We are running TMP006 inside one of our products and are having difficulty getting the measurement accuracy the sensor chip is supposed to give. By following the procedure in the user's guide -- plotting the calibration function v/s (Tobj4 - Tdie4), fitting a line to the datapoints and taking the line's slope as the calibration constant (S0) -- we experience large measurement errors. However, running the EVM on the outside of the product measuring the same surface temperature simultaneously, the calibration constant found in a similar way gave very accurate measurements. I have included calibration function plots and line fits from both cases (see a MSWord file attachment). I would much appreciate it if anyone can throw some light on the likely cause for the large errors from the sensor inside our product.
It is worthwhile to note that the datapoints in our product's case do seem to fall along a linear trend line. The source of the error seems likely because they lie on the 'wrong' line. By looking at the calibration/measurement equations, it is easy to see that the calibration function v/s (Tobj4 - Tdie4) should ideally be a line passing through origin. We suspect that the errors we experience in our product is probably due to the fact that the line fit to our datapoints doesn't pass through origin. This is visible in the plots in the attachment -- notice that the "y-intercept" in our product's case is 100X that in the EVM's case. Any idea why we are seeing this?
FYI, the nature of the target surface: aluminum case painted dull black.
Thanks!
