W6100 having unstable link status

[kenhansen]

We are experiencing an intermittent PHY link connection on a W6100 design.

Not all products show the issue but a good percentage do show the fault.

It is dropping the link when the ethernet is connected but also incredibly will sometimes show a valid link when no cable is connected.

Our software is monitoring the PHYSR register and then correctly reporting the Link status which was additionally verified by monitoring Pin 17 LNKn Link status LED.

Therefore, we do not think it is a software/configuration related issue as pin 17 is showing the “bare metal” LINK status.

We are at a loss as to how the W6100 can see a valid carrier signal with no cable connected and indicate a valid LINK on pin 17.

If this can be solved then hopefully that will also resolve the link dropping during normal operation but that is a guess at this stage.

What we have tried,

• Monitoring pin 17 which shows the issue
• We added the “bob smith” termination 75Ohm/1nF on the cable side of the magnetics between L555 and P551 with no effect.
  This was to get as close to the reference design as possible
• The PCB layer stack is impedance matched. The traces are also very short so we are struggling to see this being the issue.
• The ethernet magnetics are from Link-PP and look suitable for the application
  LP2019ANL , S16-2019​ , LAN Magnetic Transformer , Discrete Magnetic Transformer , Lan Transformer , Pulse Transformer , Telecom Transformer , Transformer Modules , Ethernet Transformer , Network Transformer Manufacturers - Low Price LP2019ANL - LP2019ANL 100% Cross 749013011 We

Any assistance would be greatly appreciated.

Thanks
Ken

[Eugeny]

How the chip was soldered, and what was the soldering profile - graph and temperatures? Please explain as much detail as possible.

@kenhansen

We are at a loss as to how the W6100 can see a valid carrier signal with no cable connected and indicate a valid LINK on pin 17.

I have the only one explanation - chip picks up its own carrier, and most possibly it happens through transformer 3V3A connection. If you look at the reference schematic it says in the top left corner “as close to transformer 2, 7 pin as possible”. Its meaning is very unclear, and I believe it means these transformer pins must be connected to 3V3A as close as possible with as thick trace as possible to minimize crosstalk between TX and RX. On your board I see this possibility was not mitigated, with long thin 3V3A trace from the source and connection to the pins 2 and 7 of the transformer with decoupling capacitor relatively far away from them.

The USBLC6-2 chip may also give a crosstalk between the TX and RX channels. This chip is not present in the reference schematic, and chip’s datasheet does not detail its use for Ethernet, without any notes or explanations, only block diagram without complete reference schematic.

[kenhansen]

Thanks Eugeny. We will investigate your surgestions once we receive the boards from manufacturer. We get back with update once that have been done.
Thank you so much for your input.