100Mbps link up issue

Hi

Awaiting for your reply.

Thanks and Regards
Sushruta

Hi Sushruta,

This is very interesting. When you say that the link oscillates between 10M and 1000M, where do you see this indication? Is it on the link led or from the status register of DP83867?

Your delays on RGMII should not impact your link establishment.

Can you use another link partner to establish 100M mode? Perhaps a 100M only switch? I want to take the PC out of the equation.

Also you can make the DP83867 handle the 100M autonegotiation by setting register 0x9 to 0x0 and restarting auto-negotiation.

Please share a schematic if possible.

Best Regards,

Hi Rob

Thank you for your reply.

1.

Rob Rodrigues said:

When you say that the link oscillates between 10M and 1000M, where do you see this indication? Is it on the link led or from the status register of DP83867?

 - Status register 

2.

Rob Rodrigues said:

Can you use another link partner to establish 100M mode? Perhaps a 100M only switch? I want to take the PC out of the equation.

-In this case, I do not see any link

Please find attached Schematics.

Thanks and Regards

SushrutaETHERNET ISSUE.pdf

Hi Sushruta,

Usually the link toggling is related to the RX_CTRL mode strap. I see you have RX_CTRL strapped as recommended, to mode 3. However, some MACs sometimes pull these lines high or low during initialization.

Please read register 0x6f. Please use the extended register access method shown in the datasheet section 8.4.2.1.4

Check to see if bit[8] in register 0x6f is set. If you are strapping RX_CTRL pin properly, then 0x6f.8 = 1

If 0x6f.8 = 0, please set 0x31.7 = 0 using extended register access method shown in datasheet section 8.4.2.1.3. After register 0x31.7 = 0, restart auto-negotiation using register 0x0.9 = 1. Total word written to register 0x0 = 0x1200

You can also try disabling EEE mode on your PC link partner.

Best Regards,

dp83867i_linux_dump.xlsHi Rob

Thank you for your support.

Please find attached register dumps.

We also tried doing the 0x0031 = 0x30 (7 bit set to 0) and it did not help.

Can you please let us know,what other things can be checked.

As this issue is very critical to us,I request you to reply at your earliest convenience.

Thanks and Regards

Sushruta

Hi Sushruta,

In the Linux dump you sent, register 0x6e shows bit[7] is set. This disable auto-negotiation of the DP83867. Auto-negotiation cannot be restarted. I think your RX_CTRL pin is being pulled high, and strapping to mode 4, which disables auto-negotiation.

Please isolate this pin from your MAC at power up. You should see the PHY able to auto-negotiate in that case. Also verify that the resistors R19 and R20 in your design are placed and of the correct value shown in your schematics.

Best Regards,

Hi Rob

1.

Rob Rodrigues said:

In the Linux dump you sent, register 0x6e shows bit[7] is set.

Please recheck the value in register 0x6e in the attached register dump sheet.

It is  0x2010; which means bit(7)=0; Auto-negotiation is enabled.

2.

Rob Rodrigues said:

Please isolate this pin from your MAC at power up.

Removed series resistor R18 in the path. Hence, RXCTL is not connected to MAC, but connected to strap resistors only. Still same issue persists.

From the register dumps, it seems to me as auto-negotiation is not complete(register 0x01 value). But not sure, why this is happening.

Please advise.

Thanks and Regards

Sushruta 

Hi Rob

Awaiting for your reply.

Can you please let us know, if there is any way to find out what could be going wrong?

Thanks and Regards
Sushruta

Hi Sushruta,

It is difficult to see what is wrong, so I must give you a large list of items to try to identify what may be happening here.

1. The magnetics used show the center taps shorted together. This is not allowed in the DP83867 as it may cause the NLP to become distorted and as such impact auto-negotiation.

2. Isolate the PHY from the MAC such that only power and MDIO control is applied to the PHY.

3. Remove the feedback resistor R53, it is not needed.
3.a. Ensure 27 pF is a good value for your crystal. With 27 pF, your crystal's required load should be approximately 18 pF in the datasheet.
3.b. Ensure the crystal ESR is 50 ohms or less.

4. Verify 1V is measured across RBIAS resistor R44 during normal operation.

5. Connect to a different link partner, I suggest using the DP83867 PHY1 on your board if available.

6. Verify power supply voltages are as expected.

Best Regards,

Hi Rob

Thank you for all your suggestions. It was really helpful.

Rob Rodrigues said:

a. Ensure 27 pF is a good value for your crystal. With 27 pF, your crystal's required load should be approximately 18 pF in the datasheet.

I changed the capacitors value to 18pF, and it is working now. However,  crystal load capacitance mentioned in the datasheet() is 18pF.

I am not sure, why it is not working for 27pF.

Attached us the board file snap shot at the crystal.

Thanks and Regards

Sushruta