Pt100 sensor precision / accuracy question , batch sample selection .

[Overspeed]

Hello

I have a question regarding Pt100 as the cost vary from few Euro / USD to $$$$$$

Does the Pt100 are sorted to select the best one in a batch ( linearity ?) or does there is other requirements ? I am not an expert in variable resistor

My goal is to built some '' accurate '' probes ( I have no problem with the mechanical parts ) 4 wires to connect to my ASL thermometers but also to my HP 34970A data loger

I have equipment as a triple point water , calibration furnace , calibrator ...

I plan to sort some for temp from 0 to 100 C

Regards
OS

[EC8010]

Yes, PT100 vary from affordable to expensive. I've used them a lot. Measuring temperature is much harder than it seems because there's always a thermal resistance between the part whose temperature you want to measure and your sensor. Combined with the thermal mass of your sensor, that makes things tricky. I use the smallest PT100 I've been able to find from Farnell: 219-1840 and I glue them with a thin film of epoxy to whatever I need to measure, so they have to be regarded as disposable. I think the manufacturing tolerance is +/-1C, which is often good enough. Usually, what is required is repeatability and resolution. But it sounds as though you're capable of calibrating your probes if necessary.

You can buy all sorts of expensive probes and they all have higher thermal mass, increasing errors. They often add thermal resistance between the Device Under Test (DUT) and their PT100. I eventually considered them to be a waste of money compared to little sensors glued to the DUT. If you're measuring the temperature of a DUT that is exposed to the environment, you may want to make a little foam house over the PT100 to prevent a thermal potential divider forming from the thermal resistance from DUT to PT100 and from PT100 to environment.

You can look up the equations for determining temperature from PT100 resistance, but many 6 digit DMMs will implement them for you. All they need to know is the PT100's resistance at 0C. Adherence to the equations is assumed.

[edpalmer42]

Decent quality PT100 probes will specify the accuracy.  I've seen reference to simple +- degree C, math formulas, DIN standards, and Class designations (which may refer back to DIN - not sure).  As expected, the tighter the tolerance, the higher the cost.

Just be aware that PT100 probes are not interchangeable.  They come with different temperature coefficients (generally called 'alpha' values).  If your probe has a different alpha value from your meter, your measurements will be total garbage.  Your meter should specify the alpha value it's expecting.  A common value is 385 ppm/K aka .000385, etc.

Ed

[MK]

Another gotcha with PT100 probes is that the fine wires involved can slip at grain boundaries when shocked/dropped, so the sensor undergoes an unknown amount of change to its resistance for every shock.

[Overspeed]

Yes, PT100 vary from affordable to expensive. I've used them a lot. Measuring temperature is much harder than it seems because there's always a thermal resistance between the part whose temperature you want to measure and your sensor. Combined with the thermal mass of your sensor, that makes things tricky. I use the smallest PT100 I've been able to find from Farnell: 219-1840 and I glue them with a thin film of epoxy to whatever I need to measure, so they have to be regarded as disposable. I think the manufacturing tolerance is +/-1C, which is often good enough. Usually, what is required is repeatability and resolution. But it sounds as though you're capable of calibrating your probes if necessary.

You can buy all sorts of expensive probes and they all have higher thermal mass, increasing errors. They often add thermal resistance between the Device Under Test (DUT) and their PT100. I eventually considered them to be a waste of money compared to little sensors glued to the DUT. If you're measuring the temperature of a DUT that is exposed to the environment, you may want to make a little foam house over the PT100 to prevent a thermal potential divider forming from the thermal resistance from DUT to PT100 and from PT100 to environment.

You can look up the equations for determining temperature from PT100 resistance, but many 6 digit DMMs will implement them for you. All they need to know is the PT100's resistance at 0C. Adherence to the equations is assumed.

