• State
  • Locked Locked
  • Replies 8 replies
  • Subscribers 45 subscribers
  • Views 834 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

ADS1232 temperature drift

theodore belogiannis
Intellectual 420 points
Part Number: ADS1232
Other Parts Discussed in Thread: ADS1131, ADS1262, ADS1261

Hi,

On a weight scale product based on ads1232 reference design I did an experiment to reverse the polarity of excitation and have measures from 0 up to 50 Celcious. I did this because I wanted to see how the system behaves in these temperature ranges so as to upgrade the product to use ac excitation.

What the measures shows, is that the drift caused by those temperatures is much smaller (about 1/3) compared to the measures in the same temperatures with normal polarity. Does it make any sense ?

Thanks a lot!

  • 0
    TI__Guru** 113765 points

    Hi Theodore,

    You should not see any difference due to just the change in input polarity.  Are your experiments using the same ADS1232?  Did you just swap the wires with everything else staying the same?  Can you share the data that you are seeing for both cases?  How are you applying the temperature for the drift experiments (oven, thermostream, etc.)?

    Best regards,

    Bob B

  • 0 in reply to Bob Benjamin
    Intellectual 420 points

    Yes.
    I use the same board, only swapping excitation wires and tests are by measuring and deep-sleep the system every 10 minutes. Its battery powered and the load cell is zemic L6E3
    First in the fridge for 4 hours after room temperature and then in the sun for another 4 hours.
    I have double check the results.
    At the same time I do the same measures with my old board which uses the ADS1131 and the same results!
    For the ADS1232 i exclude only 2 bits (from the 24bits)  with PGA=1.

    Here is what I get at reverse polarity:

    You will note that max deviation of elec weight at reverse polarity is about 16 points while at normal polarity is 38 points. (big difference as each elec point represents 12gramms of the 100Kgr load cell)

    -361 31
    -364 30
    -362 9
    -363 8
    -365 8
    -366 7
    -366 6
    -367 5
    -367 4
    -368 4
    -368 5
    -368 6
    -369 5
    -369 4
    -369 3
    -370 3
    -370 3
    -370 4
    -369 5
    -370 4
    -370 3
    -370 2
    -372 2
    -373 26
    -372 26
    -372 27
    -371 27
    -371 27
    -370 27
    -371 27
    -370 27
    -370 27
    -370 27
    -369 27
    -369 27
    -369 27
    -369 27
    -368 27
    -369 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -368 27
    -367 27
    -368 27
    -368 27
    -368 27
    -367 27
    -368 27
    -368 27
    -368 27
    -367 27
    -367 27
    -368 27
    -368 27
    -368 27
    -368 27
    -367 27
    -368 27
    -367 27
    -368 27
    -367 27
    -368 27
    -368 27
    -367 27
    -368 27
    -367 27
    -368 27
    -368 27
    -367 27
    -368 27
    -368 27
    -368 27
    -367 27
    -367 27
    -368 27
    -367 27
    -368 27
    -367 27
    -367 28
    -369 31
    -369 36
    -369 37
    -369 39
    -369 38
    -368 38
    -367 37
    -367 40
    -367 38
    -366 36
    -366 34
    -366 34
    -366 34
    -367 34
    -367 32
    -367 32
    -367 32
    -366 32
    -366 31
    -366 32
    -365 32
    -366 31
    -366 31
    -365 30
    -365 36
    -367 38
    -366 33
    -365 38
    -365 38
    -364 38
    -365 37
    -365 37
    -365 38
    -366 37
    -366 37
    -365 38
    -366 36
    -364 35
    -366 45
    -363 41
    -365 39
    -365 36
    -364 35
    -365 45
    -365 46
    -365 46
    -365 45
    -365 47
    -366 38
    -366 38
    -365 38
    -364 46
    -363 47
    -366 46
    -363 42
    -364 26
    -364 14
    -364 13
    -364 12
    -365 11
    -365 9
    -366 9
    -367 8
    -367 8
    -367 7
    -367 7
    -368 7
    -369 7
    -370 3
    -370 -1
    -371 -2
    -373 -2
    -374 -2
    -375 -2
    -376 -2
    -377 -2
    -379 -2
    -377 -2
    -378 -2
    -376 -2
    -377 -2
    -380 -2
    -380 -2
    -381 -2
    -381 -2
    -380 -2
    -381 -2
    -381 -2
    -381 -2
    -381 -2
    -381 -2
    -377 -1
    -377 -2
    -380 -2
    -380 -2
    -381 -2
    -381 -2
    -380 -2
    -381 -2
    -381 -2
    -381 -2
    -381 -2
    -381 -2
    -374 6
    -375 6
    -374 6
    -374 6
    -374 6
    -374 6
    -374 5
    -374 5
    -374 5
    -374 5
    -374 5
    -374 5
    -374 5
    -374 5
    -374 5
    -374 4
    -374 5
    -374 5
    -374 5
    -374 5
    -374 4
    -374 4
    -375 4
    -374 4
    -374 4
    -374 4
    -374 4
    -374 4
    -374 3
    -374 3
    -374 3
    -374 3
    -374 3
    -374 3
    -374 3
    -374 3
    -375 3
    -374 2
    -374 3
    -375 2
    -374 2
    -375 2
    -374 2
    -374 2
    -374 2
    -374 2
    -375 2
    -374 2
    -374 2
    -374 2
    -374 2
    -375 2
    -374 2
    -373 25
    -372 26
    -372 27
    -372 27
    -371 28
    -371 28
    -371 28
    -370 28
    -370 29
    -369 28
    -369 29
    -368 29
    -368 29
    -368 29
    -368 29
    -368 29
    -367 29
    -367 30
    -367 30
    -367 30
    -367 30
    -367 30
    -366 30


