Accuracy of Temperature Sensors: Random Error or Intrinsic?

[Vulgar]

Hi,

why does a sensor, eg a temperature sensor, have an accuracy? Is it just because of a random error, or something intrinsic to the sensor itself and is it related with the resolution?

 

[Studiot]

Perhaps if you were to explain your question more fully ?

 

[S_Happens]

Your question is VERY general, and therefore hard to give any worthwhile answer. It appears that your main question may be whether there is something such as "random error" that can account for inaccuracy in ANY type of measurement. I don't know enough about http://en.wikipedia.org/wiki/Information_theory" to tell you if there is an answer, but maybe by helping to clarify your question someone else can help you

BTW- There are many different types of temperature sensors and each one of those has it's own limitations. These are really practical applications, which don't apply across the board to measurement itself. I only bring it up to tell you that even within each measured variable there are a very wide variety of ways to obtain that measurement, which means even though you say temperature you still really didn't specify anything. If you were to specify what type such as thermocouple, RTD, IR, thermometer, etc. then we could talk about practical limitations, but I'm guessing that's not what you are really asking.

 

[Vulgar]

My question is general because I guess there is a general answer, but I don't mind talking about one type of temperature sensor, eg. an RTD sensor. Say the specification of the sensor says for accuracy +- 0,01°C. What is the cause of this (in)accuracy. Is it because of a linear fit, when really the calibration curve is non linear? Or does it have to do with resolution, i.e. the smallest detectable change or can it be explained by information theory like you suggest or other things

 

[Studiot]

My question is general because I guess there is a general answer,

I'm sorry to disappoint you but there is no general answer. If you ever find one, please let us know.

There is a whole theory of errors in engineering statistics. There is a whole branch of science devoted to this subject viz metrology.

This is because errors can come from many causes. One of the important things to understand is the idea of random errors, since you mentioned them

Take a 6 inch nail as a simple example. What % are exactly 6.00 inches long and what does this mean? Why is this so?

In your sensor example the answer is the same as with the nail - manufacturing imperfections.

But as noted above these imperfections need not be random.

Every reading from your sensor may be 0.005 deg low and the sensor still within specification.
Is the low reading therefore an error?

Every nail may be 0.005" short. It may be that the machine that cuts the nails is set wrong so the error is not random.

go well

 

[Vulgar]

You mostly describe bias errors, which can be removed well with calibration i think. Random errors can be caused with unexpected things, for instance wind breezes for temperature sensors, or other things.

But say you have a sensor which always gives you a temperature of 0,5°C lower than the actual temperature, you can easily correct it by including a term +0,5°C.

But say you remove all external factors like wind breezes, and such, and there is no bias or manufacturing imperfections, you have an area with a perfect set temperature, a perfect thermodynamical equilbrium and so on, and if you put a sensor in there with an accuracy of let's say +-0,01°C, what is the cause of this uncertainty. You can never measure the true value, but what in the sensor is responsible for the 0,01°C accuracy. I like S_Happens idea that it might be a limit imposed by information theory.

 

[Studiot]

Since you have (now) specified a perfect world, information theory returns perfect answers.

S_happens, like myself, asked for more detail.

Like I said this is a huge subject which has to deal with possible sources of error in

the measurand itself,
the measurement path
the measurement method
reading and recording and data processing
repeatability