I am installing an electric radiator cooling fan on my car in place of the original mechanical fan. The electric fan will be PWM driven and require a thermister in the radiator as well a a thermister in the condenser for control. It is desirable to use an automotive type thermister which I can find with the same type NTP radiator fitting required for the radiator. I would also like to set up a third circuit to sense under hood temperature if required.

So in the end there will be three thermister circuits connected to a 3.3V Arduino. The goal would be to make those circuits somewhat universal that can be programed as required for at least a limited selection of thermisters. So far I have selected the GM radiator thermister shown below, which has a resistance of 2500Ω at 25°C and a Littlefuse clip on thermister to clip onto the outside of the condenser tube, 10,000Ω at 25°C. I have not selected a third sensor but I would expect it to be in the range of the other two.

Usually I see these connected with a compatible resistor rated at 25°C. Which in this case would be a 2500Ω resistor for the GM sensor and 10KΩ resistor for the Littlefuse. I am not seeing any reason I cant use a 10KΩ resistor for all of them and program according to the output. That would make the controller suitable for a wider range of sensors for different applications.

So my question is it reasonable to put all sensors with a 10KΩ resistor and program as required.

 

You likely could but It somewhat depends upon the thermistor resistances at the desired trip temperatures.
Do you know what those are?

 

Normally you would select the best linearization resistor(s) transfer function in a calculation with the resistance curve of the thermistor.
https://circuitcellar.com/resources/quickbits/ntc-thermistor-linearization-2/

https://users.physics.unc.edu/~sean/Phys351/techresource/data_sheets/thermistor app notes.pdf

The combination of an NTC thermistor and a paralleled resistor has an S-shaped R/T characteristic with a turning point. The best linearization is obtained by laying the turning point in the middle of the operating temperature range. The resistance of the paralleled resistor can then be calculated by the exponential approximation:

 

Normally you would select the best linearization resistor(s) transfer function in a calculation with the resistance curve of the thermistor.

For on/off fan control, operating in the best linear region is generally not a requirement.
It would be desired if he is using the third thermistor to display the underhood temperature.
It can also be linearized by using a current-source to bias the thermistor.

 

For on/off fan control, operating in the best linear region is generally not a requirement.
It would be desired if he is using the third thermistor to display the underhood temperature.
It can also be linearized by using a current-source to bias the thermistor.

I've have no idea if it's simple on/off control but the OP says PWM fan, that would seem to have some sort of scaled temperature error signal feedback. I would just compromise on an intermediate value that works fairly well on both types of sensors.

 

I've have no idea if it's simple on/off control but the OP says PWM fan, that would seem to have some sort of scaled temperature error signal feedback.

Oops, I missed that, so linearization would likely be helpful.

 

Yes this is for a PWM controller. The hotter the components the faster the fan runs. From the web I got this data on the GM sensor.


From Little Fuse. I get this

 

As far as linearization, I don't think its a requirement. The fan controls the PWM signals in discreat % duty cycles of 10, 20, 30 etc. The Arduino controller will just output those % duty cycles whenever the signal is in range. Essential it is desreate set points. However linearizing the signal would have benefits for making adjustments.

I've have no idea if it's simple on/off control but the OP says PWM fan, that would seem to have some sort of scaled temperature error signal feedback. I would just compromise on an intermediate value that works fairly well on both types of sensors.

 

As far as linearization, I don't think its a requirement. The fan controls the PWM signals in discreat % duty cycles of 10, 20, 30 etc. The Arduino controller will just output those % duty cycles whenever the signal is in range. Essential it is desreate set points. However linearizing the signal would have benefits for making adjustments.

No, it not a requirement of an error control loop but it's makes setting the temperature control points to X values easier to calculate with just scalar and offset variables from each type of temperature sensor raw data from a half-bridge.

 

No, it not a requirement of an error control loop but it's makes setting the temperature control points to X values easier to calculate with just scalar and offset variables from each type of temperature sensor raw data from a half-bridge.

Yes and I am considering putting the steinheart equasion into the code. It will make changing the temperature set points for each speed a lot easier.

 

So I have been running the numbers and have come to some interesting observations and I do have a question.

My question is that a lot of the literature uses a 10K resistor when making these connections and in fact Little Fuse data sheet highlights the 25°C line with it associated 10K resistance at 25°C. Almost as if they are implying a 10K resistor is required. But I find when I go through the data and extract the voltages that will be put out from the circuits that a lower resistance appears to give a larger and I'm thinking may be a more appropriate signal. And as this thread had stated in the beginning there is room for an unknown thermister, which may be ambient temperature or under hood temperature. That third sensor would be like the little fuse one only it may operate at colder temperatures. Base on what I see a 1K resistor would be a good universal selection.

