I am interested in a simple non-PWM fan-temperature-controller for a server that I am modifying (the existing PWM fans scream, however I can not simply turn them down with a resistor as they must also respond to changes in server load... I am replacing 14 40mm fans with 4 80mm non-PWM fans).

I found this very simple circuit http://www.heatsink-guide.com/tempcontrol.htm



But I would like a version of it that adjusts the 12V wire rather than the ground wire so it can be used with the fan's tachometer and wondered if anyone here would be so kind as to design the circuit for me. The page actually mentions a modification of the circuit that could be used with the fan's tach using a P-Channel MOSFET instead of an N-Channel MOSFET, but I wouldn't know which modifications to make.

I am interested in other simple fan-temperature control circuits if someone knows a good one. The three component circuit above is attractive as it is so simple... I find other circuits on google that may work but are considerably more complex.

Regards, James

 

But I would like a version of it that adjusts the 12V wire rather than the ground wire so it can be used with the fan's tachometer. The page actually mentions a modification of the circuit that could be used with the fan's tach using a P-Channel MOSFET instead of an N-Channel MOSFET, but I wouldn't know which modifications to make.

To use a P-MOSFET you just reverse the plus and minus polarities.
Thus the red line would go to minus (ground) on the power supply as well as the fan's minus connection, and the black connection would go the the power supply plus connection.
The MOSFETs drain connection when then go to the fan's plus (red) connection.

Note that the P-MOSFET you select may have a different pinout so pay close attention to that.
In the diagram, the left pin is the gate, the center pin is the drain, and the right pin is the source but the P-MOSFET may be different.

Just wire the gate to the left connection, the drain to the center connection, and the source to the right connection.

 

Would I switch to a PTC thermistor or keep using a NTC thermistor?

And do you know if I would have any success substituting the R2 pot (suggested value for R2 is 5kohm) for a PTC thermistor in this diagram for an LM317?

 

Would I switch to a PTC thermistor or keep using a NTC thermistor?

You keep the NTC thermistor.

And do you know if I would have any success substituting the R2 pot (suggested value for R2 is 5kohm) for a PTC thermistor in this diagram for an LM317?

That would give a positive change in voltage with an increase in temperature, but I don't know if the rate of change would be what you want.
If you know what output voltage you want for the thermistor temperature/resistance profile then an appropriate circuit should be able to be designed.

 

Ahh thankyou very much, I may have a go with making that original circuit with a P-mosfet...

You keep the NTC thermistor.
That would give a positive change in voltage with an increase in temperature, but I don't know if the rate of change would be what you want.

You're right, the rate of change would be tiny. I am interested in the LM317 based circuit because if I'm not mistaken it would be highly efficient as it is based around a voltage regulator. I found some data for a PTC thermistor...



...and it turns out the variance between the resistance at 20 to 50 degrees celcius (the operating temperatures I am interested in, between room temperature and my hoped for upper temperature for the CPU) is very slight and would need to be amplified. Assuming a (more common) NTC thermistor (or PTC thermistor, if necessary) had a similar slight response to temperature do you know a circuit I could substitute into R2 in the LM317 application circuit that would amplify that small change in resistance and lead to a significant change in output voltage, eg 5 to 10 volts between 20 and 50 degrees celcius (input voltage would be standard ATX 12V). I am happy to do a bit of tuning.

Regards, James

 

I am interested in the LM317 based circuit because if I'm not mistaken it would be highly efficient as it is based around a voltage regulator,,,.

Sorry, but both circuits have the same efficiency, as do all linear methods of regulating voltage.
The dissipation is simply (Vin - Vout) * Iout no matter how you do it.

,,,,,do you know a circuit I could substitute into R2 in the LM317 application circuit that would amplify that small change in resistance and lead to a significant change in output voltage,,,,,

The thermistor in a bridge configuration amplified by an op amp or instrumentation circuit can provide gain to give the change you want.
The op amp output could drive a MOSFET or the LM317 provide the necessary current to the fan.

Do you have a preference for the thermistor you want to use?
It can be either NTC or PTC.

 

As I was walking to the supermarket I had an idea...

Since the response of the thermistor to temperature is very gradual we need to use the sharp response of a component to amplify it, eg a diode, perhaps a shottky diode, at its start voltage.



If we set resistor R and the thermistor to have equal resistances at 25 degrees, then at 25 degrees the LM317 will be at its reference voltage, 1.25V, as all the current will flow through the thermistor and ADJUST will be shorted to ground. When the thermistor heats to 50 degrees its resistance rises to 1.106 times its resistance at 25 degrees (assuming I use the thermistor from my previous post) and current will start to flow through the (shottky?) diode, causing the LM317 to start. We can tweak values to get the results in the range we desire eg 5-10V between 25 and 50 degrees celcius.