Hello

Thanks for your answer
I work on a differential measure system as I need a high precision delta temp measurement ( in the 0.01 Dg C ) between two parts
First problem the rigidity as a P100 is a ''foil '' and if there is strain that generate resistance variation
Connection need to be perfect as possible some supplier use Cu Ag wire

For reading / system I have no real problem I am more interested to know what procedure to use as ?:
I ''cook '' them at least at the maximum temp I plan to use , so 10 Deg C for 100 max reading and pass over Indium melting point
I ''cook '' them with a 2 or 5 mA current ?? so over the 1 mA of the measuring circuit
I install them without strain
I install them with a proven method as same soldering same wire same contacts ...
I run calibration at 0 Deg C as I have a water triple point
I run calibration by comparison at 100 Deg C ( I have no Indium triple point )
I try to trace a linearization curve
.......

Let me know
Regards
OS

[Overspeed]

Decent quality PT100 probes will specify the accuracy.  I've seen reference to simple +- degree C, math formulas, DIN standards, and Class designations (which may refer back to DIN - not sure).  As expected, the tighter the tolerance, the higher the cost.

Just be aware that PT100 probes are not interchangeable.  They come with different temperature coefficients (generally called 'alpha' values).  If your probe has a different alpha value from your meter, your measurements will be total garbage.  Your meter should specify the alpha value it's expecting.  A common value is 385 ppm/K aka .000385, etc.

Ed

Hello
I agree , 385 ppm is a common value on the market but some companies can use other bespoke values mostly to avoid aftermarket probe , the problem is the ppm/K value is not published on proprietary probe as ASL or Hart and other Wika ....

Pt100 are supposed to run up to 600 Deg C , in the real world Pt100 life time at high temp is quite short , in my project I stay close to ambient temp so 50 Deg C max but I need true accuracy in reading but also repeatability between measurements .

Regards
OS

[EC8010]

First problem the rigidity as a P10 is a ''foil '' and if there is strain that generate resistance variation
Connection need to be perfect as possible some supplier use Cu Ag wire.

Good point on the strain and foil. I don't see a way around that. When you glue a PT100 to the DUT the two different thermal coefficients of expansion almost certainly create a bimetallic strip that will strain the PT100. So you could instead put thermal grease between the PT100 and DUT to allow slippage, but you now have to apply pressure to maintain thermal contact. And that will change as temperature changes the viscosity of the thermal grease and allows it to be squeezed out. The glue is an unchanging thermal resistance.

I use thin 4-core cable in a braid screen to make my connections to the PT100. I clamp the cable to the DUT. I earth the braid screen. When I've been forced to use a connector at the DUT, I've used a 9-way D-type (as used for serial ports in old computers). Use the gold-plated brass turned pin variety rather than the folded steel. They seem to be OK. I assume that a 4-core cable uses similar conductors internally, minimising thermocouple differences.

It sounds as though you need more accurate results than I do, but I found that even for the accuracy I needed, measuring temperature repeatably and reliably was tricky and needed careful consideration of losses. Although many DMMs know about 385 and implement it to give a direct temperature reading, you might want to just set the meter to report a 4-wire resistance measurement and do the temperature calculation yourself.

[tszaboo]

Here is an overview of PT100 accuracy classes.
https://kamet-trading.com/faq/pt100-how-accurate-is-an-rtd-sensor/
The water based calibrations are tempting, but make sure your PT100 is waterproof. Also, high temperature, while the probe might be able to handle it, the wiring most likely isn't.

[EC8010]

I realise I didn't explicitly state why it's so important to use a small sensor and a thin film of glue to the DUT. The thermal mass of the sensor and thermal resistance of the glue form an RC time constant that converts step changes in temperature to exponential. Thermal time constants matter because it can take a very long time for the sensor to reach DUT temperature to the required accuracy. It was the long thermal time constants that made me decide that expensive packaged PT100 were a waste of money. A small sensor bonded by a thin film has a short time constant.

I achieve a thin film of glue by applying the glue (Araldite 24 hour) and heating with a hair dryer (typically 70C). This reduces the viscosity of the glue, allowing gentle pressure whilst glueing to squeeze excess glue out, reducing film thickness. Reduced viscosity also allows surface tension to draw the glue into surface roughness, improving the thermal bond. Care over the details leads to better measured results...

