EcoRenovator - View Single Post - Designing a General Purpose Heat Pump Controller

theoldwizard1 · 11-17-14, 09:53 PM

SIDEBAR on thermistors, wiring and A-to-D inputs

A thermistor is a devices whose resistance changes with temperature. The problem with them is that they loose "accuracy" at the ends of their range. If you need "accurate" temperature readings over a range of say -20F to 250F, you will need 2 or 3 different models of sensors. One to cover the low end and one to cover the high end.

But then again, you have to ask yourself, do I really need to know that it is -20F or is less tan 0F enough for what I am trying to control.

Thermistors are a good choice for embedded control because they are cheap and you can find one that will cover the most important part of your data range.

Another big benefit of thermistors (or any variable resistance device) is they can be wired as a ratiometric signal. Think of a potentiometer. The "high side is wired to the reference volatge, the wiper is the "signal" and the low side is wired to Ssignal Return" (notice I did not say ground). For a thermistor input, one side of the thermistor is wired to Vref. The other side is wired to the the A-to-D pin AND a fixed resistor mounted on the controller board.

Why all of this ? Well, do believe that 5V is 5.000V all of the time and at every location in the entire system ? Sorry, it is not. Ratiometeric systems return a value that is the ratio between Vref and Signal Return, so if Vref AT THE THERMISTOR is actually 4.95V, who cares !

Sensors that "generate" their own voltage, don't enjoy this benefit, so use ratiometric analog inputs when you can.

Additionally, if accuracy is important, a "calibration" channel should be setup on the controller board. 2 equal value 0.10% accuracy resistors in series between Vref and signal return with the center connected to a dedicated A-to-D channel off the input. In theory. this should always read 1/2 the counts of the ADC range (512 for a 10 bit ADC). If not, you can use this information to "correct" the other channels.

Then again, how accurate is accurate enough ?

Also remember, when you lower your reference voltage, electrical "noise" starts looking more and more like data ! Not good. Most scientific data acquisition systems use 10V or even +/- 10V Vref.

(Yeah, I did this stuff for awhile !)

11-17-14, 09:53 PM	#280
theoldwizard1 Apprentice EcoRenovator Join Date: Oct 2014 Location: SE MI Posts: 105 Thanks: 3 Thanked 12 Times in 9 Posts	SIDEBAR on thermistors, wiring and A-to-D inputs A thermistor is a devices whose resistance changes with temperature. The problem with them is that they loose "accuracy" at the ends of their range. If you need "accurate" temperature readings over a range of say -20F to 250F, you will need 2 or 3 different models of sensors. One to cover the low end and one to cover the high end. But then again, you have to ask yourself, do I really need to know that it is -20F or is less tan 0F enough for what I am trying to control. Thermistors are a good choice for embedded control because they are cheap and you can find one that will cover the most important part of your data range. Another big benefit of thermistors (or any variable resistance device) is they can be wired as a ratiometric signal. Think of a potentiometer. The "high side is wired to the reference volatge, the wiper is the "signal" and the low side is wired to Ssignal Return" (notice I did not say ground). For a thermistor input, one side of the thermistor is wired to Vref. The other side is wired to the the A-to-D pin AND a fixed resistor mounted on the controller board. Why all of this ? Well, do believe that 5V is 5.000V all of the time and at every location in the entire system ? Sorry, it is not. Ratiometeric systems return a value that is the ratio between Vref and Signal Return, so if Vref AT THE THERMISTOR is actually 4.95V, who cares ! Sensors that "generate" their own voltage, don't enjoy this benefit, so use ratiometric analog inputs when you can. Additionally, if accuracy is important, a "calibration" channel should be setup on the controller board. 2 equal value 0.10% accuracy resistors in series between Vref and signal return with the center connected to a dedicated A-to-D channel off the input. In theory. this should always read 1/2 the counts of the ADC range (512 for a 10 bit ADC). If not, you can use this information to "correct" the other channels. Then again, how accurate is accurate enough ? Also remember, when you lower your reference voltage, electrical "noise" starts looking more and more like data ! Not good. Most scientific data acquisition systems use 10V or even +/- 10V Vref. (Yeah, I did this stuff for awhile !) Last edited by theoldwizard1; 11-17-14 at 09:55 PM..