• State
  • Replies 11 replies
  • Subscribers 33 subscribers
  • Views 138 views
  • Users 0 members are here
Support feedback
Options
Options
Related

DP83867IR: Link is unstable: False Carrier interrupt and XGMII Error interrupt (cont'd)

Atsushi Takahashi
Expert 1415 points
Part Number: DP83867IR

Tool/software:

Hi Team,

This post is a continuation of the issue in the post below.

 DP83867IR: Link is unstable: False Carrier interrupt and XGMII Error interrupt 

1) The customer tried scripts 3.1 and 3.2 in Troubleshooting Guide, but there was no improvement.

2) They wrote 2053h instead of 2024h to register 53h, but there was no improvement.

I received register information when the issue occur from the customer.

Would you be able to check it and give me some advice on how to improve it?

Best Regards,

  • 0
    TI__Genius 14116 points

    Hi Takahashi-san, 

    Feel free to share me the register information. 
    For Unit C, could they also share the register information for below?


    Best,
    J

  • 0 in reply to J
    Expert 1415 points

    Hi J-san,

    Thank you for your reply.

    I'll send you the register information of Unit A by private massage for the customer information.

    I don't have any information on Unit C yet, so I'll let you know as soon as I have it.

  • 0 in reply to Atsushi Takahashi
    TI__Genius 14116 points

    Hi Takahashi-san, 

    Please keep me updated. 

    Best,
    J

  • 0 in reply to J
    TI__Genius 14116 points

    Hi Takahashi-san, 

    Thank you for the message.
    The register information didn't show much that could point to an issue. 
    Could we also take a look at the schematic/layout?
    In addition, what is the link partner? If they use our EVM against the same link partner/cable, do they see the same issue?
    Lastly, if the customer could also provide the values of the MSE registers I discussed earlier, that would be great. 

    Best,
    J

     


  • 0 in reply to J
    Expert 1415 points

    Hi J-san,

    I sent you the schematic of unit A by private massage and the link partner part number.

    I have requested the MSE register and Unit C register information from the customer, so I will contact you as soon as I receive it.

    I’ll also check whether EVM can be used for this investigation.

  • 0 in reply to Atsushi Takahashi
    TI__Genius 14116 points

    Hi Takahashi-san, 

    Please let me know. 

    I will get back to you on the schematic feedback tomorrow. 

    Best,
    J

  • 0 in reply to J
    TI__Genius 14116 points

    Hi Takahashi-san, 

    Below are my comments on the schematic:
    1. We typically recommend 4700pF for C47 instead of 1000pF. 

    Besides that, have we suggested the customer to do 0x01D5 = 0xF508? This is meant to improve VOD swing in 1G operation. 

    Otherwise, the readings from MSE register would be very helpful tracking down the root cause. 

    Best,
    J

  • 0 in reply to J
    Expert 1415 points

    J-san,

    I'm sorry for the late reply.

    Thank you for your feedback on the schematic review.

    I don't suggest 0x01D5 = 0xF508 setting yet, so I will suggest that.

    I sent you the standard register reading value of unit C and MSE register reading value of unit A and unit C by private massage.

    Best Regards,

  • 0 in reply to Atsushi Takahashi
    TI__Genius 14116 points

    Hi Takahashi-san, 

    It looks like the signal quality is good on Unit A, but the link is not good in longer cables. 
    I noticed that the customer is advertising all speeds on the PHY for the auto-negotiation. 
    Is the switch also advertising all speeds? I wonder if they can even link up on lower speeds. 

    For Unit C, if they read the 0x0013 register twice, are the values same? I noticed that some values are read twice but 0x0013 register was not. This register is cleared on read so it should've been cleared to 0x0000. 

    For Unit C, it also looks like some amount of auto-negotiation is done, but it doesn't seem to resolve. 
    Would it be possible for the customer to disable auto-negotiation and force 1G full duplex link?

    Could the same be tried for unit A?

    Please let me know. 

    Best,
    J

  • 0 in reply to J
    Expert 1415 points

    Hi J-san,

    Thank you for your reply.
    I will check the switch's advertising status.

    You said;
    "For Unit C, if they read the 0x0013 register twice, are the values same? I noticed that some values are read twice but 0x0013 register was not. This register is cleared on read so it should've been cleared to 0x0000."
    They also read the 0x0013 register twice, and it seems to have shown the same value (0x1D40) the second time.

    You said;
    "For Unit C, it also looks like some amount of auto-negotiation is done, but it doesn't seem to resolve.
    Would it be possible for the customer to disable auto-negotiation and force 1G full duplex link?"
    I understand that auto-negotiation is required for 1Gbps.
    Will the vendor(TI) allow them to disable auto-negotiation and force it to 1Gbps?

  • 0 in reply to Atsushi Takahashi
    TI__Genius 14116 points

    Hi Takahashi-san, 

    Thank you for the detail. It looks like the interrupt is not going low. It is weird that this is not going low because some of the interrupts are related to MDI crossover/auto-negotiation complete/link status change. 

    Could they verify again if reading this register multiple times clears the read?

    To disable auto-negotiation, the customer can write 0 to bit 12 on 0x0000 register. 
    To select 1G full duplex, the customer can write 1 to bit 6 of 0x0000 register and write 1 to bit 8 on 0x0000 register for full duplex advertisement. 

    Best,
    J