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.

DP83822IF: Auto Negotiation always failed

Part Number: DP83822IF
Other Parts Discussed in Thread: AM3356, TLK110, SYSBIOS

Hi,

We used two DP83822 to connect to PRU_MII0 and PRU_MII1 of AM3356 for PROFINET.Our problem now is one can work and one can't.
One PHY's address is 9 and the other one is 1.(Address is 9 ,RX_D2 is pullup).
The two PHY reset signals are connected together.The interesting part is that the PHY 9 always works fine(the link light is on),but PHY 1 always unable to work(the link light is off) after hardware reset .
Register 0x0001 of PHY 9 is 0x786D,Register 0x0001 of PHY 1 is 0x7849.Obviously, Auto Negotiation always failed for PHY 9.
I switched the addresses of two phys, the PHY 1 (Address is 1) always unable to work.
I think RX_D2 pullup is the key point.
It has something to do with Hardware Bootstrap Configurations.I never understood this part.
The ICEv2 board used the TLK110.Now it's replaced by DP83822.Is there anything I need to pay attention to?
Please give me some guidance.

Regards,

Andy Zhou

  • Hi Andy,

    Kindly read and share SOR register ( 0x467, 0x468).These register latches strap values and can give some insight on the issue.

    Regards,
    Geet
  • Hi Geet,
    0x0467 Strap Latch-In Register #1 (SOR1) =0x2001,
    0x0468 Strap Latch-In Register #2 =0x0000.
    The two PHYs have the same value.
    There is an obvious problem here,The value of the register is not consistent with the peripheral hardware configuration.Only the RX_D2 of PHY9 is pulled up with 2k ohm.Everything else is default.
    The strangest thing is that the PHY 9 always work fine and the value of this address is the same for the two phys.
    You need to know that the link light of PHY1 is off after hardware reset.This serial managemnet interface(MDC MDIO)has not been configured.
    I have no idea now, please give me some hints!
    A few questions by the way.
    Will you stop the production of TLK110,Is there a specific time?
    Is there a reference design for AM335x and DP83822.If not, I suggest you should do this,if want to use DP83822 instead of TLK110 for industrial communication, eg. PROFINET,EtherCAT.

    Regards,

    Andy Zhou
  • Hi,

    I missed to write in my last email that these register falls in extended space and you have to follow procedure mentioned in section 8.4.2.1 ( Extended Adress space access). Have you used it to read the SOR values.

    Given you mention that link light is off on PHY 9, can you check the RX_CLK pin has clock or not ? Just want to ensure Phy is coming out of reset.

    To answer your question on Industrial Communication reference design with DP83822, please refer to below TIDA reference design.

    www.ti.com/.../TIDA-00299 : It's been tested for industrial communication.


    On TLK110 v/s DP83822, we recomend all new designs to use DP83822.

    Regards,
    Geet
  • Hi Geet,

    I should have solved the problem.The culprit is that pr1_mii1_col is configured as output mode by isdk. PHY works fine when pr1_mii1_col is configured as input mode.

    Here's a screenshot of isdk(C:\ti\am335x_sysbios_ind_sdk_01.01.01.01\examples\profinet_slave_RT\appl_cnfg.h).I think it is a BUG in AM335x isdk !!!   The problem is also present in the latest isdk.

    By looking at the DP83822 manual, this should be the output pin of DP83822 and the input pin of AM3356.This pin is the same function on TLK and DP83822.

    This problem is not shown on the ICEv2 evaluation board and it should be related to this isolation circuit.It isolates pr1_mii1_col (AM3356 pin) and COL (DP83822 pin), but the two pins on my board are connected  directly.

    Is my analysis correct?Should pr1_mii1_col be set as input mode ? Please confirm.

    Regards,

    Andy Zhou

  • Hi Andy,

    Good to know you found the issue. Yes, this is COL pin used by Phy to indicate Collision in Half Duplex mode and shall be output from Phy. Also, this pin has strap function which may be getting impacted by wrong configuration .

    I will transfer the thread to Catalog Processor team to comment on SDK.

    Regards,
    Geet