• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 2 answers
  • Subscribers 49 subscribers
  • Views 179 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.

INA823: Gain Error vs Temperature

Adriana Terman
Prodigy 260 points
Part Number: INA823

Tool/software:

Hello Team, 

  • In Table 7.5, the specified Gain Drift value does not include the effects of the external gain-setting resistor Rg. Is this also true for Figure 7-18?
  • To minimize gain drift due to temperature, do we agree that the Rg drift should be matched to the drift of the internal 50 kΩ resistors? Therefore, should Rg be as close as possible to 5 ppm/°C (according to paragraph 8.3.1.1)

Kind regards, 

Adriana Terman

  • 0
    TI__Mastermind 32951 points

    Hello Adriana,

    Since the Rg can be selected by the user, and will directly affect the GE drift of the device, the datasheet benchmarks only the error terms influenced by the device. However, the direction of the GE drift is not determinate, and usually the RG will not have a determinate drift direction. Since these are random uncorrelated errors, the typical combined error will be the root sum square of the two drifts. Therefore, having a resistor drift of 25ppm/degC will combine with the INA823 drift of 5ppm/degC to yield a drift of 25.49ppm/degC. Whereas a 5ppm/degC Rg will yield a combined typical drift of ~7.07ppm/degC.

    To get the best performance, the drift of the resistor should be less than the drift/3, which means that the RSS will introduce less than 10% error. However, looking at typical resistor drifts on Digikey, the options are 2ppm/degC or foil resistors that are more expensive than the INA823, which may not be practical.

    Best,
    Gerasimos

  • 0 in reply to Gerasimos Madalvanos
    Prodigy 20 points

    Hello Gerasimos, 

    Thank you for these clarifications. As I was the one who asked Adriana these questions, I would like to elaborate on your response.

    We intend to integrate the INA823 for Wheatstone bridge signal amplification. It will be set with G=318, thus RG=316 Ohm. According to the datasheet, for G>1 the maximum GE drift is 35ppm/°C. By using an RG of 10ppm/°C, the config is optimal. Is this correct?

    You wrote that "the direction of the GE drift is not determinate", but as shown in figure 7-18, GE drift has a tendency to increase while the temperature increase (I am aware that this is averaged over 120 units). 
    Do you know, if each component tend to have a very different GE drift, or they tend to follow approximately the same curve? 
    If it's the second option, we could do a temperature compensation. 

    Regards, 
    Maximilien L 

  • 0 in reply to Maximilien Louis
    TI__Genius 15400 points

    Hi Maximilien, 

    The total drift compensation would be the root-sum square of the GE drift of the device and the resistor drift as Gerasimos previously explained. So, with the maximum drift of the INA and the RG of 10ppm/C, the total contribution would be 36.4ppm/C. 

    Please note that while the typical plot shows a large sample size, there is expected variance of behavior from lot-to-lot, so the direction always increasing for every device cannot be implied. Figure 7-18 is intended to help show the +/-3sigma (99.7%) distribution of parts that were tested to give more confidence of the gain error drift distribution.

    Does the customer have a concern over the gain error drift contribution? What error tolerance do they need to meet for their system? 

    Regards,
    Ashley

  • 0 in reply to Ashley Kang
    Prodigy 20 points

    Hi Ashley,

    Thanks for confirming the calculation for total drift contribution.

    Understood regarding Figure 7-18, you cannot guarantee the direction is the same for all lots.

    I haven't finalized the calculation yet, but we are concerned that excessive error propagation could lead to ADC input saturation in some devices. From my understanding, using RSS error is preferable as it represents a more realistic scenario than worst-case error, the latter being too conservative. 

    Regards,

    Maximilien

  • 0 in reply to Maximilien Louis
    TI__Mastermind 32951 points

    Hey Maximilien,

    RSS error does provide a more realistic error estimation. The max values typically represent statistically rare occurrences, and the worst case estimate takes the sum of multiple 5-6 sigma occurrences.

    While it is hypothetically possible to have 6 different 6 sigma events line up in the worst case scenario, it is statistically very unlikely.

    Best,
    Gerasimos