By: Brian S.       5/6/03

Keeping Things Cool

Another requirement for a fan controller is a quick reaction time to ensure that important things stay cool. To test the speed, I used a portable and mobile heat source (my finger) to quickly heat the thermistor.  With this advanced methodology, I was able to turn the fan on and off in about one quarter of a second.  Is this good enough to keep my system cool?

A better question is how well you install the fan controller.  Since the fan controller only sees the temperature of the thermistor, you must make sure the thermistor is thermally coupled to the temperature of your cooled component(s). In my case, it is the CPU. Wait!  What about the fact that I attached the thermistor to the heater core? The heater core is much cooler than my CPU, right? pHaestus noted that CPU and water temperatures typically differ by 15°C and the heater core will be cooler than the water.

Mathematically, this can be described by signal attenuation and how it relates to SNR (signal to noise ratio). Attenuation is the how much a signal is weakened.  So, a 20°C temperature swing might attenuate to a 5°C swing. The SNR is a way to measure how easily the real information can be obscured by noise.  If the thermistor is only repeatable to +/-2°C, then it won't interfere with a signal that goes from 30°C to 50°C.  The SNR would be 10 (signal range divided by error term).  Any random error is only 10% (error term of two divided by signal range of twenty).

What happens when you are measuring the water temperature which might only range from 25°C to 35°C?  (Please remember that my temperature ranges are vague guesses based upon a 23°C air temp and a 15°C temp drop at peak temperatures.) With the second situation, the signal is attenuated by 50% because the temperature only swings by 10°C, but the error term stays the same.  This means that the SNR is half and the error of +/2°C is now 20% instead of 10%.  It gets worse as the temperature sensor is moved farther from the CPU.

Thankfully, the main body of my heater core temperature has enough temperature swing for my setup to work.  If the thermistor was attached to a fin on the heater core, the temperature swing would probably not be good enough.  Worse, someone could leave the thermistor hanging out in the exhaust air to see how warm the air was.  I doubt that would work well enough, especially if the fans were not running.

The idea is really common sense, but it will bite you if you forget.

 Other things to consider are whether the thermistor is affected by any other fans.  For example, if a northbridge fan is blowing hot air right at the thermistor, the sensor may stay hot even when the CPU is idle.  For me, it wasn't a problem since my heater core is in a separate compartment within my case (similar to a YY cube). The only airflow in that compartment is the fans blowing straight through my heater core.

the size of your fan(s), as many motherboards seem to burn out with more than 2W of power being drawn from their fan headers.

Fan Noise

The noise evaluation will be fairly short for this product because I don't have the equipment to do any dB measurements.  However, I will say that it reduced the noise of my computer by what I would call half to three-quarters.  Since my computer is water cooled with a large heater core, I don't need any high velocity airflow to keep the CPU cool.  I let surface area do the work rather than velocity.

Because of my low velocity requirements, I was able to keep my CPU at a 40C temperature (still lower than stock air cooling!), and I was barely able to hear the two 120mm Nidec BetaSL fans on my heater core.  In fact, I am now annoyed by the PSU fan, which I could not hear over my previous fan configuration (full speed fans or my clicking PWM fan controller). Thankfully, my wife is also much happier.  She uses the computer next to mine and no longer complains when I turn mine on. Thumbs up.

Cost

The current cost is $39 plus shipping. Is it worth it?  Considering that the parts on the board are worth about $20, you might think "No". However, this does not include the cost for a decent looking PCB, nor does it include labor or cost of tools.

You can get a PCB like Aardil's first circuit by using Radio Shack's kit (see part number 276-1576 if that link does not work) for about $15 but that means it won't be silver plated for corrosion resistance and if it is your first board it will probably look pitted and irregular. Most of us end up using perfboard with jumper wires and end up looking much more ghetto than sleek (see this one by Macroman). Aardil is using a professional PCB company to make the boards, and that costs about $150-$200 just for setup (each board is a smaller extra amount).  For a single PCB board, you can find PCB prototyping companies that will make a double sided board for about $40, but that does not include the protective silver plating or the plated throughs.

As for tools, I spent $100 for my soldering iron, but one can be had for $10 from Radio Shack (part number 64-2055), and solder costs about $4 for a small quantity.  That is the bare minimum, because you can get waldos (to hold your PCB and wires while you use two hands for the solder and soldering iron), side cutters, desolder wick, and countless other handy items that quickly empty your pocketbook.

None of this includes the design time that Aardil and his friend put into the board. Are they making a profit?  If they are, it is slim.

If you only purchase the exact minimum components, have all the tools, and use perfboard construction, you might be able beat Aardil's price by $5. If you already have an existing fan controller, you might be able to modify it to something like Aardil's controller for less money as well.  But does it make sense to make your own? I don't think so.  I certainly spent far more money that $39 to make my own and it still didn't work as well as Aardil's unit.  I'm happy enjoying the near silence.

Conclusion

The end result is that we see a product which simply works well. There are other products that do more (such as the Crystal Fontz CF633 having a display LCD and multiple channels of fan control), there are products that do less (such as the simple rheobus controllers), and there are products that are less safe (such as those without fuses). In my mind, this is the best combination of price and features out of that bunch.

That it is not to say that Aardil has made a perfect product.  The perfect product costs nothing, does everything, and modifies itself when your mood changes.  I haven't seen one yet! Even Aardil's prototype works smoothly and safely out of the box and it looks prettier than home-made circuits.  It is not as pretty as the Vantec PWM fan controller, but it is cheaper and it includes a safety fuse. Finally, it does not have any LCD's, computer control or multiple channels, but it is much cheaper.  Now, if Aardil could add some of those extra features while keeping the price down, great!  But, it looks like a good value just as it stands.

If you have any comments or Questions please contact Brian@ProCooling.com