[Overspeed]

Hello

regarding precision / accuracy level if I refer to

There are two resistance tolerances specified in DIN/IEC751:
Class A = ±(0.15 + 0.002*t)°C or 100.00 ±0.06 O at 0ºC
Class B = ±(0.3 + 0.005*t)°C or 100.00 ±0.12 O at 0ºC

Two resistance tolerances used in industry are:
1/3 DIN = ±1/3* (0.3 + 0.005*t)°C or 100.00 ±0.10 O at 0ºC
1/10 DIN = ±1 /10* (0.3 + 0.005*t)°C or 100.00 ±0.03 O at 0ºC

I think I can sort 1/3 ( cheaper ) to select some low tolerances value and match 2 sensor as the purpose is a differential measurement as I succeed to match /sort SMD resistor I can succeed to match/sort  Pt100 sensor

I suppose than metrology Pt100 probe are selected or perhaps built as wire wound resistor

https://instrumentation.com/5615-secondary-reference-probes-instrumentation.aspx , for information this probe is in range of 1700 USD !!!!!! new
one connector = 40 USD
one meter / 3 feet of quality cable / wire = 40 USD
one cable spring holder = 20 USD
one SS 319 probe sleeve tubing = 150 USD
one handle in SS304 = 150 USD

Not included the calibration and the assembly . that let a bit of budget for the sensor itself

For the setup
the two main type of measure are contact or non contact which are the same than thermocouple temp measurement , in my case I can use both but for initial setup I prefer contact sensor to metal
I usually use 3M sealant glue over the sensor and not between the sencor and the metal part I also glue the wire to avoid any ''pull force affect '' on the small sensor

Calibration probe are not Pt100 contact probe as the probe is protected by a stainless steel shield / tube , the problem is the reaction time when measuring air temperature ( grill shield ) or thermal conductivity plain shield , same the same again with thermocouple probe

Fast temperature measurement are done by using very thin thermocouple with direct contact  on the surface , problem is the accuracy / precison as difficult with thermocouple to reach accuracy of a Pt100 sensor .

I use 2 pair wire cable with braided shield , wire are copper silver plated ( military surplus wire )

Regards
OS

[Poroit]

I managed a NATA Lab for over 40 years, so I can see from your post that you have some very good gear to carry out the calibrations.
Do you have a contact at  LNE, i.e. France's National Metrology Institute you can speak to re a method of Test?
Our NML contacts in Oz were always pretty helpful.

[Overspeed]

I managed a NATA Lab for over 40 years, so I can see from your post that you have some very good gear to carry out the calibrations.
Do you have a contact at  LNE, i.e. France's National Metrology Institute you can speak to re a method of Test?
Our NML contacts in Oz were always pretty helpful.

Hello
Yes we have the LME or even COFRAC Lab but that not really both open mind organization and they are more useful to '' send the equipment and pay the invoice for the calibration certificate '' and additional measurement / service are extremely costly even big companies with large budget try limit these costs.
 
My goal is to learn and save money to spend money in things I cannot do .
I have do that with my calibration / reference resistor boxes home made built and I plan to do that with my Pt100 probes set 

If you check the picture I have linked of the glass tube Pt100 thermometer lab that not rocket science and not far from a lab reference electrical resistor a two stage resistor with the Pt100 wire resistor and connections .

Regards
OS

[dietert1]

Let's correct something: The usual PT100 sensor has alpha = 0.003851 /K = 3851 ppm/K, not 385 ppm/K. Its resistance is 100 Ohm nominal at 0 °C and it increases to 138.5 Ohm at 100 °C. The 1000h stability spec of 0.04 % of the Farnell 219-1840 mentioned above translates to about 0.1 K.
I have been using the thin film PT1000 with good results. Probably easier to measure in 2 wire mode. When soldered into the PCB next to the DUT it did not exhibit significant thermal lag. I could unsolder it and reuse it for something else.

Regards, Dieter

[Conrad Hoffman]

You say you want a differential measurement. How far apart are the two temperatures? How hot? When doing very high resolution measurements I've found that the platinum sensors may not be the best choice. For mid temperatures I've used thermistors with success because of their high sensitivity. Linearization is harder, but todays processors make it easier than the old funky bridge arrangements. OTOH, there are many times when the platinum sensors are the only way to do the job properly.

[EC8010]

... the problem is the reaction time when measuring air temperature ...

