New to electronics, looking for help building the best temp sensor possible

Thread Starter

jumpkickpunch

Joined Dec 14, 2013
8
I'm starting a project where I want to use a PID controller to hold water temperature in the ~150F range +/- 15F with as high a precision as *reasonably* possible. I know it's a bold goal but I'm shooting for .05C error. I doubt I'll get there but it should set the tone for this project.

This is the beginning stages right now and since the PID is entirely dependent on the quality of the temp input data, I'm just trying to focus on creating the best temperature sensing circuit I can. Once I'm sure I have quality input, then I'll move on with the rest of the circuit.

Like I said, I'm new but from what I've gathered from this article:

http://www.bapihvac.com/content/uploads/2010/07/RTD-vs-Thermistors.pdf

given my relatively narrow temp range, a thermistor would be the best type of temp sensor. I expect to have access to some top notch lab equipment at a medical research facility so the non-linearity of the thermistor *I think* can be resolved with proper calibration. That said, in one of the links below, it was mentioned a thermistor will be out of spec in a week if not "pre-aged".


Also I haven't decided if I should be using the on board ADC of my MCU. It's been suggested to me for the sake of my tolerances I consider a seperate ADC all together to minimize interference. Thoughts?? Suggestions for possible components?

I'll need to amplify the signal somehow. Any thoughts on the best way to do this while minimizing interference?

I came across this: http://www.ti.com/lit/ds/symlink/ina330.pdf and it appears to be a good solution in addition to reducing number of components, simplifying design and thereby cost but I'm not sure since it doesn't use a bridge circuit like most of the other examples I've found online.

Long term stability and self heating are most definitely concerns that will need to be addressed. If anyone can point me to a schematic I can reference, tips on circuit design or in general-please feel free to comment.

So far I'm referencing the threads linked below but most of the discussion seems to be about which probe to use. I'm still interested in what anyone has to say regarding probe selection but I would love to hear thoughts regarding some of my other concerns as well.

http://forum.allaboutcircuits.com/showthread.php?t=54303

http://forum.allaboutcircuits.com/showthread.php?t=35509
 

wayneh

Joined Sep 9, 2010
17,496
...I'm shooting for .05C error.
Do you mean accuracy, or deadband control within 0.05°C, like 65.55±0.05°C? You can accomplish the latter with an LM35 thermometer, an op-amp and comparator.

I think you will need calibration against NIST standards to ensure that level of accuracy with any solution.
 

shteii01

Joined Feb 19, 2010
4,644
I'm starting a project where I want to use a PID controller to hold water temperature in the ~150F range +/- 15F with as high a precision as *reasonably* possible. I know it's a bold goal but I'm shooting for .05C error.
That's where I stopped reading.

Either you are doing it in °F or you are doing it in °C. Choose one.
 

tracecom

Joined Apr 16, 2010
3,944
A couple of comments. First, it's interesting that you state your anticipated temperature range in Fahrenheit, but your precision objective in Centigrade. In terms of holding water temperature at +/- .05°C, I think that depends entirely on the amount of water, the type of container, and what happens to the water once it comes to temperature. Given a substantial amount of water in a well insulated container with no sudden introductions of off-temperature materials, I think it's doable. I doubt however that you simply want to hold water at 150°; probably, you want to use it for something.

What is going to happen to water once it's at temp?
 

Thread Starter

jumpkickpunch

Joined Dec 14, 2013
8
wayneh,

I'd like to be able to hold water temp within .05C of a setpoint so I guess that means deadband control, right? In nearly every thread I read, the LM35 is being suggested. I can't tell if its because of familiarity or because it's actually a superior component. After a quick scan of googles results for silicon junction temperature sensors, I see there are also digital ones (ex LM70, LM35) Do you think these would be a better choice since I could eliminate the ADC?

tracecom,

I'm trying to focus on the temp sensor circuit right now. Once I'm confident I have quality input data, I'll begin working on the rest of the project. That information was only there for context. But to answer your question, nothing. Once at temperature, the objective is to keep it there.
 

