DP83867IR: Failed to connect to 100M Cable and Reg0x0008=E00A

Hi Steve,

Thank you for the register dump. Could you provide more details as to what went wrong? Is our PHY dropping link, dropping packets, etc? Is this issue only observed with a 100M cable (is this shielded or unshielded)? Could we use a shorter cable to see if the issue persists?

Regards,

Alvaro

Hi Alvaro,

Thanks for your respond. Customer has successfully connected to the Cable (unshielded) with 10M and 1000M, only the 100M cannot link, the Auto-negotiate failed.  We also tried the shorter cable, it still failed to auto-negotiate.

Hi Steve,

Apologies for not replying today, I did not have time to review your query today, please allow me another day to respond.

Regards,

Alvaro

Hi Alvaro,

Here is customer's schematic. And I have found that the TX_EN/TX_CTRL pin is float, while datasheet is recommended to set as MODE 3/4. Is this caused the problem with 100M, but still work with 1000M？

Hi Steve,

In the register log that you sent me, it shows the DP83867 linking up at 100Mbps. I can see this by seeing Reg 0x1 = 796D, which means link is up, and Reg 0x11 = 7F02, which means it is operating at 100 Mbps. I'm sorry to have made you wait Steve but I don't believe I'm understanding the problem. 

Steve Sun said:

100M Cable

In your initial question you mentioned 100M cable, is this 100Mbps or 100 meter? I suggested using a shorted cable initially because I thought you meant 100 meters.

To confirm the strap settings, could you read Registers 0x6E & 0x6F? Please note these are extended registers, please see the following FAQ for how to access them.

https://e2e.ti.com/support/interface-group/interface/f/interface-forum/1271487/faq-extended-register-space-access-for-ethernet-phys 

Regards,

Alvaro

Hi Alvaro,

I mean 100Mbps, and we also try the shorter cable, but no use.

Steve Sun said:

We also tried the shorter cable, it still failed to auto-negotiate.

Regarding to the register value comparison between successful case (test board) and failed case (CNC board), does these different value have reasonable explanation if 100Mbps all linking up, especially reg0x0008=E00A)?

and the reg0x6e=0x0，reg0x6f=0x100

Hi Steve,

Thank you for clarifying, I misread the register table you clearly provided. Thank you for providing registers 0x6E and 6F, everything seems to be in order here.

The Test board is working as expected, but the CNC board is not. 

When the cable is unplugged, Reg 0x5 & 0x8 are expected to be 0. Once the cable is plugged, what the link partner claims it's capabilities are, are stored in these registers.

After the cable is plugged in, Reg 0x5 = CD81, which is expected. We can see that the Link Partner is capable of 100Mbps communication.

Who is the link partner during this test? Does the CNC board have a problem linking up with multiple different link partners, or does it only have a problem with a specific link partner?

Regards,

Alvaro

Hi Alvaro,

The PC is the link partner during this test, there is only one link partner, they modifies the communicate rate (10Mbps/100Mbps/1000Mbps) of PC to test. And as pre-tested, the CNC board can successfully link 1000Mbps, but only can not link 100Mbps.

Hi Steve,

Could you give me a register dump (similar to the one you provided) when the CNC board links up at 1000Mbps? Could you also send me a pdf of the full schematic, I want to check the MDI portion of the schematics. This is very strange.

Regards,

Alvaro

Hi Alvaro,

I have checked again with customer. Here is the register dump when CNC board links up at 1000Mbps, and also the reg0x6e=0x0，reg0x6f=0x100. Due to customer's confidentiality, even the screenshot can not be provided. Here are the pictures. Besides, I have checked the CNC and test board. The schematic is the same, but the PCB is a little different. The RJ45 interface, the transform and the PHY are on the one PCB board (the test board). But, in the CNC board, there is another interface adapter which the only the interface including RJ45 and transform are on it, the Ethernet signals are connected between the two boards by connector DT11-080S-10-T, as shown in the below.

I‘m not sure if the long PCB wires effect the signal quality and caused the issue. But the 1000MHz can successfully connect.

DP83867 _Issue.xlsx

DT11-080S-10-T.pdf

Hi Steve,

Thank you for the register dump and connector schematic. This connector configuration is strange, the trace lengths will definitely have an impact, but I would expect it to fail gigabit. Please allow me until the end of the week to review this.

Regards,

Alvaro 

Hi Steve,

I didn't catch anything suspicious from the register dump or the limited schematic photo. I'm skeptical about this DT11 connector, could you confirm if the customer is able to ping/transmit data when linked at 1000Mbps? 

Regards,

Alvaro

Hi Alvaro,

I'm sure that all is normal under 1000Mbps. Is that possible that the long signal wires effect the ethernet signal only in 100Mbps, not in 1000Mbps? Because the only different is the signal wire length including the DT11 connector between CNC and Test boards.

Hi Steve,

It is possible but I would expect the board to be affected at 1000Mbps and be okay at 100Mbps. Please confirm if successful communication is possible at 1000Mbps. Do they have a way to transmit packets and check how many were received/packet-loss?

Regards,

Alvaro 

Hi Alvaro,

We have confirmed again, they can transform packets and files successfully without any abnormal at 1000Mbps.

Thanks

Steve

Hi Steve,

Could we try enabling/disabling Mirror Mode? This can be done by writing to Reg 0x31[0].

When linked at 1000Mbps, could you read registers 0x225, 0x265, 0x2A5, & and 0x2E5? These registers read the link quality of each channel (A-D).

Please note that the above mentioned registers are Extended Registers and cannot be accessed directly. Please refer to this FAQ for more information:

https://e2e.ti.com/support/interface-group/interface/f/interface-forum/1271487/faq-extended-register-space-access-for-ethernet-phys

Regards,

Alvaro

Hi Alvaro,

Here are the values of the registers 0x31, 0x225, 0x265, 0x2A5 and 0x2E5.

Steve

Hi Steve,

Alvaro is OoO, and will be returning next week. Please expect delay in response.

Sincerely,

Gerome