W5500 PHY Link failed when system reboot

Hello developers,

we are now using a controller with W5500 chip for UDP communication.
And it works normal (ip address, mac) at normal operations.
But when the system reboot, the W5500 chip will sometimes lost network connection.

The power input of the controller is 24V, we use regulators to 5v and then to 3.3v, follow the official design.

When the problem comes, the IP connection is lost, ping is not available, the ARP can not be responded. And we need to repowered the controller to make it work again.
we guess its the PHY of the 5500 chip that cause the problem.
Because we do some modify in the code to use SPI to read the W5500, and it works.

So we measure the power supply of the 3.3v of the W5500 chip, see as the following pictures.

My problem is Will this kind of power supply fluctuations cause the Phy not working?

And if you meet the same problem ever?
Could you give me some insights?

Hi I’m jaden

Noise is visible in the measured oscilloscope waveform.
What part of the waveform did Ref1, 2, and 3 each measure?

Can you please confirm that the Link LED on the W5500 also does not light up when the IP connection is lost?

Hello, Jaden.
Ref1, 2 and 3 all measure the 3.3v analog voltage which supplies the W5500.
Just measure 3 times. I put all the 3 waveform in one table sheet.

And it’s very interesting that the Linked Led will sometimes not lighting up, sometimes still lighting up.
But network connection in both cases are already lost.
In most cases, the Linked Green Led is off when it happens.
will the PHY functions bad but it kept the status(maybe lights on, maybe lights off)?
we tried to reset the PHY when the SPI detected the Linked status is 0(off), but it not work.
The error status (ping not available) will keep unless we perform a whole reset of W5500 or repower.
and now in code when we meet this situation, we reset the W5500 in a loop until the Linked Status is 1(lights on). And it worked. In my mind, this workaround works is another evidence that it’s only the PHY of the chip functions bad.

We are super curious about this problem and want to know the root cause.
In which Situations will break the PHY only but keep the W5500 working.
And as a beginner of hardware, me and my team what to go further about this topic.

if you want more information, we are glad to share more.

Send me a schematic and I’ll take a look at the W5500 HW part.

Is the VDD, 3V3, 3V3A is same power source?
I mean are they connected 0R or bead in other page?

“The system reboot” is mean 24V power on-off?
What happens if you try to reset W5500 when lose network connection after a system reboot

Not really, the 24V power keep supplying.
“The system reboot” means the device which the controller(network cable) connects perform a reboot. It’s like a personal computer perform a reboot process. But the device connected has a solenoid valve which will also perform on and off in reboot and will let other 24V power on and off.
But it will not impact the 24V power for the controller with W5500.

Do you think some electromagnetic interference will impact the behavior of W5500 chip?

Yes, I saw the noise measured by your oscilloscope and asked.

Does this noise only occur during system reset?

Yes, I make a trigger point, and the noise come form the system reset.

However, it’s not the case that each time this kind of noise is encountered and the 5500 lose network connection.
Some times the 3.3v pressure waveform is quite good and the 5500 chip still functions bad.
The following picture shows the error case for another 3 different measurement.
However, the first one and the second one has not much fluctuations.

if the electromagnetic interference will probably be a cause, we would go further to have an investigation about it.

Because we have several systems, and only in one system this error situation will reproduce frequently.
By some measurement, we could have a comparison upon the electromagnetic interference on different systems.

Do you have other suggesstions?

I’ll check the running environment cause there’s nothing wrong in your circuit and it’s not a problem with all devices.
If there is a noise source (solenoids, inverters, motors) close around the W5500, would you like to remove that noise source and change the Ethernet cable to Cat6 to test it out?