Hello Abhi,
I do understand that error will vary between an analog and digital out part. In this case, I'm talking about a specific analog out part (TMP20). I'm asking for information specific to this part and not temperature sensors in general.
What I'm saying is that the error curve over the full temperature range - for 39 parts - follows a 2nd degree function. Since this function can be estimated (best fit method) it would greatly increase the accuracy of the output. I don't understand why the error looks more deterministic than random.
Could Figure 7 be depicting values for row 1 in Table 1 (Temperature limits and maximum deviation seem to correspond → See ±1.41 on graph @ 130°C)