DS280DF810: The text issue for DS280DF810 CDR

Part Number: DS280DF810

Hi,

We use the DS280DF810 for our server project, and now, our 100G card cannot be linked using the AOC cable. I measured the CDR receiving eye diagram and found that it didn't meet the standard. I would like to ask TI to confirm if it is caused by the equalization process. The eye diagrams of the DS280DF810 receiver at the test points of the DAC cable and the AOC cable have been tested, and the problem has also been explained in the below document:

48_TI_Eye_Data_AOC_OverEqualization_EN.xlsx 

 

  • Hi Zhang,

    1. I think that over-equalization is a valid concern.  The reason why AOC would show over-equalization is that it would typically be transmitting a clean electrical signal.

    Question: Have you tried using an optical module?  If so, do you observe similar behavior.  If over-equalization is the root cause, you should observe similar behavior for all signal sources that provide a clean electrical signal.

    I see you're also trying to sweep TX FIR settings while using AOC.

    Question: Does the AOC have a CDR?  Can you share details on the AOC signal chain?  If AOC has CDR, sweeping TX FIR settings may not be effective.

    Question: I'm wondering if AOC may have some TX emphasis settings.  Is it possible to check and ensure all TX emphasis is disabled on AOC?

    2. Regarding retimer suggestions for low loss channels with AOC/optical modules, there are a few suggestions:

    • Enable DFE (and use adapt mode 2)
      • Can also try adjusting the start polarity for DFE tap 1.  It might attenuate if over-equalization is an issue.  Set 0x12[7]=0.
    • Following sequence can help with low loss cases:
      REG        Value    Mask     Comment           

      13           04           04           //Set eq_limit_en=1 to enable CTLE limiting mode

      1A          00           08           //Reduce CTLE bias current, to reduce applied CTLE boost

    Thanks,

    Drew

  • Hi Drew,

    Question: Have you tried using an optical module?  If so, do you observe similar behavior.  If over-equalization is the root cause, you should observe similar behavior for all signal sources that provide a clean electrical signal.

    We use the Optical module+Optical Fiber, and our end customer are also use the Optical module and they still don't link up;

    Question: Does the AOC have a CDR?  Can you share details on the AOC signal chain?  If AOC has CDR, sweeping TX FIR settings may not be effective.

    The optical module uses JPC: P59HN0ERCAS1-1, and I am currently unable to confirm if the optical module contains CDRs;

    Regarding retimer suggestions for low loss channels with AOC/optical modules, there are a few suggestions:

    • Enable DFE (and use adapt mode 2)
      • Can also try adjusting the start polarity for DFE tap 1.  It might attenuate if over-equalization is an issue.  Set 0x12[7]=0.
    • Following sequence can help with low loss cases:
      REG        Value    Mask     Comment           

      13           04           04           //Set eq_limit_en=1 to enable CTLE limiting mode

      1A          00           08           //Reduce CTLE bias current, to reduce applied CTLE boost

    We have adjusted the values of the relevant registers as suggested by TI, but there is currently no improvement in the eye diagram.

    The below is the updated form for our eys pattern:

    eye_data_optical module-20261019.xlsx

  • Hi Zhang,

    I reviewed your results.  If I understand correctly, K1+K2 improved HEO, although there was a tradeoff with VEO in some cases.

    Question:

    If you are still observing link down, have you tried measuring BER?  You can use the retimer PRBS generator and checker to assist with this.

    Although retimer HEO/VEO are not optimal, it is often still possible to observe link up.  Looking at BER will help us better understand the issue.

    Thanks,

    Drew

  • Hi Drew,

    Please refer to below BER test data:

    eye_data_optical module-20261019(1).xlsx