    Here is what I get at normal polarity:
    elec weight -  temperature
    -373 30
    -376 30
    -370 2
    -364 0
    -360 -1
    -356 -1
    -354 -1
    -352 -2
    -350 -2
    -348 -2
    -347 -2
    -346 -2
    -345 -2
    -345 -2
    -345 -2
    -345 -1
    -346 -1
    -346 -2
    -345 -2
    -344 -2
    -344 -2
    -371 34
    -374 39
    -379 42
    -383 41
    -383 41
    -384 41
    -383 40
    -382 37
    -381 36
    -381 35
    -380 35
    -379 34
    -380 35
    -378 33
    -378 34
    -378 33
    -377 30
    -376 29
    -375 28
    -375 27
    -373 27
    -373 26
    -373 26
    -373 26
    -372 26
    -372 26

  • 0 in reply to theodore belogiannis
    TI__Guru** 113765 points

    Hi Theodore,

    How are you determining the temperature of the ADS1232?  Also, the numbers you have provided do not tell me much.  It would be much easier for me to determine what is going on by seeing the raw codes from the ADC as opposed to a computed value. 

    For these tests, is there any load on the load cell?  Or is this a 0 load condition?  Also, did you ever initiate the offset calibration after you power-up the ADS1232?  You should always initiate an offset measurement if you are not chopping the input to remove the offset.

    Have you determined what the offset of the load cell is?  If the load cell has an offset, when you swap the excitation wires you have also reversed the offset.  Some of the 'drift' affect differences you are seeing may be related to load cell offset.

    The ADS1232 offset drift is not determined by the polarity of the load cell input or excitation.  The ADS1232 offset drift is relative to the drift of the PGA. Here you would short the inputs together within the proper input common-mode and then cycle your temperature to determine the ADS1232 offset drift. 

    One further item is possibly a typo.  In your post you said you were using a gain of 1, is that correct?  A gain of 1 bypasses the PGA and lowers the input impedance of the ADS1232.  As the output of the load cell is very small, it is difficult to use a gain of 1 to determine the performance.

    Best regards,

    Bob B

  • 0 in reply to Bob Benjamin
    Intellectual 420 points

    Hi Bob,

    How are you determining the temperature of the ADS1232?  Also, the numbers you have provided do not tell me much.  It would be much easier for me to determine what is going on by seeing the raw codes from the ADC as opposed to a computed value. 

    I use an external temperature sensor. These are raw codes from the ADC and after the 'space' is the ambient temperature

    For these tests, is there any load on the load cell?  Or is this a 0 load condition?  Also, did you ever initiate the offset calibration after you power-up the ADS1232?  You should always initiate an offset measurement if you are not chopping the input to remove the offset.

