[10°C,60°C] Temperature Measurement with +-0,01°C of accuracy

A1) One solution is to review the need for such accuracy. 1/100 of a degree celsius? The specification sounds too tight to be meaningful.

A2) Unheard of. Pt100 RTDs are connected into a Wheatstone Bridge circuit, the balancing of which indicates the temperature to the rest of the circuitry. There is no need to consider the measuring device solo and disregard the rest of the equipment!

Consider the measuring device. It is a resistor. Current passing through it will dissipate heat, thereby affecting the measurement in a small way. There is an analog here with Heisenberg's Uncertainty Principle.

thank you PWSlack for your reply ,

1) the need for such accuracy is a real need expressed by some industrials. i have reached +-0,05 °C of accuracy using a pt100 and an ADDI-DATA´s msxe-3211 system. but it´s not enough . that´s why i have asked if it is possible to improve the sensor´s performances with a signal conditionner for example, and how much accuracy i will gain theorically with this. So the question is : how a signal conditionner improves the performances of an RTD and how to calculate theorically the accuracy gained usig a conditionner.

thank you

Good luck with the quest. Reprogramming the industrials seems just as difficult.

It is possible to improve the resolution of a platinum probe with signal conditioning but there will always be consequences for doing this. To explain the consequences requires an understanding of sampling theory, uncertainty, thermodynamics, and phonons. All of these aspects will change with the media to be measured and the time period that one measures.

????

... like this:

Secondary SPRT -sport -sports -sort

That worked for me here in California. Here's a screen dump of most of the first page:

Dick

I'm having trouble coming up with any industry that could need that kind of resolution.

Protein crystallography requires a very stable temperature for single crystal structure growth.

Yes but, what are the odds that any of OP's "industries" are growing crystals.

I almost put a disclaimer in about analytical work, but decided not to bother.

Here are a few pointers:

Self heating error is likely if the reference current is high and the coupling with the media is poor. Try to use a high thermal conductivity probe with a large surface in contact with the media. Reducing the current reference will reduce the self heating but also reduce the signal to noise ratio. Do it only in "quiet" environment.

The precision of your current reference is very important. Whether you use a current source or a Wheatstone bridge makes a big difference. The source is simple but accuracy really depends on the quality of the components used. The bridge is more forgiving but you should use 0.1% resistors in close proximity to balance the thermal effect.

The wire gage, its length, and the type of connection used to wire your sensor can have an effect. You are measuring a millivolt level signal. Minimise the voltage drop in the cable and connectors. Gold plating or soldering is almost a must. Use a twisted pair sheilded cable grounded at the amplifier end only.

The amplifier you use and the resistors / capacitors are very important. If your amplifier's gain varies by 1% from cold to warm, forget about 0.1% precision on your measurement.

Make sure that your A/D converter has the resolution and stability to obtain your objective. It is particularly important if your amplifier's output is in the fraction of a volt range.

If your sensor and signal processing system have a repeatable error, make a curve and subtract it from future measurements.

If the equipment is affected by thermal drift, place it in a temperature controlled oven.

Start with this and call me in the morning if it is not better...

To give you an idea of the sort of values you're dealing with, standard pt100 RTDs have a characteristic value change of 0.00385 Ω/Ω/degC.

That translates to about 0.4 Ω/degC in your range, or 0.004 Ω change for a temperature change of 0.01 °C.

With an excitation current of say 10mA, the output voltage would be about 1V, and would change by about 40μV per 0.01 °C.

You would definitely need laboratory grade sensors and measuring equipment, not suitable for an industrial environment or application.

The folks at NIST are the experts on precision temperature measurements. I suggest you contact them. If you can't find the right people to talk to I can get contact information. I suspect one of the first questions they will ask you is whether you actually require .01 deg. accuracy, or if .01 deg. resolution is adequate.

You might read up on the literature of low temperature physics and Bose Condensates where temperatures in the order of mili-Kelvin are measured. Or, call a university near you, and talk to someone in the physics department.

Click on the instrument...

I find that objecting the need for 0.01C accuracy is not the real issue.

To solve this problem, back in the 70's I made a study on temperature measurement accuracy for my Bachelor's degree thesis. I found that the 0.01 C accuracy was required for reaction kinetics studies.

My conclusion was that the best sensor was to use a Thermistor. Although it has a non linear resitance vs temperature characteristic, it can be calibrated versus high accuracy laboratory master thermometers. I agree that the accuracy and stability of the circuit is as important as the sensor accuracy. The coupling of the sensor to the process is of paramount importance. The thermowell may have a greater impact than all the other components and shall be specially designed for this application. The lead wires shall be thick and made of an alloy with very low resistance vs temperature coefficient, such as Constantan.

Thanks for the enlightening comment. What I find intensely curious, the OP's complete disengagement from the discussion.

You are welcome.

In my 1970's thesis work I found that Thermistors are best for 0.01C accuracy because these devices have the largest change of physical property vs temperature. in this case, ohms/°C.

In the late 70s and 80s precision current sources were developed in an Integrated Circuit form. Feeding a forward biasedsilicon diode by it, the voltage drop on it only depends on bandgap pysics. The voltage is precisely calculable, and - if my memory serves - declines at -26mV/C degree.

So, the OP has now 2 excellent ways to get the repeatibility he desires. I am sure, we could dig up a 3rd one too.

By the way, a fellow brought the picture of an instrument with too many digits. They look sexy, have nothing to do with absolute measurements. In tuning something, they might indicate slight variations, whose absolute value we have not the foggiest notion of.

I thank you all for your responses. They are really very useful for me. although I noticed the lack of a quantitative study that consider the sources of measurement errors for RTDs and shows by following a theoretical ( mathematical and physic ) raisonment how signal conditionning can correct these errors to obtain the desired accuracy. all that I found your answers there included, are, rather general advice. and if you can indicate me a link to a quantitative study on RTDs and how conditionning can improve their performance. It will be really appreciable.Alpha

Sorry to disappoint you.

Did you not find anything enlightening when you Googled RTD signal conditioning?

Excellent answer.

There is one additional limitation that you did forget in all of these measurement approaches and that is the thermodynamics of the set up and the collection of differential equations that this will entail. All of these probes only report their temperature and not the temperature of the media they reside in. There will always be a small but potentially significant difference between the probe and media to be thermally measured. The tiny mass of a thermocouple will make the time constant of this thermal exchange quicker however it will also make it more prone to self heating. All of these methods will also have the added complication of the wire leads themselves transferring heat into or out of the probe.

Does a thermocouple exhibit self-heating? Just asking.

Depending on how one reads the voltage produced, yes it can self heat. The problem is the bias currents of the sensing circuitry and almost never is a problem until one is working with cryogenics but it does happen.

I guess that could be a problem - but not very significant in [10°C,60°C] .

True but to get four degrees of valid significant figures, one should resolve five degrees of figures. Right now I would not casually dismiss any plausible errors.

I'm with John on the thermocouple self-heating.

Some of the smallest thermistors are smaller than any thermocouple I've seen, and their really tiny wires can't conduct much heat in or out. Their self-heating is small and fairly predictable with low sensing current.

It is possible, but going deeply into details is required. It would take too many pages to write here. Write to me directly.

CR4 ADMIN - email removed

From the CR4 Rules: Do not post phone numbers or email addresses. The CR4 Admin will delete all phone numbers and email addresses posted in threads or comments. You can share this information via the CR4 internal messaging system.