Precision Temperature Measuring

ref200 pt100 14 bit adc

Hello to all

Could anyone give me some tips on building a precision (0,1°) Temperature measuring system with PT100 RTD.

It's not so easy as it sounds. :roll:

Nice Weekend

Michael

I do not believe you can build such a precise thermometer without performing a calibration. The problem is not building the thermometer but to have a precise temperature standard. you can use icecubes in watercup at a known atmospheric presure and boiling water at a known atmospheric pressure as calibration points but I doubt if that will be precise enough. I suggest to borrow or rent a calibrated precision thermometer for calibration purposes.

If you see Laser trimmed temperature sensore from National they give best accuracy of .5 to 1 degree . 0.1 is way too much you are demanding .

i hope these files helps you.
Regards.


/ Warning #1 - Files deleted. Don't upload files from the net! Just post link instead!:
**broken link removed**
**broken link removed**
https://focus.ti.com/lit/ds/symlink/ina114.pdf
https://focus.ti.com/lit/ds/symlink/ref200.pdf

take a look at this appnote from maxim :

**broken link removed**

it tells you how to increase the DS1624 0.5° precision to a better 0.05° with a good calibration.

hi kripton,
you need a nist level thermometer for you can this. i dont think so its not practical solution.
Regards.

find information in www.smartech.nl
easy

hystersis

Hysteresis is another problem. If you take the device outside the temperature range it was calibrated at and then back into the range there will be systematic errors in the reading.

there are manys devices to realize this function
you can hunt and buy it

Hi!

You can build a circuit which corrects the non-linearity error in PT100
temperature sensor. The total error is about 2.5 deg C over the range of 0 to 200 deg C. The correction has to be provided in the constant current source used as a supply to the PT100 sensor. It is possible to achieve
+/- 0.1 deg C over a range of -100 to 200 deg C. The constant current source is controlled in a closed loop circuit which provides correction in the current source, supplying to the PT100 sensor, proportional to the sensed temperature.

I could have provided the circuit but the design is owned by some company manufacturing temp. controllers.

Bye!

dudu said:

find information in www.smartech.nl
easy

Click to expand...

I just check your link. It's nothing there!

I don't know if it's still in production but there was an IC: XTR103, produce by TI (BurrBrown) special design for RTD excitation and linearization.

cold junction

Don't forget that your cold junction compensator contributes to the error.

[I am embarrassed about forgetting what the topic was about. I should have said that the self heating from measuring the resistance will produce error. Pulsed methods should be used.]

Thermocouples only introduce additional error due to cold junctions.
The PT100 is platinum resistance thermometer (RTD), so don't require for a cold junction compensator.

[Don't be embarassed. It happened to me too after hours of moderating job.]

How achieve high accuracy temperature measurements is not then an easy matter, however some suggestion may be given here.

1. Four wires PT100 should be selected or at least 3. This will compensate (sensing) for leads resistance.

2. Choose good sensor. Rosemount or Minco o maybe Heraus. In PT100 the worst error is long time stability, due to PT100 contructions factors.

3. Select a class A PT100. That's means 0.1 + 0.002|t| (°C). Class B is 0.3 + 0.002|t|.

4. When you have purchased PT100, send it to NIST if you live in USA or EAL if you live in EUROPE and ask for certification with not less of 6 temperature points, one of them passing from 0°C (you will need it, because you require R0, that is R at 0°C). From the certificate, calculate the coefficients A, B and C. R0 is reported at 0°C.

5. If you cannot calculate the coefficient for your RTD, you may keep the standard coefficients, that is A=3.90830e-3, B= -5.7750e-7, C=-4.2735e-12 and R0=100 ohms. In that case you will have and error of +/-0.5°C, provided that you have purchased a class A PT100.

6. Linearize RTD using VanDusen equations. Hence : Rt=R0+(1+At+Bt^2) from 0 to 850°C, or Rt=R0+[1+At+Bt^2+Ct^3(t-100)] from -200 to 0°C. I guess you will use a micro to do this job. Analogue methods are difficult to implement if you need high accuracy.

7. Excitate the PT100 with a costant current generator. The value should be less of 1mA to prevent the auto power error (the PT100 dissipate a power so it will generate heat due to joule effect). A current value of 0.5mA or better 0.2mA should be nice. Obviously a good ADC with high resolution is a must. Select a delta sigma one. ADI or Linear Technology would be ideal.

8. You may use a DC current to excitate RTD. Some ones suggest to use an AC current to null the effects of parassite cold junctions that would exist from Platinum and Copper junction (cable and connectors) and prevent the autopower due to Peltier coefficients of above junctions. I don't think so. The Peltier auto power maybe calculated from the FEM of the parassite junctions multiplied by excitation current, let say 1mA, so given that the FEM is not more of 15uV for platinum, the Peltier auto power is 15uV x 1mA=15nW (joule effect due the Peltier juction). This value is unmeasureable compared to auto power of PT100 excitated at 1mA, for instance at 0°C, that is 100*1mA^2= 0.1mW. To reduce the effects of parassites junctions you can revert excitation current at low frequency, 5 or 10 time per seconds and then take averages of absolute values. However this add an additional complessity to your system. From my experience a DC method will suffice to achieve 0.1°C or bit better, provided that you have considered what above said.

Followings note maybe useful.

Don't forget that if your sensing system is converting the voltage from the sensor with a ADC, the accuracy and stability of the voltage reference will have a big effect too! Sometimes the voltage reference can be as troublesome as the sensor itself.

TDC

Yes sure ! The Analog Device ADR 291 is a good one.

If some looks at the expression

Rt=R0*(1+At+Bt^2)
A=3.90830e-3, B= -5.7750e-7, R0=100

Then the RTD thermal coefficient will be appr

TEMPCO = R0*(A + 2*Bt) = 100*(3.9E-3 - 6E-7*t) = 0.39 - 1.8E-4 Ohm/K at 300K

For bias current = 100 uA the thermal sensitivity will be about 40 uV/K. For 0.1K resolution some have to have readout with 2*PI times less combination of input voltage noise and current noise, i.e less than say 500 nV rms. If some assume 1 Hz signal bandwidth, then the combined noise of the front end amplifier shall be less than 500 nVrms/RtHz.

Also the bias current generator should have noise less than Rt*in < 500 nVrms/RtHz.

If some uses DC current bias then the frontend OP should have low 1/f noise ( a chopper stabilized amplifier is preferable, but for less demanding applications OP97 should be OK).

Using AC current bias is equivalent to using chopper stabilized amplifier with the important difference that the RTD component is located after the modulation and before demodulation, hence the additive (not the multiplicative) out of band noises will be rejected. Building a lock-in amplifier is real fun and good training in a bit more advanced signal processing technicues. The AD624 is good choice for a lock-in chip.

If some assume 1 Hz signal bandwidth, then the combined noise of the front end amplifier shall be less than 500 nVrms/RtHz.

Click to expand...

how do you suggest bulit 1HZ low pass filter ?