• State TI Thinks Resolved
  • Locked Locked
  • Replies 14 replies
  • Answers 7 answers
  • Subscribers 47 subscribers
  • Views 3679 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.

ADS8688A: about the precision of ADC

user4788502
Prodigy 50 points
Part Number: ADS8688A
Other Parts Discussed in Thread: OPA4180

Hi,

      I am going to build a data acquisition system(DAS) with a precision of less than 0.05%, I want to choose ADS8688A which has a 16 bits resolution. But in the file " tidu365b.pdf" a projest is provided, which shows that the accuracy over entire input range is 0.2%(much higher than 0.05%). So is there any other ADC recommended?

     I have another two questions about the file " tidu365b.pdf" .

     1 In  the file " tidu365b.pdf" , figure 26 to figure 31 show the test accuracy results, how is the FSR(full scale range) Error calculated ? For example, there are two groups of datas for an analog input, group one is tested by a multimeter, group two is the output of ADS8688A, so how can we calculate the FSR Error with the two groups of data ?

    2 In the figures we can see the  FSR Error is much smaller after calibration, so what is the exact method of calibration ?

figure 26 to figure 29 are shown bellow:

  • 0
    TI__Mastermind 45285 points

    Hello ,

    Thanks for your query on E2E, we have forwarded your questions to the author for the precision design (tidu365b.pdf), they should answer to you soon.

    Thanks&Regards

    Dale

  • 0
    TI__Intellectual 1035 points
    Hi,
    The system accuracy will be different from the ADC accuracy due to errors introduced by different elements in the signal path. In the above mentioned TI design, there are few resistors in the signal path, which contribute to the error for the entire range (+/-2.5V to +/-10V).
    The difference between the 2 sets of data is the "error". The percentage of "error" with respect to FSR is in the plot. The calibration was done as post processing, were the offset error and gain error were computed and compenstated.

    Sankar
  • 0 in reply to Sankar Sadasivam
    Prodigy 50 points
    Thank you for your reply. Is there any operational amplifier(low drift) that you can recomend for me to build the conditioning circuit(such as filter circuit) before the analog signal input the ADS8688A ?
  • 0 in reply to user4788502
    TI__Intellectual 1035 points

    You can look at OPA4180. Also suggest to look at the TI design TIDA-01214.

    regards

    Sankar

  • 0 in reply to Sankar Sadasivam
    Prodigy 50 points
    Hi Sankar,
    Could you please explain further how the calibration is carried out practically. For the ADS8688A how are the Gain and offset errors calibrated out. By looking at the design example i can only see that for calibration, all eight channels require a precision input voltage at the terminal blocks, surely this would not be practical for a production run of 1,000 boards.
    If we take the ADS8688A specification and ignore any external component errors i calculate the TUE as a maximum of 0.07% (RSS method), is it necessary to perform any calibration on this if the ADC accuracy alone is targetted to be 0.1%
    Best Regards,
    Martin
  • 0 in reply to Martin Thomas17
    TI__Intellectual 1035 points

    Hi Martin,

    If there are no other external components in the signal path or if we can ignore the external components errors, then TUE of ADC can be calculated by RSS method and that should be sufficient for estimating the signal chain performance. The TUE will depend on the input range setting and operating temperature range.

    If the signal chain performance meets your overall system performance requirement without calibration, that would be good in terms of less test time and cost. However if calibration is required, I guess you would need at least two measurements to perform calibration for a given operating temperature. The following is an example, which could be adapted according to your needs and results might vary accordingly. The first step would be to calibrate the offset. You can short the input and read a series of samples. The samples need to be averaged to reduce the effect of noise. The average value of the set gives the approximate offset. You need to subtract this offset value in all the subsequent measurements, which will be the offset calibration. Next you can calibrate the gain, by feeding a stable input which could be 0.9xFS input and also have a higher precision multimeter (10x higher precision than ADC) to measure the value that you are feeding. Now, read a series of samples and average them to get a value. The ratio of ADC average value and the multimeter reading should give the gain error. You can repeat this for different input values and temperature settings to get better accuracy for the entire input range and temperature range.

    After this intial calibration, the resulting residual error is called "linearity error".

    regards

    Sankar

  • 0 in reply to Sankar Sadasivam
    Prodigy 50 points

    Hi Sankar,

    Many thanks for that explanation. Could i please just expand on the requirement for calibration. If we are using six of the eight channels on the ADS8688A am i correct in assuming that we would need to carry out the offset and gain calibration for each of the six channels. Does the offset and gain error relate only to the ADC performance or does it include the errors from each PGA. Is it possible to use the spare two channels (7 and 8) on this ADS8688A to calibrate out the Gain and offset errors.

    Best regards,

    Martin

  • 0 in reply to Martin Thomas17
    TI__Intellectual 1035 points
    Hi Martin
    Calibration for all six channel might be needed depending on the PGA gain setting and the required system accuracy. As mentioned in the datasheet, the channel-to-channel error matching is comparable to channel offset. If this error might affect your system accuracy, you might need to do calbration for all six channels individually.

    regards
    Sankar
  • 0 in reply to Sankar Sadasivam
    Prodigy 50 points

    The document tidu540a on high accuracy AC voltage and Current measurement using the ADS8688A refers to Accuracy performance in section 7.9 with graphs (Figs 21-26) for FSR Error before and After Calibration. For this design example could you please tell me what the calibartion method was.

    For the hardware design in this example was the calibration carried out on each individual channel of the ADC to obtain these curves, did the calibration mechanism include both Offset and Gain error adjustment. I

    Best Regards,

    Martin

  • 0 in reply to Martin Thomas17
    TI__Intellectual 1035 points
    The calibration was done for one channel only and that included both offset and gain error calibration. I got this information from the author of that TI Design. I hope that clarifies you question.
  • 0 in reply to Sankar Sadasivam
    Prodigy 50 points
    Hi Sankar,
    I assume therefore that this "calibration factor" was then applied in software to the other seven channels.
    The document does not describe the method as to how the calibration is carried out in hardware. For the one channel that was used (do you know which channel was used) for the calibration of both offset and gain error was the calibration "stimulus" of 0v for offset applied at the terminals marke "Jn" on the schematic (Section 4.5), what voltage was used to calibrate the gain error.

    The AUX xhannel on the ADC has a voltage input of +2.5v applied to it via the two 47k resistors, is this also used in the calibration routine.

    Best Regards,

    Martin
  • 0 in reply to Martin Thomas17
    Prodigy 50 points

    Hi Sankar,

    Are you able to comment on my previous questions from Nov 20th

    Best regards,

    Martin

  • 0 in reply to Martin Thomas17
    TI__Intellectual 1035 points
    Had already mentioned that only one channel was calibrated. The calibration procedure is also mentioned in this thread. Please read it again.
    regards
    Sankar