    There is no load on the load cell. ) load condition. I never initiate offset calibration of the ADS1232. 

    Have you determined what the offset of the load cell is?  If the load cell has an offset, when you swap the excitation wires you have also reversed the offset.  Some of the 'drift' affect differences you are seeing may be related to load cell offset.

    For sure. I have also asked from the load cell manufacturer for their opinion

    The ADS1232 offset drift is not determined by the polarity of the load cell input or excitation.  The ADS1232 offset drift is relative to the drift of the PGA. Here you would short the inputs together within the proper input common-mode and then cycle your temperature to determine the ADS1232offset drift.

    One further item is possibly a typo.  In your post you said you were using a gain of 1, is that correct?  A gain of 1 bypasses the PGA and lowers the input impedance of the ADS1232.  As the output of the load cell is very small, it is difficult to use a gain of 1 to determine the performance.

    Yes I use gain of 1 so as to reduce noise and the analysis is accepted. 

    Regards

    Theodore

  • 0 in reply to theodore belogiannis
    Intellectual 420 points

    Do you think that reversing polarity will cause any unexpected issues as I'm planning to supplying weight scales with this setup from now on.

  • 0 in reply to theodore belogiannis
    TI__Guru** 113765 points

    Hi Theodore,

    Can I ask what you hope to gain by chopping the excitation polarity using AC excitation?  You may not get the results you are expecting relative to the effort involved.  AC excitation will only remove the error and offset of the load cell.  It will not affect the ADS1232 offset which most likely dominates.

    I have concerns about your measurement and method.  In both polarity you show very similar numbers.  You said you see "the same results" for both the ADS1131 and the ADS1232.  What does this mean?  The ADS1131 is fixed gain of 64 and 18 bits.  The ADS1232 is gain of 1 and 24 bits.

    You tell me the values given are the raw ADC codes, but earlier you stated you drop 2 bits from the ADS1232 results.  So a few things are not adding up.  The lowest noise for the ADS1232 is when gain is applied.  This seems to be backwards, but input referred noise actually lowers with gain.  It is true that more codes appear to be noisy, but the value of one code changes in resolution.  Note the noise tables on page 5 of the ADS1232 datasheet and how the peak-to-peak noise lowers as gain is increased.  So your tests should be using a gain of 128 if you want the lowest noise.

    When you say you "deep-sleep the system every 10 minutes" does this mean you are powering off the ADS1232 or using the PDWN pin to put the ADS1232 in power down mode?  What about the excitation?  Is this being turned off?

    When you bring your system out of "deep-sleep" what is your start-up process?  I would highly recommend that you issue an offset calibration at ADS1232 power-up after all supplies have stabilized and the reference has reached a settled voltage.  The offset calibration is started when the ADS1232 receives 26 or more SCLKs.  This will place any ADS1232 offset at the level of noise of the conversion.  Any offset measured will be the no-load offset of the load cell.

    If you issue the offset calibration each time you bring the ADS1232 out of "deep-sleep" the temperature offset and drift of the ADC should be removed.  You may see that this has much more value than using the AC excitation which requires multiple measurements and time for input settling.

    I would highly recommend reading the TI Design using the ADS1262 and AC excitation.

    http://www.ti.com/tool/TIPD188

    You may also want to take a look at the ADS1261 which has built-in GPIO control for using AC excitation as well as input chopping to remove PGA offset.

    Best regards,

    Bob B

  • 0 in reply to Bob Benjamin
    Intellectual 420 points

    Can I ask what you hope to gain by chopping the excitation polarity using AC excitation?  You may not get the results you are expecting relative to the effort involved.  AC excitation will only remove the error and offset of the load cell.  It will not affect the ADS1232 offset which most likely dominates.

    It is just that the application is a weight scale on a beehive which means that there is steady load in outdoor conditions. Temperature correction will be a good feature I think.

    I have concerns about your measurement and method.  In both polarity you show very similar numbers.  You said you see "the same results" for both the ADS1131 and the ADS1232.  What does this mean?  The ADS1131 is fixed gain of 64 and 18 bits.  The ADS1232 is gain of 1 and 24 bits.

