Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

[SOLVED] How to use NTC temperature sensor in production

Status
Not open for further replies.

peeyushsigma

Member level 4
Joined
Apr 15, 2012
Messages
77
Helped
19
Reputation
38
Reaction score
17
Trophy points
1,288
Location
Bengaluru, Karnataka, India
Activity points
1,773
Hi There,

I want to use NTC temperature sensor in one of my product with MCU having 12 bit ADC..any idea how to take this into production, How we will maintain its calibration and will every sensor will require calibration if we want to achieve accuracy of -+0.5 deg C across -40 to + 40 Deg C.
 

Hi,

We don´t know which NTC you choose...
Thus my recommendation: read the datasheet, as we can´t do this.


Klaus
 

Industry standard NTC used for climate control have about +/- 0.5K tolerance in the positive temperature range and slightly more, e.g. 0.6 - 0.7 in the negative range. Additional electronic tolerances should be considered.
 

@KlausST

https://en.tdk.eu/inf/50/db/ntc/NTC_Probe_ass_M2020.pdf

is the NTC thermistor I want to use.

My question is if I use good tolerance analog components, should I be able to achieve the 0.5 deg C accuracy over (-40 to +40 deg C) range, without doing any calibration?

- - - Updated - - -

You possibly may get more error from
thermal resistance to heat source.

Have you not learned yet to do a tolerance stackup?

No I do not have any Idea..

- - - Updated - - -

You can get pretty accurate thermistors for not a lot of money. For example: https://www.littelfuse.com/~/media/...rmistors_ultra_precision_pr_datasheet.pdf.pdf

However, you also have to consider the intrinsic non-linearity of the device, as well as the tolerances of your application circuit.

If I use linearisation circuit for overcoming the non linearity of NTC in short range say 50 degree C then, can they be production ready without any calibration ?
 

If I use linearisation circuit for overcoming the non linearity of NTC in short range say 50 degree C then, can they be production ready without any calibration ?
Analog components have tolerances that in the set give a total characteristic tolerance on the circuit larger than the value of each one individually, so without analysis of the circuit the question is vague, but intuitively the answer is: No. Another fact to consider is that the nominal value of these components varies with temperature, and therefore if the whole circuit is subjected to the temperature to be measured, this will add an extra uncertainty to the circuit. In short, if you want bigger accuracy, you should consider a digital solution, either in the sensor (which often come pre-calibrated from factory) or in the conversion, as this turn more reliable the calibration result.
 
I'm afraid I disagree with my esteemed colleague andre_teprom. With proper design and analysis you should be able to determine the overall system performance. There is no "uncertainty"; there are tolerances, temperature coefficients, etc., which are specified by the manufacturer(s) and which can be used in a comprehensive analysis. Digital or not, at the fundamental level this is still an analog problem.

The OP already stated that he's using a 12-bit ADC. I would suggest you do the linearization in the microprocessor. You'll need a very accurate current source or a very accurate bridge amplifier.
 
Agree with you with regard to the inappropriate terminology, as 'uncertainty' could give the impression of unreliability, but I was rather referring to the OP's phrase "If I use linearisation circuit", as if he wanted to do the non-linearity compensation with analog circuit, which is often more complicated to calibrate in production when compared with any method done via firmware.
 
If I use linearisation circuit for overcoming the non linearity of NTC in short range say 50 degree C then, can they be production ready without any calibration ?

Of course you can linearise any range: but it comes at a price. How much inaccuracy you are going to accept?

It is designed to be used with a look up table and you can linearise the device within a small range: say 2-5 deg (and perhaps you can get <0.5 accuracy over this small range)

They are resistive devices and need calibration. If you want a device that is absolute, you may use a band-gap device. But they are voltage devices.
 
-40 to + 40 Deg C this is a very long way ... and linearity this would be a major problem.
I did several experiments with NTC temperature sensors, they are slow, even if my samples was Vishay / Philips quality parts.
My application was adding lost NTC in to rechargeable NiCD battery for cordless tools.

Safe use for NTC this is as high temperature cut-off (fuse circuit) and not as thermometer.
As lab testing gear I did made my own Thermoelectric DIY Thermistor Tester - Minus 5~110 Celsius
https://www.ittsb.eu/forum/index.php?topic=1395.0
 

Hi,

I assume there is a microcontroller...
So why don't you use a digital temperature sensor and avoid all analog problems?

Klaus
 

Can you please suggest any band gap device for the same purpose..I just want avoid calibrating all the sensor in production.

i suggest you get some basic idea from the article; **broken link removed**

One of them perhaps can meet your requirements.
 

So my question is what are you measuring? A temperature on a PCB? A chassis? A dangling temperature sensor at the end of a cable?

Your response to the previous question implies you're not measuring a PCB.

An LM35 in a thru hole package would be pretty versatile. It can be PCB mounted and glued to a thing its sensing. Or it could be manufactured (carefully) into a cable assembly for longer distance remote sensing.
https://www.digikey.com/product-detail/en/texas-instruments/LM35CAZ-NOPB/LM35CAZ-NOPB-ND/148180
 

So my question is what are you measuring?

He will never answer that, for now he is taking advantage of this forum for getting ideas supposedly cutting costs in some sort of production line, that it could be his own bedroom room.

Yesterday I was watching a video, K type thermocouple for our multimeter this cost just few dollars, quality made one and calibrated this cost 3000 ~ 5000 USD ..
Therefore there is no high accuracy sensor's in the price of few dollars.
 

K type thermocouple for our multimeter this cost just few dollars

Why so costly? The material is inexpensive and if you want to use it at temperatures above 1000C, you need some protection (the metals work well in reducing atmospheres but using in furnaces need special care).

If you are not into very high temperatures, I would suggest iron constantan - both the wires are very cheap and easily available.

Quality is not really an issue but impurities in the metals can cause changes in the EMF and one calibration (during their life) is usually enough.

But I cannot guess what thermocouple can cost 5000USD (unless they are selling to some middle east military)- Pt and Pt/Rh is apparently the most expensive of all thermocouples.

Perhaps it is not difficult to get 1C accuracy with a cheap thermocouple in the ambient temp range (0-400C) and you will not get much better with a 5000USD thermocouple.

High accuracy is relative; 0.1C is considered accurate in general measurements (say a thermostat or the temp of the junction) and 0.01C is very accurate in the lab (around the room temp).

Temp accuracy of 0.001C is tough to get, measure or maintain (around the room temp or ambient temp). Close to the absolute zero, we often hear of temp of microdegree accuracy.
 


He will never answer that, for now he is taking advantage of this forum for getting ideas supposedly cutting costs in some sort of production line, that it could be his own bedroom room.
.

Why do you think I will not answer that..can you please clarify..why one should not take advantage of this forum ?? I have been using this forum for the last 8 years and everyone helps as per his/her capability and knowledge.

- - - Updated - - -

So my question is what are you measuring? A temperature on a PCB? A chassis? A dangling temperature sensor at the end of a cable?

Your response to the previous question implies you're not measuring a PCB.

An LM35 in a thru hole package would be pretty versatile. It can be PCB mounted and glued to a thing its sensing. Or it could be manufactured (carefully) into a cable assembly for longer distance remote sensing.
https://www.digikey.com/product-detail/en/texas-instruments/LM35CAZ-NOPB/LM35CAZ-NOPB-ND/148180

I want to measure cold storage temperature and in some places ambient as well.. thats why I mentioned -40 to +40 deg C span will be sufficient. LM35 Accuracy from –55 °C to 150°C is +-1deg C.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top