Post Go back to editing

AD9697: bit corruption after few minutes

Category: Hardware
Product Number: AD9697

Dear all,

I have a board with 3 AD9697 running at 1280 Sampling/second. The readout is done with a kintex ultra-scale (xcku035)

After initialization, I read samples and every thing is fine. Without signal, I got a baseline with an RMS of 500 uV:

The problem is that after some times (few minutes) I got this kind of patterns:

where clearly some bits are corrupted in ADC3, few also in ADC2 and none in ADC1.
If I redo the init procedure, everything is fine again.

There are few points to know:

- ADC3 is the farest to the FPGA (~12 cm) and closest to the clock generator.
- If I send patterns (ramp, midscale, full scale, custom, etc.) in the ADC, I do not see any corruption
- The PLL lock status bits are correct (ADC PLL-reg 0x056F and JESD PLL-reg 0x1262)
- Playing with pre-emphasis does not change anything
- Using scrambling or not gives exactly similar patterns

This latter point makes me think that the problem is inside the ADC before the serialization otherwise I would see different corruption patterns with or without scrambling.

I would like to know if this kind of behavior rings some bell to any of you...

Thank you very much for any advice.

Best regards

Marc

  • Hi  

    Thanks for your interest in AD9697.

    Your query is acknowledged. Kindly give time for the product owner to look into this and provide their response.

  • Thank you very much for your attention.
    Few more information:
    - The ADC is used for unipolar signals, so the 0 of the ADC corresponds to the lower rail
    - If I tune the baseline around 200 ADC counts, I don't get such a behavior, even after few hours of running.

    Could it be linked with the dithering ?
    Is it possible to turn it off ?

    Best regards

    Marc

  • Hello,

    These ADC's are designed for broadband/high frequency AC applications where AC metrics such as THD/SFDR as well as SNR are considered important.   As such.....they are meant to be driven with a differential input vs unipolar input that you mention above.  If you were still in the early design-stage, one would recommend a single-ended to differential ADC driver to interface with the AD9697 but it would appear that you are beyond that state. 

    The issue may (?) be related to the large single-ended voltage swing (relative to the ADC buffer supply of 2.5 V when trying to achieve full-scale for default setting of 1.59 Vpp.  Perhaps you can investigate if a setting of 1.36 Vpp improves the issue.

  • Dear PMH2023,

    Thank you from you answer. I will make a try reducing the dynamic range of the ADC.


    To be more clear, my signal is unipolar, which means that it is always positive but I feed the ADC with a differential signal by using ADA4927 with which I set the baseline near the lower ADC rail:

    Best regards,

    Marc

Before You Switch


Switching languages will make ADI Explorer unavailable. Resume your session by switching back to English and reopening ADI Explorer.