• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 45 subscribers
  • Views 1023 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.

ADS1231: Delta sigma ADC for loadcell bridge application

Andi76
Prodigy 160 points
Part Number: ADS1231
Other Parts Discussed in Thread: ADS1230, ADS1232, ADS1120, ADS1131, ADS1220, LP5907, TPS732

Hi forum,

my application is a loadcell full-bridge and I try to find a suitable ADC (maybe ADS1231?)
What resolution do I really need for the ADC?

My loadcell accuracy class is 0,03% of FS, that means max. sensor resolution is about 12 bits, right?
In TI's AppNote "Delta sigma ADC design for bridge sensor applications" (SBAA184.pdf) there is a system accuracy calculation on page 3.
It says that I lose accuracy, because:

  • mismatch between ADC input range and loadcell output voltage range
  • ADC's linearity error
  • noise
  • etc.

24-Bit ADC ADS1231 gives an INL of 8ppm which translates to an effective resolution of only 16,93 bits without INL compensation, is that correct?

How to deal with the difference of 24 bits - 16,93 bits = 7,07 bits? Do I have to drop 7 bits for example by 7 times shift right to get error-free bits without compensating the INL?
Do I have to compensate for INL error, I'd prefer not to compensate.

Similar errors are input offset drift and gain drift that reduce resolution even more. How is this usually handled?

Thank you and best regards,

Andreas

  • 0
    TI__Guru** 113765 points

    Hi Andreas,

    It would be really helpful to know more about the load cell, in particular the load cell sensitivity in mV/V.  Also, it would be important to know the excitation voltage of the load cell.  All of these factors come into play when attempting to calculate a reasonable prediction of performance. 

    Also, it would be helpful to know the measurement resolution desired. For example if you have a 100kg load cell and want to measure to 1g resolution, then you would need a measurement resolution of 100,000 counts.

    I can tell you from my experience that the greatest factor you will encounter is noise.  As far as linearity, as the application note has stated "In practice, the nonlinearity bottleneck is the linearity of the bridge sensor, not the INL of the ADC."  The reason being is that the full-scale output of the load cell is much smaller than the full-scale range of the ADC where the measurement stays within a more linear region of the ADC.

    Offset and gain drift can be an issue.  Some devices, such as the ADS1230 or ADS1232 have the ability to self-calibrate offset.  Gain drift is usually not a large factor as gain error is usually insignificant with respect to the small portion of the full-scale range used.  Also, excitation drift as it relates to the measurement can be eliminated by using the excitation source as the reference source which allows the measurement to become ratiometric.

    Best regards,

    Bob B

  • 0 in reply to Bob Benjamin
    Prodigy 160 points

    Hi Bob,

    thank you for quick reply, here some more application data:

    • Load cell sensitivity is 1,8mV/V +/-0.1mV/V,
    • ratiometric measurement Uref,ADC = Ubridge = 5V (optional possible would be 3,3V),
    • load cell capacity 10kg, resolution 10g

    Best regards,

    Andreas

  • 0 in reply to Andi76
    TI__Guru** 113765 points
    Hi Andreas,

    If the ADS1231 is used as an example we find that with 5V excitation (and voltage reference) the full-scale range of the ADS1231 is 39mV (+/-19.5mV). With 1.8mV/V sensitivity and 5V excitation the maximum output voltage of the load cell is about 9mV. This means that only about 1/4 of the total FSR of the ADS1231 will be used. As noise will limit the total available counts from 24 bits to the noise-free bits, we can determine the noise-free bits from the noise tables in the datasheet for the desired data rate. For almost every case the noise-free bits will be approximately 16 bits, which is 65535 decimal counts for FSR. As only 1/4 of the total available will be used, then there would be about 16000 counts noise-free.

    Based on a 10kg capacity and a resolution of 10g, you will only need noise-free (or flicker-free) stable counts of 1000. So the ADS1231 should work well and go well beyond the desired resolution to perhaps 1g or maybe a little better (excluding external noise sources). Using the same design criteria you probably could use the ADS1131, or ADS1120.

    Take a look at the ADS1120 and the bridge application example on page 58 in section 9.2.3. as this example is very close to your design parameter. Another advantage is that this part can operate at higher data rates with good performance and also has the advantage of the low-side switch to save power when not measuring (similar to the ADS1231 and ADS1131). The ADS1x31 devices have the advantage in simple pin controlled setup where the ADS1120 requires communication to setup the device registers. Another advantage of the ADS1120 is if you require better performance you could use the pin compatible 24-bit device, the ADS1220.

    Best regards,
    Bob B
  • 0 in reply to Bob Benjamin
    Prodigy 160 points

    Hi Bob,

    thank you for this very good explanation. I like the ADS1120. Would the LP5907 be an adequate voltage regulator to supply bridge and ADC's VREF or what is your suggestion?

    Thanks for support and best regards,

    Andreas

  • 0 in reply to Andi76
    TI__Guru** 113765 points
    Hi Andreas,

    The LP5907 would be a good low noise choice, but the output is a maximum of 4.5V. That may or may not be an issue. I think the TPS732 device family would also work even though it has slightly higher noise.

    Best regards,
    Bob B