Or we can swap the positions of the variable resistor and thermistor, and change the thermistor to a more common NTC thermistor.

 

Note that a 1.106 change in resistor value gives only a 5% change in voltage from the junction of the two resistors so I don't see how you can get a 5V change without some added gain.

 

Actually looking at datasheets for NTC thermistors their response is much larger than for the PTC thermistor I posted a chart of

eg

from a 15ohm themistor and which shows resistances as a proportion of the resistance at 25 degrees. In both cases the resistance has approximately halved between 25 and 50 degrees.

This makes the circuit very simple, we simply use a 500ohm NTC thermistor for R1 in –



– and use a 1.8kohm resistor for R2... Vout=1.25(1+R2/R1)
At 25 degrees –
1.25(1+1800/500)=5.75V
At 50 degrees –
1.25(1+1800/250)=10.25V

Excellent.

 

Great.
Looks like you solved your problem.

Below is the LTspice simulation of your circuit for a temperature change of 25°C to 50°C based upon the right-side NTC thermistor chart you posted (-297.5Ω change for a 500Ω thermistor).
(The thermistor resistance is 500Ω at the start of the simulation and changes to 202.5Ω at 1 second.)

Note that I had to reduce the value of R1 to 1500Ω so that the maximum output voltage for the LM317 (≈10.5V with a 12V input) wasn't reached until near the maximum temperature.

 

I am back a week later with the results of my first test setup of the "LM317 variable voltage regulator + NTC thermistor + trimpot" based fan controller.

I went to the local electronics component store to buy a LM317 and 500 ohm NTC thermistor, but they had run out of 500 ohm NTC thermistors so I bought a 1k one. To compensate I adjusted the value of R2 from my previous suggestion from 1.8k to 3.07k (I used a 25k variable resistor to set what I thought would be a reasonable resistance). I started by putting the thermistor with a temperature probe into a plastic bag, and placed that into a jar of freshly boiled water, and recorded the resistance as the temperature decreased from over 60 degrees to under 30 degrees celcius (I took readings at 60, 50, 40, 30 degrees).

I then put the thermistor hooked up to the whole circuit (LM317, fan, R2) into the jar of boiled water again and recorded its voltage along with the fan RPM on my computer's fan controller https://community.oracle.com/message/14328715 as the temperature dropped again back to 30.

Unfortunately when I pulled the thermistor out of the jar the thermistor and temperature probe had been soaked... the bag had leaked. If the water shorted the thermistor leads (the temperature probe was fully sealed) it should have decreased the resistance, increasing the fan voltage/speed... this is exactly what I recorded. I thought the leak would have completely ruined the results, instead the results are only slightly skewed away from predictions. Because it took about 40 minutes for the temperature to drop from 60 to 30 degrees I won't run the experiment again lol

It worked, basically. One would use a trimpot for R2 and adjust it to achieve the desired fan/temperature response (while acknowledging that the fan can not run at 12V with the LM317 in the circuit). Note: the LM317 got pretty warm, if you wanted to run multiple fans off one LM317 you might want to bolt a slice of aluminium to it or position it near a fan.



Here are pictures of the samples -

Thermistor resistance (multimeter) vs temperature (red):



LM317 voltage (multimeter) + fan rpm (fan controller) vs temperature (red):

 

So I was wondering about my soaked thermistor and what would happen if I used a lower value for R2 (about 2k), so I reran the experiment, first doing a run to check the value of the thermistor at each temperature, then checking the output voltage and fan RPM at each temperature.

Yesterday when I did the test the output voltage didn't budge from 9.89V until it hit 48 degrees, today it was stuck at 9.7V till about 60 degrees.

I added (to the attached image) readings I took yesterday at 70 degrees and 25 degrees, and the value at 13.4 degrees that I took today before I dunked the thermistor and temperature probe in the jar of boiled water (it's pretty cold here in New Zealand at the moment... the voltage at 13.4 degrees wasn't high enough to start the fan spinning).

The measured voltages from yesterday were higher than the values calculated, in line with a slightly shorted thermistor during the voltage measuring run, today they were mostly fractionally lower than calculated... the thermistor resistances yesterday and today were similar.

The fan is a Titan 120mm standard computer fan.

 

Note: I didn't use any smoothing capacitors (voltage regulation for a fan is probably not that important).

 

The LM317 can oscillate if it doesn't have a capacitor (100nF ceramic) connected near its input pin to ground.
It doesn't need an output capacitor.