• State TI Thinks Resolved
  • Locked Locked
  • Replies 13 replies
  • Answers 6 answers
  • Subscribers 32 subscribers
  • Views 643 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.

TMP61-Q1: achieve +/- 1C temp error

Elec Cheng
Expert 7055 points
Part Number: TMP61-Q1
Other Parts Discussed in Thread: MSP430F5529, MSP430G2553, TMP61

Hi Expert,

Our customer would like to achieve +/- 1C temp error with TMP6131QDYARQ1. I see the requirement in this e2e thread. But there are several items they cannot meet

https://e2e.ti.com/support/sensors-group/sensors---internal/f/sensors---internal-forum/1014127/tmp61-q1-accuracy/3787779?tisearch=e2e-sitesearch&keymatch=tmp61%2525252525204th#3787779

First of all, the ADC the use is 12bit. And due to MCU resource limit, they cannot implement 4th-order polynomial. They plan to use 2th-order instead. Under these conditions, can they achieve +/- 1C temp error result? 

Another question, what's the procedure of applying 4th-order polynomial? What does it mean marked yellow? 

And what about the coefficient? Does it apply to all TMP61-q1 ? Or is it just an example and the customer need to calculate themselves? Thanks.

BR,

Elec Cheng

  • 0
    TI__Guru* 91206 points

    Dear Elec - 

    Thanks for the post - here if you go to the 4th Order Polynomial TMP vs Vdc tab, in the Thermistor Design Tool you referenced, you will see the recommended code example to use. The one you were looking at was for using a calculated R value, which is really unnecessary, since the device output voltage is all that is needed for input to ADC.

     

    Also, please do make sure to look at the Averaging and Low Pass Filter tabs and leverage either of those examples in your code, too. 

    Lastly - please advise as to the MCU being used, this exact code has been implemented on general purpose 16 bit and 32 bit MCUs with 12 bit ADCs and does not require any special resources - it was done first on 32 bit MCU with FP capability (MSP432), but we have also implemented same code with same results on 16 bit MCUs such as MSP430G2553 and MSP430F5529. So we would be curious to know that detail so as update our code example to support whatever we have missed here. Thanks! 

  • 0 in reply to Josh Wyatt
    TI__Expert 7055 points

    Hi Josh,

    Thanks for pointing out. Yes I will use the "4th Order Polynomial TMP vs Vdc" sheet. The customer will use our C2000 also so I believe the ADC is in C2000. 

    Josh Wyatt said:
    Also, please do make sure to look at the Averaging and Low Pass Filter tabs and leverage either of those examples in your code, too. 

    What's the difference between them? Can we use both? If we can use only one method, which one do you propose? 

    As for Low pass filter sheet, I know Alpha determines the filter result. But how to choose the right one? Do we have foemula? 


    Elec Cheng said:
    First of all, the ADC the use is 12bit. And due to MCU resource limit, they cannot implement 4th-order polynomial. They plan to use 2th-order instead. Under these conditions, can they achieve +/- 1C temp error result? 

    Also, how about this question? Can they achieve? Thanks.

    BR,

    Elec Cheng

  • 0 in reply to Elec Cheng
    TI__Mastermind 43730 points

    Hi Elec,

    Either the averaging or low pass filter can be used with similar effects.  There is not really any benefit to using both.  This is just to reduce the noise that is seen in the raw measurements.  The low pass filter is quite effective and does not require much code to implement.  Typically an alpha of 0.1 will provide a good smooth result, but if you require a faster thermal response time, a higher alpha may work better.  In the thermistor design tool spreadsheet, you can alter the alpha value and see the resulting filtered data vs. raw results.  

    +/- 1C temperature results are possible with this setup and the 4th order polynomial as we show in the thermistor design tool which also provides the generated C code.  Keep in mind that each thermistor would also need to have a single point offset calibration in the system.  This can be performed at room temperature.

    Is there a reason you are not able to use the 4th order polynomial?    

  • 0 in reply to Eddie LaCost
    TI__Expert 7055 points

    Hi Eddie,

    Eddie LaCost said:
    Keep in mind that each thermistor would also need to have a single point offset calibration in the system

    Do we have guide for this?

    Eddie LaCost said:
    Is there a reason you are not able to use the 4th order polynomial?    

    The customer says they have many other items to calculate so they need to save some C2000 capability here. Is there really big difference between 2th org\der and 4th order? Thanks.

