ADS1256: ADC Accuracy comes from INL

Hi Takumi-san,

Your reasoning is correct. I would make a couple of additional comments to your analysis:

1. By using +/-10 ppm (20 ppm), you are effectively doubling the INL error which never exceeds 10 ppm. Therefore, I would propose that the worst case accuracy should be 16.61 bits (+1 bit higher than you initially calculated).
    
2. Worst case INL is a very conservative specification. In most cases, you will probably observe the typically 3 ppm performance, which provides 18.35 bit ( = log2[1,000,000 / 3]) accuracy.
   
   
3. Accuracy and resolution are quite different, and with any high-resolution system you will always find that accuracy is worse than the resolution.
   
   
4. INL is indeed a very important specification as it is very difficult to remove this error through calibration; therefore, INL effectively places a limit on the maximum accuracy you can expect to achieve in your system (after calibration). However, few systems calibrate out all drift errors, so in most cases the accuracy due to temperature related effects will be the biggest source of error. By providing 24-bits of resolution, the ADS1257 allows you the ability to calibrate your system to achieve greater than 16 bits of accuracy; whereas a 16 bit ADC cannot achieve an accuracy above its resolution.

Best regards,
Chris