Exactly. Air to the PT100 sensor is a large thermal resistance and the thermal mass of the PT100 sensor once buried in a tube is high. That's why you see a very slow reaction time. What I have done there is to mount the PT100 on a small piece of 0.1mm folded tin plate (to protect it) at the end of the sleeve so that air can pass over it. The tin plate (being steel, and thin) has quite high thermal resistance so it effectively decouples the thermal mass of the tube from the sensor.

If you don't have glue between the sensor and the DUT, you probably have an air gap... That's why transistors are mounted on heatsinks with a really thin film of thermal grease - to fill the air gaps within the surface roughness. Usually, people use far too much grease and increase thermal resistance rather than reducing it.

[Overspeed]

You say you want a differential measurement. How far apart are the two temperatures? How hot? When doing very high resolution measurements I've found that the platinum sensors may not be the best choice. For mid temperatures I've used thermistors with success because of their high sensitivity. Linearization is harder, but todays processors make it easier than the old funky bridge arrangements. OTOH, there are many times when the platinum sensors are the only way to do the job properly.

Hello

Yes that a solution as 1504 Tweener Thermistor Readout show : Typical accuracy of a 1504 is ±0.002 °C with a resolution of 0.0001 °C.

Agilent 34970A can read thermistor as there is : KEYSIGHT 34308A Thermistors (x 5 pcs, 10 kΩ, 0-75°C, for 34970A /344xxA series )

Can a quicker solution to reach easy accurate solution and afordable as Mouser sale around 5 Euro per piece

Regards
OS

[mendip_discovery]

You might want to take a look at Euramet's website, the one about calibrating temperature block calibrators might help you with getting good results and getting rid of some of the errors that can be found.

https://www.euramet.org/publications-media-centre/calibration-guidelines

[KT88]

when measuring gasses self heating becomes a challenge. convection makes it even more unpredictable…

[Overspeed]

Hello

As measurement done on Pt100 for calibration that seems to be

a 0 deg C resistance measurement
two or three point measurement on the sensor range , can be done by sensor comparison or triple point measurement

I link a screen print of Isotech doc

So first step is to measure each Pt100 at 0 Deg C with the triple point of water as triple point liquid is alcool ( to link water outside the tube to the inside / internal sensor to calibrate , alcohol shall not damage ( normally the bare Pt100 sensor .

That a mandatory first step to sort / match sensor

Regards
OS

 

[dietert1]

This log shows a sigma of about 10 ppm (0.001 Ohm) which translates to about 1/385 K = 2.5 mK.
In order to get there the resistance measurements need to be in 4 wire mode. In 2 wire mode there will be much larger errors due to connector and wire resistances.

Regards, Dieter

[tooki]

Regarding self-heating: is there any reason not to energize the PT100 only for a brief moment to perform the measurement, then discontinue the excitation current until the next measurement?

[Overspeed]

This log shows a sigma of about 10 ppm (0.001 Ohm) which translates to about 1/385 K = 2.5 mK.
In order to get there the resistance measurements need to be in 4 wire mode. In 2 wire mode there will be much larger errors due to connector and wire resistances.

Regards, Dieter

Hello

The measuring setup is a ''little'' concern as even in 4 wires setup that request high precision bridge and bridge PSU but also connections , normally Pt100 are feed with a 1 mA current

Question what is the best ( realistic regarding feasibility ) 1mA current source does a voltage ref and a high performance OP AMP is enough as source or does that request other setup as a mirror circuit or ??

Measuring voltage is more easy than generate precision current

Regards
OS

[Overspeed]

Hello

as 1mA is low level of energy and the cold mass of the triple point is high I don t think that generate a problem even is the liquid is stagnant

That just a point of view

Regards
OS

[KT88]

a liquid medium is less critical indeed. however it won‘t hurt to calculate the power dissipation and at least get an estimation of the thermal resistance involved. in case the deviation is concerning a Pt1000 might be a better option…

[dietert1]

10 ppm of 100 mV is 1 uV. So the measurement needs to include voltage reversal in order to suppress thermal EMF.
A good 1 mA current source is a 10 V reference with a 10 KOhm resistor. It isn't constant current though, but needs a little calculation.
Power level will be about 100 uW. With a little effort one can make a constant power driver to reduce effect of self heating.

Regards, Dieter