#12

Joined Nov 30, 2010
18,224
Has anyone told you that water stratifies according to density (and that includes temperature)? That heated water mixes by convection?

Are you going to keep this under a pound of water and have a stirring motor going at all times? Or is a swimming pool size job?
 

Thread Starter

jumpkickpunch

Joined Dec 14, 2013
8
It will be relatively small. Approximately 1 gallon that will my mixed by a separate device. But again, I can't stress this enough, these are all issues to be addressed later. Right now, as the title states, I'm trying to focus on the best possible temperature sensor design for water in the ~150F range. (or 65C if you prefer metric)
 

shteii01

Joined Feb 19, 2010
4,644
It will be relatively small. Approximately 1 gallon that will my mixed by a separate device. But again, I can't stress this enough, these are all issues to be addressed later. Right now, as the title states, I'm trying to focus on the best possible temperature sensor design for water in the ~150F range. (or 65C if you prefer metric)
If I remember right, LM35 has the precision of + or - one °C. To me that means that it does not meet design requirements.
 

Thread Starter

jumpkickpunch

Joined Dec 14, 2013
8
I'd be inclined to agree but I think wayneh believes I can still achieve my goals using an LM35 with an op-amp and comparator.

What would you suggest?
 

tracecom

Joined Apr 16, 2010
3,944
I'd be inclined to agree but I think wayneh believes I can still achieve my goals using an LM35 with an op-amp and comparator.

What would you suggest?
Wayneh may have a method in mind to improve the accuracy of the LM35, or he may be planning to calibrate the circuit to the particular LM35 in use. Otherwise, an LM34 is intrinsically more accurate than an LM35 simply because one degree F is smaller than one degree C. I think that, with the relatively narrow temperature band you are trying to cover (135 to 165°C), a thermistor can be more accurate than either an LM35 or LM34. But that's just based on my experience, which is not exhaustive.
 
Last edited:

ronv

Joined Nov 12, 2008
3,770
Since your range is small it is all in the calibration. I would follow the IC data sheet recommendations. I don't think the LM35 will be stable enough.
 

#12

Joined Nov 30, 2010
18,224
Guaranteed initial accuracy is one thing. Calibrating to the "real" temperature is another. You are not going to find ANY temperature sensor with a guaranteed initial accuracy of .05C. You are going to calibrate whatever you use, so plan on it.

This fact places the LM34 or LM35 back into the available options. They are very convenient, in my opinion.
 

ronv

Joined Nov 12, 2008
3,770
The long term stability is .+/- .08 degrees C.
But calibration be for each use would fix that. But having said that I'm not sure what you could use for a calibration source.:(
 

#12

Joined Nov 30, 2010
18,224
I see it. .08C per thousand hours. That is a maximum drift promise from the manufacturer and would hold to within .05C for 26 days, and I believe that number is the sum of all factors for consideration of long term drift.

Mark this in your mind. You will need to compare it to other sensors.

As for what to calibrate this to ... You can't buy a box of temperature and connect it to your volt meter. This has me stumped.

Edit: Holy smoke! You CAN buy a box of temperature!
 

Thread Starter

jumpkickpunch

Joined Dec 14, 2013
8
I'm not certain I understand the calibration process fully but my plan was to use one of the thermometers at a medical research university near me. I thought I could bring water up to my temp range and log the readings of their thermometer with my temp sensor to make a table. Would something like that work?
 

#12

Joined Nov 30, 2010
18,224
That would only be as good as their thermometer.
The triple point of water is an indisputable constant.
 

Thread Starter

jumpkickpunch

Joined Dec 14, 2013
8
True but I thought if wanted max precision with whatever ADC be it 10 12 or 24 bit, I'd want to use all the bits within the narrowest range. Since I only need precision from about 55C-70C those would be the temps I need to use, right?
 
Top