DP83867IR: [DP83867IRRGZ]Only one board works at 100/1000Mbps but not at 10Mbps.

Hi Satoshi,

Glad to hear that you are using our DP83867IR in your design! Do you have register access to the PHY? If so, could you compare Registers: 0x0-0x1F, 0x6E, and 0x6F; between a working and non-working board? Would want to check if there is a register setting that is preventing 10Mbps for whatever reason. Reg 0x4 decides if the device allows 10/100 speed communication, and register 0x6E and 0x6F are the strap registers, which will confirm what mode the PHY is configured in.

Regards,

Alvaro

Sorry. There is an error in the incident information.
The incident is as follows.

Incident:
In a ping connection test with a PC,
there is no link-up at all at a speed of 100Mbps.
It takes a long time to link up at 10M/1Gbps.

Analysis status:
・No abnormalities in the signal between CPU and PHY
・No abnormalities in the PHY clock
・No abnormalities in the PHY chip and surrounding components as far as visual inspection is concerned
・No abnormalities in the conductivity of the pulse transformer built into the connector
・No abnormalities in the communication LED lighting/blinking

Confirmation contents:
・Is it likely that the device is defective?
・Are there any factors such as a small design margin?

Hi Satoshi,

Thank you for clarifying the problem, but could you still provide me with the register reads I request in my previous reply?

Regards,

Alvaro

Hi,
I develop that prototype board with Satoshi.

Here are register values.

When PC set force 100Mbps, board PHY is linked down, and register 0x11 is repeated different values:
 - 0x0000,0x0002,0x0300,0x0302,0x1002,0x1302,0x6302,0x7002,0x7302.
It seems to me PHY repeats negotiation process.


Equipement:
 - Magnetics/RJ-45 : L829-1J1T-43 (bel)
 - Cable Length : 3m, unshielded
 - Test connecttion to Windows10 PC.
 - Force Speed/Duplex in Windows10 device property.


Could you advise what I need to do next ?

RegisterValue.xlsx

Hi User,

Thank you for providing the register table, especially for formatting it so neatly. The only register that seemed odd is Board A's Register 0x13 in the Force 100Mbps case, and of course Register 0x11 as you mentioned. Instead of forcing speed on the PC side, can we leave the PC in auto-negotiation and change the speed from the PHY side?

To be clear, we want to keep auto-negotiation on in both the PC and the DP83867. To "force" the speed, we will be de-advertising their abilities. Reg 0x4 contains the PHY's 10/100 speed advertisements and Reg 0x9 contains the gigabit. For example, writing Reg 0x9 = 0000 will disable gigabit capabilities with the 867, so it should link up at 100.  

Just to confirm, when you read register 0x6E and 0x6F, did you follow the extended register space process?

Regards,

Alvaro

I am sorry for the late response.


I tested to force speed from PHY side.
 - PC : Enable auto-negotiation.
 - PHY : Set register 0x09 and 0x04.

It results still PHY link down at 100Mbps.
Refer excel.


I noticed different value of 'Link Quality MSE Registers'(in document SNLA246C) between board A and B.
That registers value of board A are larger than board B.
Another board C is about the same as board B.

Should I try replacing Magnetics/RJ-45 of board A ?
Or is it possible that PHY Exposed Pad is not enough solder jointed ?

As for confirmation, the answer is yes.
I followed the extended register space process to read 0x6e and 0x6f.

ForceSpeed_LinkQuality.xlsx

Hi User,

I did not get a chance to check this today, please allow me another day to respond.

Regards,

Alvaro

Hi User,

Thank you for the register log and again for formatting is so clearly. The MSE values you found is a good debug step, but even though they were slightly larger, it is not enough to explain this issue in my opinion.

Changing the magnetics could be a good change, but instead could we change the PHY itself? Take the non-working PHY_A (from Board A) and place it on Board B (and vice versa, take PHY_B and place it on Board A). This will let us know if the issue is with the part or with the board.

Regards,

Alvaro

Hi,
Thank you for for advice.

First, I try to change PHY of board A with a new one.

I'll try to place PHY_A on board_B next.


I would to like place working PHY_B on board_A following your advice if possible, but I can't now make working board_B to not-work.

Hi User,

Take your time and let me know when you have an update.

Regards,

Alvaro