in the
circuit to get the result you want. The
exception is if you get a quadratic
equation (like you will if you
work through this example) with
one positive and one negative root.
Then the positive root is the resistance
To complete the 7V conversion,
U1B would be set to produce a 7V
output by making R13 equal to 2.5K.
Most thermistors have very low
mass, so if you change component
values, you want to be careful with R1.
If the value of R1 is too small, the
thermistor will dissipate enough
power to heat itself and affect the
sensed temperature. The maximum
power dissipated by the thermistor in
this circuit is about eight milliwatts.
The thermistor can be used to
measure air temperature inside the
enclosure, but you typically want to
measure the temperature of an IC,
heatsink, or other component.
In this case, I used a small bead-type thermistor, glued into the
barrel of a ring terminal (Figure
2). The ring terminal was then
■ FIGURE 3. The thermistor has
an approximately logarithmic
This means that the fan speed
will increase faster near lower
temperatures, and increase
more slowly as the temperature
attached to the hot component with a
machine screw and nut. This would
work with a component such as a
hard drive, or a part with a heatsink. If
the part is, say, a TO-220 transistor,
you could attach the thermistor with
the same screw that attaches the
transistor to its heatsink.
If you are measuring the temperature of an IC, you would want to glue
the thermistor to the IC package or to
a heatsink that is then glued to the
package. Thermally conductive epoxies for mounting thermistors are available from companies such as Omega
www.omega.com). Various manufacturers, including RTI Electronics (www.
rtie.com) and Vishay (
com) make thermistors already mounted to metal tabs with mounting holes.
Testing is fairly simple: Connect
+12V and +5V, and (optionally) connect a fan. The output voltage should
be about 5V. Use a heat gun or hair
dryer to heat the thermistor; the fan
speed and voltage should increase as
the temperature goes up. To test the
fault output, connect a 10K pullup
from the fault output to +5V; the fault
output should go low around 70°C.
The control circuit does not
attempt to maintain a constant
temperature. Instead, the temperature
is allowed to float. For a given ambient
temperature and circuit power dissipation, the temperature will tend to
stabilize at some value. However,
given the significant thermal mass in
this enclosure (and associated delay
between changing fan speed and
resulting temperature change),
maintaining a constant temperature
would be difficult, anyway.
Although maintaining a fixed temperature would seem more “normal”
for a control system, in this case, it
would result in higher fan noise. At
higher ambient temperatures, or
under increased load, a constant-temperature circuit would run the fan