Your opinion on a Temperature-Sensing circuit I designed

Thread Starter

EngIntoHW

Joined Apr 24, 2010
143
Hey,
I designed the following circuit.

The resistors (R32, R33) and NTC thermistor have Resistance's tolerance of 1%.


I'd like to ask a few questions please.

1.
The ADC supports 12-bit resolution - does it mean that its max error would be 1/(2^12) = 0.024% ?

2.
The way which I tell the temperature is as follows:

The tolerance of B is ±3%.
T is the NTC Thermistor's temperature.
R is the thermistor's resistance, which I calculate using the ADC and the following formula:

To what extent of accuracy can I tell the temperature of the NTC thermistor, considering all the above data?

3. Would you consider this design good enough for temperature sensing?

Thank you very much :)
 

Attachments

tom66

Joined May 9, 2009
2,595
NTC thermistors are tricky. I recommend an IC like a LM35 or perhaps a more modern variant. More than likely, it will be much easier to implement. It's going to be tricky to calibrate this.

Also, the 0.024% "error" is not an error at all. 12 bits is the resolution; it does not mean the ADC is that precise, it just means that it has so many states.
 

Thread Starter

EngIntoHW

Joined Apr 24, 2010
143
Hey Tom,
Thanks.

NTC thermistors are tricky. I recommend an IC like a LM35 or perhaps a more modern variant. More than likely, it will be much easier to implement. It's going to be tricky to calibrate this.
I'd like to remain with this design currently.
Why would you say a NTC thermistor is tricky?
Moreover, could you please refer to the temperature accuracy this circuit can offer?

Also, the 0.024% "error" is not an error at all. 12 bits is the resolution; it does not mean the ADC is that precise, it just means that it has so many states.
Thanks,
I should look for its tolerance in the datasheet then.
 

tom66

Joined May 9, 2009
2,595
The ADC on modern microcontrollers has no specified tolerance, because it depends on too many factors. You should be more worried about your reference voltage tolerance, which may be ±1% or more. Errors don't necessarily add up, but assuming worst case, your temperature sensing circuit could be out by ±5% (resistor tolerance + 'B' var tolerance + ref voltage tolerance.)

Additionally, I say that NTC thermistors are tricky because the response is nonlinear. You have to have a lookup table, or use tricky math, which is sloooow on most microcontrollers.
 

Thread Starter

EngIntoHW

Joined Apr 24, 2010
143
Hi again Tom,
I really thank you!

Actually, I'm about to be interviewed for an Electrical Engineering student position, and I'd like to show some of the circuits I got to design in my current job, and this circuit is one of them.

So it'd be great to know also all the disadvantages of this circuit and of using a NTC thermistor instead of an IC.

What are the advantages of using a thermistor over an IC?
I could think of some:
- Cheaper.
- Smaller.

The ADC on modern microcontrollers has no specified tolerance, because it depends on too many factors. You should be more worried about your reference voltage tolerance, which may be ±1% or more.
Would it be reasonable to say that the ADC tolerance is the VREF tolerance?
I think that Ucs also don't specify the VREF tolerence, do they?

Additionally, I say that NTC thermistors are tricky because the response is nonlinear. You have to have a lookup table, or use tricky math, which is sloooow on most microcontrollers
You're correct.
Actually, the Uc only calculates the resistance of the Thermistor, using the above voltage divider's formula, and it sends it the PC, which calculates T.
 

tom66

Joined May 9, 2009
2,595
The advantage of NTC are it's good enough for threshold detection or a system where fan speed might be controlled. The NTC voltage divider would be more than sufficient for controlling fan speed, even if the response is nonlinear. Or you could have a system when the temperature above 40°C, the system switches off a heater. But it's not suitable for accurate measurement.
 

Thread Starter

EngIntoHW

Joined Apr 24, 2010
143
@Tom
But it's not suitable for accurate measurement.
Say that I pick an NTC with B tolerance of 1% and Resistance tolerance of 1%, and let the PC derive T (temperature) from the exponential function, would you consider it accurate?

@Marshall
There are many others on the web, some manufacturers even have their own online ones you plug some numbers in and it generates an .xls file for you.
Thanks, that is absolutely great, I wasn't aware of such published curves.
 

marshallf3

Joined Jul 26, 2010
2,358
Beats the heck out of those formulas, some of which haven't even given me the correct results.

A B Curve is a B curve and probably the most common., that's why I posted that one particular chart. You can extrapolate between values for component values not listed.

I think Vishay has a ton of charts, some you just input their part #, a starting and ending temp and how many points you want it to display and it will resolve down to something like 10 decimal points.
 
Top