• State
  • Locked Locked
  • Replies 8 replies
  • Subscribers 83 subscribers
  • Views 783 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.

F2837xD sigma delta comparator offset error

jimj2713
Expert 2385 points
Other Parts Discussed in Thread: AMC1200

We have been using the F2837xD for almost one year now, but using the ADC inputs to measure current. I have recently converted over to Sigma Delta hardware and software, and I have it all working except the comparator thresholds. The data-path and comparator-path are configured using the same Sinc3 filters, the data-path uses DOSR=40 and the comparator-path used DOSR=32.

The problem is all 6 comparators require an offset equal to ~1675 counts of the 0x7FFF range (mid-point is ~18059 instead of 16384). I calibrate out any measurement offset from my current-sensors, and the Sigma Delta data-path measures all 3 phase currents accurately (so I know the values of the currents entering the comparator filters). The comparator filters are the same as the data-path, yet I need to add this ~1675 count offset to ALL 6 comparators to get them to trip at the correct threshold.

I have reviewed, and re-reviewed, my Sigma Delta register settings and I find no errors. Even if there was an error in the filter-setting or OSR/resolution, those errors would cause a linear/scalor error, yet my errors on all 6 comparators are simply offset -- with this 1675 count offset, all 6 comparators trip at an accurate trip point across a broad measurement range.

While there is some errata on the comparators for the pre-released silicon, there is nothing about a comparator offset. Do you know if the rev-A silicon has any known problem in this area? Do you have any suggestions on how to troubleshoot this?

Thanks,

Jim

  • 0
    Expert 2385 points
    I received an email that Fulano de Tal replied to this question, but it is not shown here; he wrote:
    Unlikely, but consider that they have a different gnd reference electrically.
    Is your ADC offset-calibrated and temperature-compensated with OTP values?

    To that response, I am confused. I thought the ADC was completely independent from the sigma-delta peripheral. To re-state my question, the sigma-delta comparators trip at a consistent (1675 count) offset from the sigma-delta data-path. Will the ADC-offset calibration (described in the "ADC Calibration" section of the datasheet) affect the sigma-delta comparator outcome?

    Also which ground-references are different? different between the ADC and SD, or different between the SD comparator and data-paths?

    Thanks,
    Jim
  • 0 in reply to jimj2713
    TI__Guru 55700 points

    Jim,

    The comparator filter and data filter differ in the way they handle input data. 

    The comparator filter unit translates a low input signal to '0' and high input signal to '1'  whereas

    The data filter unit translates a low input signal to '-1' and high input signal to '1'. So, this offset is inherent in comparator and data filter design.

    Regards,

    Manoj

  • 0 in reply to Manoj Santha Mohan
    TI__Guru 55700 points
    Yes, ADC and SDFM are completely independent of one another.
  • 0 in reply to Manoj Santha Mohan
    TI__Guru 55700 points
    Jim,

    Please close the thread if your queries got answered.

    Regards,
    Manoj
  • 0 in reply to Manoj Santha Mohan
    Expert 2385 points
    I believe my question was misunderstood. I understand the data-input differences between the data unit and the comparator unit. Let me restate my problem. With my chosen DOSR and OSR, the resolution of my data unit is 64000 and the resolution of my comparator unit is 32768. If I measure 0-current, my data unit correctly produces an output count of 0. What should happen is my comparator unit should trip at about mid-scale of 16384. What I have found, is that the comparators have an offset of up to 1675 counts from the ideal. Also the 6 comparators (hi/lo for 3 phases) each have a different offset. One board varied from ~900 to 1675 counts.

    Because my data-paths measure accurately (over the entire range), I would expect the comparator paths to be equally accurate, but my temporary solution is to measure the offset of each of the 6 comparators in production, and save those offsets permanently.

    In all of the register settings I have found, a mistake in this area would affect the slope (input vs output), but these offsets are additive. I have not found any register settings that would affect/create a comparator offset.

    Thanks,
    Jim
  • 0 in reply to jimj2713
    TI__Guru 55700 points

    Jim,


    SDFM is a digital peripheral. I don't expect any offset if the data channels are fed with the same input.


    Are you feeding the same sigma delta modulator output to all the channels? (or) do each channel have different sigma delta modulator connected?


    Regards,

    Manoj

  • 0 in reply to Manoj Santha Mohan
    Expert 2385 points

    Manoj,

    We have individual AMC1200 isolation amplifiers connected to each sigma delta channel. The current is measured across a shunt-resistor -- basically the same as the AMC1200 app-note diagram.


    I agree about this being a digital peripheral, so no offset is expected, but I have seen this on multiple boards now, so I am applying a production calibration to subtract off each comparator's offset. Could it be related to the fact that the data-path sinc3 filters are reset each PWM cycle (using FILRES) and the comparator filters cannot be reset? That would still not explain why each of the 6 comparators (hi and low for 3 phases) has a different offset.

    Thanks,

    Jim

  • 0 in reply to jimj2713
    TI__Guru 55700 points
    AMC1200 doesn't seem to provide modulated clock and modulated data output. This clock data pair should be fed to F2837xD SDFM channels. You CAN'T fed analog input directly to SDFM.