    BR,

    Elec Cheng

  • 0 in reply to Elec Cheng
    TI__Mastermind 43730 points

    No, we don't have a guide for offset calibration.  It is really just measuring the temperature with a high accuracy temp sensor, then measure with the TMP61 and add or subtract the offset to between the two to your TMP61 measurement.  Very simple and can be performed at any temperature since the TMP61 is very linear.

    We recommend the 4th order polynomial as this is what our testing is done with.  Let me consult with the expert about using a 2nd order polynomial and I will let you know what I find out.

    Another thing that I did not mention earlier is the PPM rating of the Rbias resistor.  This will play a big role in the accuracy that can be achieved over temperature.  For +/-1C accuracy, I recommend a 0.5%, 50ppm or lower Rbias resistor.  

  • 0 in reply to Eddie LaCost
    TI__Expert 7055 points

    Hi Eddie,

    Thanks for your feedback.

    Eddie LaCost said:
    We recommend the 4th order polynomial as this is what our testing is done with.  Let me consult with the expert about using a 2nd order polynomial and I will let you know what I find out.

    For this question, could you please help confirm today since this question has been pending for several days and the customer is pushing. Thanks for your understanding.

    BR,

    Elec Cheng

  • 0 in reply to Eddie LaCost
    TI__Mastermind 43730 points

    Regarding the 2nd order polynomial question, it can be done, but accuracy will be lost(on the order of +/-0.1C).  The 4th order polynomial ensures that we are tracking the curve more accurately.  You can find the 2nd order polynomials for TMP6131DEC with 10k Rbias and Vbias = 3.3V.

    It does not seem to me that a 4th order polynomial would be much of a strain for the C2000 capability since this is pretty powerful controller.     

  • 0 in reply to Eddie LaCost
    TI__Expert 7055 points

    Hi Eddie,

    Sorry that it's still not very clear to me.

    If our customer meets all the requirements mentioned in below thread, except that ADC is 12bit, and they will use 2th order polynomial. They customer can apply "low pass filter" as well. Under these conditions, can they get +/- 1C temp errors? 

    https://e2e.ti.com/support/sensors-group/sensors---internal/f/sensors---internal-forum/1014127/tmp61-q1-accuracy/3787779?tisearch=e2e-sitesearch&keymatch=tmp61%2525252525204th#3787779

    Eddie LaCost said:
    It does seem to me that a 4th order polynomial would be much of a strain for the C2000 capability since this is pretty powerful controller.   

    Since you think 4th order adds to0 much work to C2000, do you verify 2th order polynomial? If not, will you perform? 

     

    BR,

    Elec Cheng

  • 0 in reply to Elec Cheng
    TI__Mastermind 43730 points

    I have an unfortunate typo and will fix it in my post.  Sorry about that.  I meant to say that it "would not" be too much of a strain.  I think performing a 4th order polynomial will be very easy for the C2000 to perform and is what I recommend.  

    They can meet the +/-1C error using a good PPM Rbias resistor and 4th order polynomial.  Our testing is with a 4th order, so I cannot say for sure that 2nd order would be acceptable.  You would need to test this.  My recommendation is to use the 4th order which will provide best accuracy.

  • 0 in reply to Eddie LaCost
    TI__Expert 7055 points

    Hi Expert,

    Thanks for your help. The customer can use 4th order. They want to achieve +/- 1C temp error with TMP6131QDYARQ1 in full temp range. Could you please tell the specific requirement they need to meet? Also, do you have an example for that calculation? Thanks.

    BR，

    Elec Cheng

  • 0 in reply to Elec Cheng
    TI__Mastermind 43730 points

    The temperature accuracy over the full temp range will be highly dependent on the Rbias tolerance and PPM.  The single point offset calibration is also very important.

    Figure 1-7 in the note below shows some worst case numbers for differing Rbias resistor types. 

    https://www.ti.com/lit/ug/sbou246/sbou246.pdf     

  • 0 in reply to Eddie LaCost
    TI__Expert 7055 points

    Hi Eddie,

    There is a sentence below the chart, saying that it assumes ideal Vbias at 5 V and an ideal Vref with no ADC error. But we need to take that into account. Could you please share the conclusion with Vbias and ADC error? Thanks.

    BR,

    Elec Cheng

  • 0 in reply to Elec Cheng
    TI__Mastermind 43730 points

    Elec,

    I will send you an email separately with some calculations that can be used for derating Vref and Vbias.