• State TI Thinks Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 2 answers
  • Subscribers 32 subscribers
  • Views 493 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Original question:

TMP117: Accuracy when there is no averaging

TMP117: Accuracy of the temperature sensor

DELJO DAVIS
Intellectual 270 points
Part Number: TMP117

We working on a home indoor camera and need to measure the ambient temperature inside a home. In our device we have TMP117 temperature sensor and it will measure the temperature of a metal part inside the device. TMP117 IC top will touch the metal by a thermal pad. The change in external ambient will change the metal part temperature. So by correlating the result we can measure the external ambient temperature.

But we want a higher accurate measurements and measurement time is not a issue here. Below shown is the accuracy graph mentioned for TMP117

And a table also mentioned at sec 6.5  for accuracy

I want to know the max possible accuracy of TMP117 for the temperature range.

Can you confirm the max possible accuracy of TMP117 only mentioned in above table. Or we can improve that by averaging?

We can see a Average plot and 3 sigma plot on graph. 3 sigma plot should be the distribution of accuracy for each time measurement. So 3 sigma should be the real accuracy of the device right?. But the dark black color plot have higher accuracy than 3 sigma plot. So is the accuracy really increased here or on the next time measurement the dark black color plot position will be change but it will be definitely inside 3 sigma plot?

P/N we using is TMP117AIYBG. We want to reduce the thermal conduction from board and need to focus temperature coming on the top of the IC. From the thermal resistance values found BGA package version of TMP117 will be good and that's why selected TMP117AIYBG.

  • 0
    TI__Expert 7170 points

    Hi Deljo,

    When measuring ambient temperature inside of a house, we can assume that this is a still-air environment with small changes of temperature over long preiods of time. In this setup, thermal response is not an issue because of the slow rate of change of temperature. Furthermore, the TMP117 is guaranteed per the datasheet for ±0.1°C at room temperature which is already a high accuracy measurement. By default, the TMP117 averages 8 samples, and this can be increased to 64 samples if needed--post-processing averaging is not needed. What the ±3σ curves show, is the process deviation to gurantee the temperature accuracy spec from IC production/manufacturing. 

    What is important to consider is the PCB layout to thermally isolate the sensor from any heat sources present on the PCB. My recommendation would be to use distance and cutouts on the PCB to achieve this--please see layout example below. Thus, improving the accuracy of the temperature read with respect to ambient temperature.

    Best regards,

    Simon Rojas

  • 0
    TI__Expert 3910 points

    Hello ,

    In addition to Simon's response, I wanted to ask a follow-up questions:

    • What accuracy are you trying to achieve?
    • What is driving the need for better accuracy?
    • Are you interested in a specific temperature range?

    We might have some soon-to-release device that has improved specs and some package and software similarities to TMP117. I think it might be of interest to you, depending on the temperature range, and accuracy level you are seeking to achieve

  • 0 in reply to Victor Salomon
    Prodigy 10 points

    Hello , btw what it would be this soon-to-release device part number?

  • 0
    Intellectual 270 points

    Please find my reply below

    • What accuracy are you trying to achieve?>> +-0.1C
    • What is driving the need for better accuracy? We not measuring direct ambient temperature but a metal part exposed to external ambient from inside of the device. And we need to correlate the metal temperature with external ambient
    • Are you interested in a specific temperature range? yes 5 to 70degC
  • 0 in reply to Victor Salomon
    Intellectual 270 points

    Hi victor,

    Please  find reply below

    • What accuracy are you trying to achieve?+-.1C
    • What is driving the need for better accuracy?We not measuring direct ambient temp but a metal part temp which exposed to ambient and we need to correlate the metal temperature with external
    • Are you interested in a specific temperature range? yes 5 to 70degC
  • 0 in reply to DELJO DAVIS
    TI__Expert 7170 points

    Hi Deljo,

    Thank you for clarifying. When trying to measure the temperature through a medium, the best way to achieve this is to isolate the sensor from the undesired sources of heat and improve the thermal conductivity with the source of heat to be measured. As mentioned before, cutouts on the PCB and distancing the sensor from other heat sources are the best layout practices for this purpose. Besides this, I would also recommend using a copper pour at the bottom layer--opposite to the temperature sensor along with vias in it. This way, you could have thermal contact from the metal block to the IC (through the copper pour and vias) for better thermal performance.

    Ultimately, the TMP117 is a ±0.1°C temperature sensor that is trimmed in production to achieve high accuracy. Thus, the best way to get an accurate result is to improve your overall system to rely on this guaranteed spec. 

    Best regards,

    Simon Rojas