Hi folks,
I am working on a temperature measurement project. The current design is to connect a RTD(PT100) and a reference resistor(0.01%) in series, drive them with a constant current source and then measure the voltage across both RTD and reference resistor by SD24 ADC simultaneously, calculate the resistance of RTD and deduct temperature from the resistance of RTD. The circuit works reasonably well in terms of repeatability and precision. The problem is accuracy. I am trying to develop a calibration strategy which will rely on error analysis and calculation.
The errors come from the following sources: 1) nonlinearity of SD24 ADC; 2) offset error of SD24 ADC; 3) the gain error of PGA; 4) inaccuracy of reference resistor.
The datasheet of MSP430F6720 that is used in the project states the error of 1) is about 0.01%, when RTD is close to reference resistor, this part of error is really neglectable. error of 2) can be as high as 73mV, but this can be calibrated in software so not a big deal.
What really beats me down is the error of PGA gain, it is as high as 2%. Even when gain = 1, it is still 1%. That means roughly 4% error in the measured resistance of RTD if my calculation is right. If the reference is 80ohm, the error could be up to 3.2ohm, an error of rough 8 degree C in temperature. Of course in reality it is not this bad. The initial accuracy of my first 10 boards is usually within +/-0.3C.
My question is, is the error of pga gain consistent over time for each particular MCU? If that is true, since this is largely offset error, I can calibrate the measurement circuit against a reference temperature.
Please advise and thanks in advance.
