Thermal controlled fans

Since many years I am fan of quiet computing, but I also like power computers and videocards because I like do play games from time to time. That's why I have tried number of different solutions to control the fans in my computer.


Couple of years ago I stumbled across a two schematics of temperature controlled fan in one russian site. I tried both and they worked fine, but of course they are not very efficient and can control only one or two fans.
Another solution that I designed was a schematic that connect the fans to the 5V rail and switched them to the 12V rail when temperature crosses certain limit.
Lately  there is a gain in popularity of PWM controlled fans. I have tried that too, but I am not satisfied because usually PWM controlled fans produce more noise.
So here is my solution: DC-DC converter with temperature controlled output voltage. As a base, I used the first schematic from the aforementioned russian site. In my boxes I found one LM2575 in TO-220 package and one LM2595 in TO-263 package, and I made two pieces with both. Here is the first variant I made:

The first comparator is connected as non-inverted amplifier and amplify the voltage difference so the output voltage changes between almost 0V to about 12V as temperature changes form 25°C to 80-90°C. Zener diode D3 prevents the voltage at the negative input of the second comparator to drop below 4.5V. If the connected fans cannot start at 4.5V, then D3 can be replaced with lower voltage zener diode. The minimum voltage is calculated with the formula: Umin = 12V - Uzener. For example with 6.8V zener diode the minimum voltage will be 5.2V
The second comparator control the feedback input of the  DC-DC convertor so that the output voltage is equal to that in its negative input.The resulting output voltage varied between 4.5V and 11V. It cannot reach 12V because the saturation voltage of the DC-DC converter is about 1V.

There can be connected as much as 5 fans and the current can be up to 1A.
The transistor used for thermal sensing can be PNP or NPN and is connected as VBE multiplier. The case of the transistor must be electrically isolated because usually heatsinks are connected to ground.
I tested the schematics with KSE350 and KSE340. The two resistors are soldered directly to the transistor pins and are isolated with a heat shrink tubes.

Here is a little video demonstrating how it works. In the video I heat up the little heatsink with my solder iron:

After I finished the two boards I realized that the more convenient solution will be if the thermosensor is connected to the ground and the first comparator is connected as inverted amplifier:

With the collector of the thermal sensing transistor connected to the ground, there is no requirement for electrical insulation, so it can be BD140 or MJE350.

I am not tested this variant yet, but it work in simulation.
Also, you can notice that in the schematic I used LM2596. If the rest of the power elements (D2, L1) are selected properly this can control up to 3A load.

I will post the files for the project in couple of days.

P. P.  The two finished boards are for sale - if anyone is interested, the contact info is in the "about" page.