• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 33 subscribers
  • Views 1627 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.

DP83867IS: interruption of the SGMII connection

T. Klein
Prodigy 50 points
Other Parts Discussed in Thread: DP83867IS

Hi,

I'm using two DP83867IS PHY as a repeater. I connected the SGMII interface of the first PHY over a backplane to the second PHY. This works fine.

To test the repeater, I'm using an Ethernet Tester to generate a high traffic load with random payload data. The ports of the tester are forced to 100 Mbit Full Duplex. The DP83867IS is configured for Auto neg. (SGMII enable is the only external strap setting, the other settings are default) . The 100 Mbit FD link is established successfully on both PHY. I send data through the repeater and loop the data back at the other side. So both directions are tested and working. However sometimes I discover that the connection is interrupted and no Ethernet data is transferred. I'm not able to reproduce this behavior, it happens randomly. Another problem is, that the SGMII connection is not reestablished. So only resetting the PHY will work. This can be reproduced sometimes if I interrupt the SGMII connection manually. In this error state the following data can be measured on the SGMII interface:

Phy 1 sends the following data (decoded 8b10b data stream):

K28.5+, B5- , 1-, 94-, K28.5+, 42-, 1+, 94+, K28.5-, B5+, 1+, 94+, K28.5-, 42+, 1-, 94- .....

Phy 2 sends the following data:

K28.5-, 50+, K28.5-, 50-...

After resetting the PHYs the connection is established again.

I tried to disable the SGMII auto negotiation. However this seems to be worsening the error. 

Do you have any suggestion what causes the interruption of the SGMII link? Is there any Ethernet package data which may causes a critically signal on the SGMII connection (e.g. long high, low sequences)?

How can I force the PHY to reestablish the link? Is there any loss of signal indication which I can use to generate a reset?

Thanks and best regards,

Thomas

  • 0
    TI__Mastermind 18425 points

    Hello Thomas,

    A few follow-up questions;

    1. Does the PHY loose link in between communication or does it sometimes not link at all?

    2. What are the trace lengths between PHY and RJ-45 connector for both PHYs?

    3. What are the SGMII trace lengths between both PHYs? Are they differential 100ohm traces?

    4. Cable length of both PHYs?

    5. Could you try to run the same tests but with the test ports in Auto-Neg mode instead of Force. You can still ensure that the PHYs link in 100M by disabling 1G advertisement.

    Write 0x0 to Register 0x9

    Write 0x1E1 to Register 0x4

    The PHY has Link LED which is de-asserted when PHY looses link. If LEDs are already used then GPIO pins can be used as LEDs. Refer to Registers 0x18, 0x19 and 0x172 for more information. Are you planning to use the loss of link indicator to prompt a microcontroller to reset the PHY?

    -Regards,
    Aniruddha

  • 0 in reply to Aniruddha Khadye
    Prodigy 50 points

    Hello Aniruddha,

    To your questions:

    1. The PHY does not loose the link to the Ethernet Tester. The register 0x1 of both PHY is 796D when the error occurs.

    2. The trace lengths from the PHY to the external magnetics are matched to 14.1 mm (variation +- 0.4 mm), from the magnetics to the RJ45 (15.5 mm +- 0.7mm)

    3. Yes, the traces are 100 ohms differential. Currently I am using 50 ohms coaxial cables with a length of 200 mm. A 100 ohm termination is placed between the SGMII_SIx data lines at the target.

    4. Cable length form the Ethernet Tester (VePAL TX300E) is 2 meters to each PHY.

    5. I tried this using the same setup and steps as mentioned above. I interrupt the SGMII connection manually between the PHY. This time there seems to be a reset of the PHY. Both ports of the Ethernet Tester negotiated to 1000 Mb FD speed. With forced 100 Mb speed on the Ethernet Tester the link remains up or I didn't see the down time.

    In this error state the first Phy sends data over SGMII:

    K28.5+, B5- , 1-, 94-, K28.5+, 42-, 1+, 94+, K28.5-, B5+, 1+, 94+, K28.5-, 42+, 1-, 94- .....

    The second PHY sends really strange signals (no LVDS??). Please find attached the oscilloscope prints. I got no response from both PHY at the MDIO interface. I can't read any registers in this state. After an reset its working again.

    The worst case that can happen to our application is when no communication is possible at all. So the reset of the PHY was the first idea to at least reestablish the communication.

    Best regards,

    Thomas

  • 0 in reply to T. Klein
    Prodigy 50 points

    Another way to reproduce the error is to connect the PHYs SGMII to a SFP module and establish the SGMII connection between the PHY via fiber optics. Now If you hot plug the SFP module a few times the connection is interrupted and will not be reestablished. However if you just unplug and reconnect the fiber optic cable, the error does not occur.

    I've attached the register settings of both PHY in the error state.

    /cfs-file/__key/communityserver-discussions-components-files/138/3750.is_5F00_dp_5F00_sfp_5F00_1.txt

    /cfs-file/__key/communityserver-discussions-components-files/138/4382.is_5F00_dp_5F00_sfp_5F00_2.txt

    Please advise how I can solve this problem.

    Thanks and best regards,

    Thomas