The attached tables show the show the range of temperatures the radiator and condenser need to operate in along with the calculated Vout for there respective sensors. The first table has R1 set to 10KΩ for both the radiator and condenser. The second table shows it at 1KΩ.

Are there any pros and cons of selecting the larger or smaller Ohm resisters?

Radiator			Vin =	3.3​
			R1 Ω =	10,000
Speed %	Temp °C		R Ω	Vout
20​	85​	​	295.3214​	0.0946605​
30​	90​	​	250.4919​	0.0806423​
40​	95​	​	212.6699​	0.0687196​
50​	100​	​	180.7279​	0.0585815​
60​	105​	​	153.7254​	0.0499614​
70​	110​	​	130.8774​	0.0426316​
80​	115​	​	111.5273​	0.0363981​
90​	120​	​	95.1252​	0.0310955​
100​	125​	​	81.2104​	0.0265835​
Condenser			Vin =	3.30
			R1 Ω =	10,000
Speed %		Temp °C	R Ω	Vout
20​		40	5,326	1.1467963​
30​		45	4,368	1.0032294​
40​		50	3,602	0.8738862​
50​		55	2,986	0.7588018​
60​		60	2,488	0.6574632​
70​		65	2,083	0.5688902​
80​		70	1,752	0.4919673​
90​		75	1,480	0.4254355​
100​		80	1,255	0.3679698​

Radiator			Vin =	3.3​
			R1 Ω =	1,000
Speed %	Temp °C		R Ω	Vout
20​	85​	​	295.3214​	0.7523698​
30​	90​	​	250.4919​	0.6610385​
40​	95​	​	212.6699​	0.5787319​
50​	100​	​	180.7279​	0.5051138​
60​	105​	​	153.7254​	0.4397007​
70​	110​	​	130.8774​	0.3819118​
80​	115​	​	111.5273​	0.331112​
90​	120​	​	95.1252​	0.286646​
100​	125​	​	81.2104​	0.247865​
Condenser			Vin =	3.30
			R1 Ω =	1,000
Speed %		Temp °C	R Ω	Vout
20​		40	5,326	2.7783433​
30​		45	4,368	2.6852459​
40​		50	3,602	2.5829205​
50​		55	2,986	2.4721024​
60​		60	2,488	2.3538991​
70​		65	2,083	2.229614​
80​		70	1,752	2.1008721​
90​		75	1,480	1.9693548​
100​		80	1,255	1.8365854​

 

The GM Coolant-Temp-Sensor, and the Intake-Air-Temp-Sensors are exactly the same
as far as Resistance values go.
The IAT-Sensor has an open Element surrounded by a plastic-cage,
this allows faster response times to Temperature changes.

I no longer use direct Coolant-Temp-Sensing for my Fans, I go by Under-Hood-Temps.

Here's a Schematic that You may get some ideas from .................
.
.
.

 

The GM Coolant-Temp-Sensor, and the Intake-Air-Temp-Sensors are exactly the same
as far as Resistance values go.
The IAT-Sensor has an open Element surrounded by a plastic-cage,
this allows faster response times to Temperature changes.

I no longer use direct Coolant-Temp-Sensing for my Fans, I go by Under-Hood-Temps.

Here's a Schematic that You may get some ideas from .................
.
.
.
View attachment 259409

Thanks for the reply LowQCab

Interesting that you are using only under hood temps.

I noticed 2 things on your schematic that interest me.

1) Your sensor shows a 0.1uF cap in the sensor. Do all the GM sensors have a 0.1uF cap built it. I have a 0.1uF cap on my controller board. Is it OK to have one at each end.

2) Your scematic shows twisted pair wiring? How do I know if I need twisted pair or not?

 

The Sensors don't have Capacitors.
The Twisted-Pair-Wiring and Capacitors are simply insurance against
under-hood "Electrical-Noise", that "may" or "may-not" even exist.
Precautions prevent headaches, it's a very noisy environment.

With a PWM Controller, the Fans never stop running.
Notice that there are Pots for "Minimum-Speed" and "Minimum-Speed-with AC-Running",
and there's even an input for a special Mega-Squirt-Output that
switches the Fans On Full-Speed when the Engine is under Heavy-Load,
even before the Coolant noticeably changes Temperature.

With the plastic GM-IAT-Temp-Sensor Ty-Rapped to the frame of one of the Fans,
the Temp-Control is very accurate, and fast to respond.
.
.
.