Hi Frank,
The datasheet spec 283 mV may typically represent room temperature, but it could vary with temperature and due to process variartion across mutliple parts as you have noticed from the above posts. I should have said that it also varies with throughput.
I didn’t see anywhere in these posts, where the other users are explicitly trying to measure two separate and controlled temperatures. I'm not sure if they are running the part at 500 ksps or some other throughput. Have you tried to use an external reference for this?
You are making this 0.1 mV/C assumption based upon possible process and throughput variation across few parts and It’s a flawed analysis of our temperature sensor.
Regards,
Maithil