ADS1248/ADS1247 accuracy calculation on 3/4 wire RTD

Hello Stefan,

The ADS1248 can easily meet and exceed Class A RTD accuracy standards; however 0.01C may be pushing the limits of the accuracy of the device. When using a PT-100 with a 1mA excitation current; a 0.01C change corresponds to approximately ~3.9uV change at the inputs of the device.  The 0.01C absolute accuracy spec will be hard to meet after adding all the errors due to gain, offset, INL and drift errors; and after accounting of any errors due the resistors and passive components involved in the system. 

For example, a very quick error estimate using typical offset, Gain and INL errors when performing a 4-wire PT-100 RTD measurement with 1mA excitation current source, with RBIAS= 2400 Ohms (VREF=2.4V)...

At 25C, the PT-100 is 109.73ohms; this corresponds to a input voltage of ~109mV at the inputs of the ADS1248.

Offset error (typ @ 25C) =  1uV

Gain error (typ @ 25C) = +/- 0.005%*(109mV) = +/- 5.45uV

INL Error (typ) = 6 ppm; this will correspond to 6ppm*VREF/Gain = 0.9uV  (where VREF=2.4V, PGA Gain =16)

Total Error Accuracy Error = square root(offset error^2 + INL error^2 + Gain error^2)

Total Accuracy Error = ~ 5.6uV

This corresponds to a temperature accuracy error of ~0.014C.

I will provide a more detailed analysis tomorrow accounting for the drift of the resistors, however, the device may be able to achieve absolute accuracy in the order of >0.015C.

Thank you and Best Regards,

Luis 

Hi Stefan,  

1)  In the typical Class A RTD application, where the customer does not calibrate the RTD; the ADS1248 device gain error is trimmed for each gain setting and provides a typical gain error in the order of 0.005%.  The user typically only perform a self-offset calibration.  The device offers a system monitor that may be used to detect large temperature changes in the PCB board  (changes of temperature in the order of 10 C); and this feature may be used to minimize the offset drift errors.  Please see the recommended offset calibration procedure attached.

3443.ADS1248_Offset_cal_question.pdf

In the typical ratiometric configuration (as shown on SBAA180), since the voltage across the RTD sensor and the VREF resistor are generated with the same excitation current source, any changes due to the drift of the excitation source will be reflected across the sensor at the input path and across RBIAS at the reference path of the ADC.    In this ratiometric circuit, if the RTD and RBIAS resistances remain unchanged, the digital output of the ADC is unaffected by changes in the excitation source.  The user needs a high precision low drift resistor to generate the voltage reference.  The errors due to the absolute value of the excitation and the errors due to the excitation drift are virtually eliminated. 

In the 3-wire RTD application, any small error due the mismatch of the IDAC and current sources could be corrected using the attached procedure.  The IDAC1 and IDAC2 could be alternated on AIN0 and AIN1; and by performing two measurements and averaging the results, any small errors due to IDAC mismatch can be corrected.  Please see procedure attached.

0523.IDAC_Mismatch correction.pdf

 2)  On very high precision applications, where the user calibrates the complete system, including the RTD, the user could use a very accurate temperature source such as a thermal bath (with guaranteed accuracies in the order of ~0.01C) and perform complete calibration measurements on the required RTD temperature range.  The user could perform a first order gain/offset linear calibration using a two point calibration including the RTD using the thermal bath at the temperature reference, as you have mentioned.  Alternatively, the user could create a look-up table for each temperature.  The drift/accuracy of the precision resistors used play a direct role on the achievable accuracy. 

I apologize for the delay,  will provide some approximate error calculations making assumptions about the precision/passive components ASAP.  if you have drift/accuracy specs of the resistors (Rbias and Rcomp) that will be used, I will include these on the estimate. 

Thank you and Best Regards,

Luis