Fan Control Circuit

Controlling two 12V fans is what this little project is about. Something a case modder would slobber on, and any self-respecting geek dad should try with his daughter, as soon as she can hold a perfboard the right way up and can refrain from putting those brightly coloured resistors in her sparsely toothed mouth.

I know. As of May 2012, there are already projects on the net along the same lines. Just google and you'll find some in the first few hits. Unfortunately they seem to be mostly incomplete, poorly documented, or just wrong. So I set off to brew my own.

An ATtiny45 monitors an NTC and starts a fan once a certain temperature is reached. Using PWM it varies the fan speed according to the temperature. When the top speed is reached, a second fan is started and its speed is changed according to the temperature, just like the first, but for the upper range.

There are two potentiometers, one to set the minimum temperature threshold for the first fan, the other one to set the temperature range. That is, how much hotter than the threshold temperature it must get before both fans is run at top speed. In other words, the fan1 is started at the temperature set by pot1 and fan2 is started at pot1+pot2. Fan1 reaches top speed at pot1+pot2, and fan2 reaches top speed at pot1+2*pot2.

If you are rather more familiar with PID controllers (who isn't? my grandma used to run a blast furnace in her backyard!) this circuit can be seen roughly as implementing a controller of type P (K_i = K_d = 0), where the first pot selects the SP and the second pot lets you choose the K_p. Similarities end there, as this circuit doesn't strive to keep the SP; anything below the SP is fine.

"Is that a fan controller in your pocket or are you just happy to see me?"

A couple of details have been taken care of, to make it interesting:

• Fans are always started at full speed for a brief period, as many fans can keep running at a voltage that is not enough to jolt them into movement.
• There is a delay on the turn-off. It's not quite an hysteresis, but in practice can effectively avoid quick and repeated on/off/on transitions, that wears nerves as much as motors.
• The first fan to start is alternated between the two, so that wear is evenly spread between the two fans. If you need to disable this feature, for instance if fans are not the same type/size, there is a knob called SWAP_FANS that you can define to zero.
• A cumulative moving average is applied to temperature readings, with the effect that fan speed variations are smoother (see CMA_WINDOW) .
• At boot time (that is, when the circuit is powered up) a brief test is performed on the fans to make sure that everything is in working order:
  1. the first fan is briefly turned on at top speed and then off
  2. the same is done for the second fan
  3. after a pause to let the fans settle, both are turned on simultaneously and then off again
  If the self-test keeps repeating or looks like it doesn't run at all, you may have a power problem; the fans draw too much current and cause a brownout of the IC.
• A reset button and a LED for each fan are there to help you debugging your circuit.

It is possible to use the circuit with just one fan, but if you don't intend to ever use the second header, you may want to get rid of one MOSFET (and related components), and you'll have a more compact board.

At the beginning of the source file there are some constants you may want to fiddle with. Especially:

• FAN[12]_MIN_DC, the minimum duty cycle, which is different for every fan
• SPINUP, the larger the fan, the longer the spin-up phase required
• MIN_TEMP, that depends on the use you'll make of this device

As you may have guessed from the picture above, my prototype employs a 7805 voltage regulator in TO-220 packaging, which is way too big for the job, and two MOSFETs that could easily drive industrial motors the size of my fist. Those were the parts I had at hand. You are free to use more reasonably sized components that would waste less room on your board.

The source code is on GitHub. You'll also need avr-gcc, avrdude and, of course, an ISP programmer. If C programming is not your cup of tea, you may try to upload a pre-compiled image I had laying around. Just make upload and good luck.

Afterthought

After struggling for a couple of weeks trying to cram this circuit onto a stripboard, you will start appreciating those little contraptions made in Chinese sweatshops for few cents and sold to you for an arm and a leg.

No, not the same arm or leg the little Chinese boy injured while making your device.