    My mistake. I mean same deviations.

    You tell me the values given are the raw ADC codes, but earlier you stated you drop 2 bits from the ADS1232 results.  So a few things are not adding up.  The lowest noise for the ADS1232 is when gain is applied.  This seems to be backwards, but input referred noise actually lowers with gain.  It is true that more codes appear to be noisy, but the value of one code changes in resolution.  Note the noise tables on page 5 of the ADS1232 datasheet and how the peak-to-peak noise lowers as gain is increased.  So your tests should be using a gain of 128 if you want the lowest noise.

    I only exclude 2 bits from raw data. I thought that the lower the gain the better results. I didn't look at the datasheet. 

    When you say you "deep-sleep the system every 10 minutes" does this mean you are powering off the ADS1232 or using the PDWN pin to put the ADS1232 in power down mode?  What about the excitation?  Is this being turned off?

    Yes. Both PWDN and excitation using the analogue switch.

    When you bring your system out of "deep-sleep" what is your start-up process?  I would highly recommend that you issue an offset calibration at ADS1232 power-up after all supplies have stabilized and the reference has reached a settled voltage.  The offset calibration is started when the ADS1232 receives 26 or more SCLKs.  This will place any ADS1232 offset at the level of noise of the conversion.  Any offset measured will be the no-load offset of the load cell.

    If you issue the offset calibration each time you bring the ADS1232 out of "deep-sleep" the temperature offset and drift of the ADC should be removed.  You may see that this has much more value than using the AC excitation which requires multiple measurements and time for input settling.

    I had underestimated the value of offset calibration of the ads1232. I will start working on this and I will be back for results when ready.

    Thanks a lot for your help Bob.

    Theodore

  • 0 in reply to Bob Benjamin
    Intellectual 420 points

    Can I ask what you hope to gain by chopping the excitation polarity using AC excitation?  You may not get the results you are expecting relative to the effort involved.  AC excitation will only remove the error and offset of the load cell.  It will not affect the ADS1232 offset which most likely dominates.

    It is just that the application is a weight scale on a beehive which means that there is steady load in outdoor conditions. Temperature correction would be a good feature I thought.

    I have concerns about your measurement and method.  In both polarity you show very similar numbers.  You said you see "the same results" for both the ADS1131 and the ADS1232.  What does this mean?  The ADS1131 is fixed gain of 64 and 18 bits.  The ADS1232 is gain of 1 and 24 bits.

    My mistake. I mean same deviations.

    You tell me the values given are the raw ADC codes, but earlier you stated you drop 2 bits from the ADS1232 results.  So a few things are not adding up.  The lowest noise for the ADS1232 is when gain is applied.  This seems to be backwards, but input referred noise actually lowers with gain.  It is true that more codes appear to be noisy, but the value of one code changes in resolution.  Note the noise tables on page 5 of the ADS1232 datasheet and how the peak-to-peak noise lowers as gain is increased.  So your tests should be using a gain of 128 if you want the lowest noise.

    I only exclude 2 bits from raw data. I thought that the lower the gain the better results without looking at the datasheet.

    When you say you "deep-sleep the system every 10 minutes" does this mean you are powering off the ADS1232 or using the PDWN pin to put the ADS1232 in power down mode?  What about the excitation?  Is this being turned off?

    Yes. Both PWDN and excitation using analogue switch.

    When you bring your system out of "deep-sleep" what is your start-up process?  I would highly recommend that you issue an offset calibration at ADS1232 power-up after all supplies have stabilized and the reference has reached a settled voltage.  The offset calibration is started when the ADS1232 receives 26 or more SCLKs.  This will place any ADS1232 offset at the level of noise of the conversion.  Any offset measured will be the no-load offset of the load cell.

    If you issue the offset calibration each time you bring the ADS1232 out of "deep-sleep" the temperature offset and drift of the ADC should be removed.  You may see that this has much more value than using the AC excitation which requires multiple measurements and time for input settling.

    I had underestimated the value of offset calibration of the ads1232. I will start working on this and I will be back for results when ready.

    Thanks a lot for your help Bob.

    Theodore