DP83825I: RX of ethernet phy seems defective

The device seems active BTW:

[ 0.000000] psci: probing for conduit method from DT.
[ 4.811151] fec 30be0000.ethernet eth0: registered PHC device 0
[ 6.713884] TI DP83825S 30be0000.ethernet-1:00: attached PHY driver [TI DP83825S] (mii_bus:phy_addr=30be0000.ethernet-1:00, irq=POLL)
[ 12.837180] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
[ 12.837361] fec 30be0000.ethernet eth0: Link is Up - 100Mbps/Full - flow control off

Hi Danny,

What do you mean by the device seems to be defective sometimes? Could you elaborate more?

Is there link but no data transmission? Or, is it something else? How reproducible is this behavior? 

Only thing I noticed on the schematic is the 4.7ohm resistors on the MDI line which is not needed. Have you tried downgrading to 10M link instead of 100M? Also, does it work all the time if PoE is not enabled?

Best,

J

Hi J, 

I suspect a short circuit occurred on the primary side of the Ethernet transformer, causing a spike on the secondary side. This then fried the RX side of the IC. The Ethernet PHY can no longer connect to the network at all.

We've already tightened the TVS diodes a bit so they activate sooner, but this doesn't seem to solve the problem.

I'm having a hard time reproducing the problem myself because the defect was caused by a customer, and I don't know how it happened. However, there are several cases where this has occurred.

Regards, Danny

Hi Danny, 

Thank you for your information. Because it is hard for you to reproduce this problem, would it be possible for you to submit a failure analysis request via your FAE so we can systematically analyze what might have gotten wrong in the PHY? 

Also, I thought the link was up on the defective PHY from the MAC side. If this is the case, I wonder if RX side is fried since the auto-negotiation has to be done for the link to be up. 

Best,
J

Hi J, 

Yes you where right! the phy does work at 10Mbit's. Do you have an idea what the problem for 100Mbit could be?

Danny

Hi Danny,

Interesting that the 10M link is up and working, but 10M works.

Since this is the case, could you put back the PHY into 100M link and send packets onto the PHY and then check register 0x0015? This register counts any incoming packets with symbol errors. 

You can also put the PHY in reverse loopback mode and then send packets onto the PHY from the link partner and see if the PHY is able to send back the packets to the link partner without any issue. You can refer to the troubleshooting guide.

I would like to exclude any HW issue since the PHY can link up and have traffic in 10M and the PHY can still link up in 100M.

Please let me know what you think.

Best,

J