• Locked Locked
  • Replies 7 replies
  • Subscribers 84 subscribers
  • Views 994 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.

TMS320F28035 temperature sensor inaccuracy

David Clemans
Prodigy 10 points

I am seeing very bad accuracy of the internal temp sensor.  We have 2 lot codes of parts that had a 30C positive offset.  Recent lots exhibit anywhere from -19C to +20C offset at room temp.  After discussing with TI and engaging their FAE, TI verbally has said the sensor is no better than +/- 20C. I don't know what application can use that poor of an accuracy. They claim it is primarily for calibrating the internal oscillator. 

We made the naive mistake of setting an overtemp limit with the internal temp sensor. 

They claim that they improved their process in Jan 2011 so we should not see any parts that are worse than 20C error. 

The sensor variation within a lot code ranges about 20C after looking at a few samples from several lots.  We really didnt look at a statistically significant number of samples however. 

The temp sensor should not be used.

  • Prodigy 60 points

    I may be seeing what you're seeing. I'm using the internal board temperature with a 3v reference and I'm seeing 95/100C as the feedback.  What ACQPS value are you using?  I do the dummy read, so I don't think it's an issue, but I'm trying to see what I can do to make this more accurate.

  • in reply to Nicholas Cross
    TI__Intellectual 1105 points

    Nicholas,

    Please see the following post on temp sensor for an explanation.

    (bad post deleted)

    Regards,

    Ricky Rosinski

  • in reply to Nicholas Cross
    TI__Intellectual 1105 points

    Nicholas,

    Sorry for the bad link.  Here is a summary of the post.

    The primary function of the internal temperature sensor is to calibrate the internal 0-pin oscillators.  As a secondary function, it can be used to measure absolute temperature, but there are production limitations on the achievable accuracy against absolute temperature.

     
    For the F2803x (and other piccolo devices) TI is in the process of upgrading the production capabilities to give better absolute temperature calibration.


    There will be datasheet and other documentation updates in the future, but until then the temperature sensor is provided "as is" with no specification of absolute error bounds.  It is worth noting that the limitations on accuracy are calibration limitations only; the temperature sensor itself is quite linear with reasonably precise and reproducible readings.  If the user can provide a temperature reference point, later sensor readings should be faithful to within a couple degrees. 

    Regards,

    Ricky

  • in reply to RickyRosinski
    Prodigy 60 points

    Thanks! So the scale and offset will probably be different for each chip then.  I'll see what I can do.  Thanks for the help.

  • in reply to RickyRosinski
    Prodigy 20 points

    Hi Ricky,

    I didn't found new datasheet update concerning the accuracy of the temperature sensor of the currently produced CPU.

    I made some tests with the CPU I have now in production and I get very accurate temperature measurements (checked with an IR camera to have the real temperature of the silicone). But I need some specifications to be sure I can use this information on a large production volume.

    So, do you have any news?

    Thanks for quick answer


    Thierry

  • in reply to Thierry Bieler
    TI__Intellectual 1105 points

    Thierry,

    Based on characterization data, you can expect the temperature sensor accuracy to be +/- 20C, but this will not be guaranteed by test or in the datasheet.  If you need an accuracy better than +/- 20C, you will need to use an external temperature sensor.  Just as a reminder, the temperature sensor's main application is to calibrate the internal oscillators to within +/- 1%.

    Regards,

    Ricky