CN0287 accuracy analysis

Hi,

One of our new designs, requires us to implement RTD & Thermocouple measurement circuit.

We need to get a accuracy of better than "0.2% of FSR" over the operating temperature range of "-20°C to 60°C".

We have looked at CN0287 board & planning to use the same circuit in our board if the required accuracy is met.

The CN0287 specification says that accuracy of 0.5°C. Is this accuracy guaranteed over a specific temperature range?

For our accuracy analysis , we have identified the error introducing components as below :

RTD measurement :

  • Resistor Rref  - Tolerance & Temperature coefficient
  • ADC AD7193 - ENOB , INL , Offset error, Gain error , Offset drift , Gain drift

Due to ratio-metric cancellation,  change in excitation current & ADC reference voltage cancel out each other. Thermo-couple measurement :

  • ADC AD7193 - ENOB , INL , Offset error, Gain error , Offset drift , Gain drift
  • ADC Reference ADR3425 - Initial Accuracy, Line regulation , Load regulation, Output Noise
  • Cool Junction Compensation ADT7310 - Accuracy ( 0.5°C )

Please let us know if we have missed out any parameter in our analysis.

Also, it will be helpful if you can share any document related to accuracy analysis done on CN0287.

Looking forward to your response.

Regards,

Suresha N S

  • 0
    •  Analog Employees 
    on Dec 10, 2013 8:35 PM

    Hi Suresha N S,

    CN0287 system performance varies under different type of sensor. The 0.5°C is the accuracy for thermocouple due to the limitation of Cold Junction Measurement Accuracy from ADT7310. We actually have ADT7320 with 0.25°C accuracy for performance upgrade. I’m pretty sure that the board could get the accuracy of 0.16°C (Calculated from 0.2% of “-20°C to 60°C”) very easily if using RTD to be the sensor.

    One thing need to pay more attention when using RTD: The protection components will generate small leakage current that cause difference between the current going through the RTD and the current finally going through reference resistor and generate additional error as calculated in the circuit note. I would highly recommended you to select the low leakage TVS and diodes if the protection function is necessary. The TVS(PTVS30VP1UP) and clamping diodes(BAV199LT1G) are recommended unless you could find other components with lower leakage.

  • Hi Brandon,

    Thanks for your response.

    My apologies, I have missed out on mentioning the actual temperature measurement range in our design.

    Our requirement is to support RTD to measure temperature for -60°C to +250°C. We need to achieve 0.2% FSR accuracy which means a accuracy requirement of 0.62°C.We will try to implement the protection circuit as per your suggestion.For thermocouple our requirement is to achieve a accuracy of 0.2% over the measurement range of "-200°C to +800°C". This results in a accuracy requirement of 2°C.

    We also need to maintain the above mentioned accuracy over the operating temperature range of "-20°C to 60°C" of our board. Thus we would like to know if any accuracy analysis was done on CN0287 circuit over a operating temperature range ?

    Also, we had one query regarding your response.

    You have mentioned that the 0.5°C accuracy of CN0287 is due to the limitation of cold junction temperature measurement. We assume the accuracy of measurement in case of thermocouple is also limited by the accuracy of ADC & ADC reference voltage.

    Is there a specific reason for limiting the accuracy analysis to cold junction measurement only ?

    Regards,

    Suresha N S

  • 0
    •  Analog Employees 
    on Dec 12, 2013 8:36 AM

    Hello Suresha N S

    For this circuit, we assume customer doing the system calibration to cancel out all the linearity error. Besides that, the non-linearity error (drift, noise) would decide the final accuracy performance. We only tested the circuit under room temperature and ignoring the drifting effect. By testing, we find that the noise performance which could decide the final resolution, is good enough to satisfy the system accuracy requirement. The real bottleneck  to get better accuracy comes from ADT7310 with 0.5°C accuracy which is used for CJC in this circuit without considering drift.

    For general analysis, first, you need to find the dominated error items in your circuit. For CN0287, the dominated error items for thermocouple circuit is from 1st: voltage reference, 2nd: Accuracy of CJC. The both error items effect the final system accuracy directly. For 80°C working temperature range, ADR3440(8ppm) will generated 640ppm = 0.064%. Assuming your signal covers 50% of dynamic range of ADC by configuring the gain block in AD7193 properly, the final effect by reference is 0.13%. For -200~800°C measurable range, the accuracy would be 1.3°C. Use square root of 1.3 and 0.5 = 1.4°C.

    In summary, the system accuracy limitation would come from ADR3440 if considering the 80°C temperature changing.

    For CN0287 using RTD sensor, the main error items would be the reference resistor with 10ppm temperature coefficient. So, 800ppm for 80°C temperature difference. RTD has common mode voltage that limit the gain of the signal chain. For -65°C~250°C, the resistance of PT1000 would be 763.3 Ω to 1941 Ω. With 4.02K reference resistor, actually, we could set the gain to 2 but this is not available for AD7193. The only available gain would be 1. So that (1941-763.3)/4020 ≈30%. The final system accuracy would be around 800ppm/30% = 2400ppm = 0.24%.

    So you need to use reference resistor with 5ppm to meet your system performance.

    Also, other error items such as gain drift, offset drift, leakage would effect the final performance as well. You need to do more calculation to get closer to the real result but the real testing result beats all the calculation.

    And some times real testing will be more easier than calculating.

    Wish you all the best.

    With best regards

    Willie.

  • Hello Willie,

    Thanks a lot for detailed response.


    We had one query with regards to your response.

    Our operating temperature range in -20°C to +60°C. Thus the maximum variation from 25°C (Considered as ambient temperature of our board) is 25-(-20)= 45°C.

    Thus in thermocouple calculation, ADR3440 (8ppm) temperature change will generate 45 x 8 = 360ppm variation. This reduces our final effect of reference on accuracy to 0.72°C & RMS accuracy value will be 0.87°C.

    Similarly, in case of RTD, resistor variation will reduce to 10 x 45 = 450ppm. This will increase RTD accuracy to to 0.15%.

    Is our calculation correct ? Please comment.

    Also, we are planning to do a complete accuracy analysis for our design based on CN0287,including ADC parameters such as INL, gain drift ,offset drift etc.

    Will it be possible for you to review our accuracy analysis document ? This would really help for us to go ahead with the design.

    Regards,

    Suresha


  • 0
    •  Analog Employees 
    on Dec 12, 2013 5:31 PM

    Hello Suresha N S

    Your analysis is correct from my opinion. It's more reasonable to analysis the temperature refer to room temperature.

    My pleasure to review your accuracy analysis document too.

    Regards,

    Willie.