WG82574IT in failure

[BMatt4]

This is my first time to the community so hello everybody.

We are using two ethernet controllers WG82574IT in a motherboard for some specific projects. The board are in the field and we have some failure feedback on these components.

The controller in failure doesn't respond on the PCIe but generates clock and data signals on the EEPROM side, so it is not completely dead.

By investigating, we have found that the ethernet transfomer has bad solders on both side (cable side and controller side).

Is it possible that this solder issue damaged the PCIe part of the controller by creating spikes or other phenomena?

Thanks

Best Regards

Matthieu

[Gabriel_T_Intel]

Hello Baquematthieu.

Welcome to the Embedded Community, we are very glad to have you as a new member.

We are going to review your case and will get back to you very soon.

Best Regards

Gabriel Thomas.

[Adolfo_S_Intel]

Hello baquematthieu

Are the Ethernet controllers part of retail cards or are they soldered in the motherboard?

The motherboard was manufactured by your company or was purchased from a third party vendor?

The PCIe was working before, or has it always been malfunctioning?

Have you determined if there are short circuits, or continuity problems in the PCI lines?

Best Regards

Adolfo S.

[Natalie_Z_Intel]

Hello, Matthieu! So glad that you found the Embedded Community. Would love to know how you found us. Please know that we will do our best to assist you!

I worked with, John, an Intel engineer and he said, "The magnetics are attached to the MDI pins, and the PCIe interface is attached to the PCIe pins. The MDI pins and the PCIe pins are on opposite sides of the package. Any voltage spikes would have to damage the chip itself to kill the PCIe interface. My suggestion is to investigate further. Since you have bad solder connections on the magnetics, I would look for additional bad connections on the 82574. If the NVM clock is working, then the crystal must be operating, so the chip can't be entirely dead, but it is always possible that it was damaged. However, it seems unlikely that both chips were damaged in the exact same way resulting in the exact same symptoms. I have a weekly meeting with the hardware engineers shortly and run it by them for additional comment."

I will let you know when John updates me again. Have a great day! LynnZ.

[Natalie_Z_Intel]

John met with the hardware engineers and they agreed with John's assessment. They feel it is most likely some other manufacturing defect causing the issue.

[BMatt4]

Hi all,

Thanks for your support, I found the community by searching a way to contact the technical support of Intel on the Intel Embedded web-site.

For Adolfo S

The Ethernet Controller are soldered on the motherboard.

The motherboard is designed and manufactured by our company.

The PCIe was working before the failure.

There is no short circuit or open connections on the PCIe links. I already tried to swap the PCIe link with another right one on the board and the result is the same.

The next step for us, according to your answers:

We will make X-Ray controls on the soldering of the controller and PCIe link.

We will swap the controller between two boards. We will unsolder the controller in failure, replace it by a new one and re-solder it on a board which is working.

I give you an update as soon as I have the result, hoping next week.

Great day to you too.

Matthieu

[BMatt4]

Hi LynnZ,

I have some news on this issue and so some questions. I have seen that the issue is linked with the resistor connected at the pin 48 (RSET) of the WG82574. On the board in failure, this resistor is dead (value is more than 150Kohms). We are now searching the root cause of that. It seems not related to mechanical or thermical stress.

The resistor we use is a 4K99, 1%, 63mW in a 0603 package. I have seen that you use a 100mW one in the reference deisgn. Could you confirm me that our resistor is OK?

Do you have any idea of what could damage this resistor (spikes on the ethernet part,...)?

After your answer, we will discuss with our manufacturer.

Thanks

Regards

Matthieu Baque

[Adolfo_S_Intel]

Hello baquematthieu

I have checked the Datasheet, and the Design Guide (doc# 475120) for the 82574IT. It does mention that you the RSET should be 4.99kΩ+1%, but it doesn't define any power requirement, so I will check it with our engineers.

In the meantime, you might want to check the following sections to make sure that everything is in order with the PHY:

On the datasheet please check sections 13.5.5.1 and 13.5.5.2 regarding the placement of the Magnetic.

It could be useful to read Section 12.8 to guarantee that the Magnetics that you are using in your board comply with the required specifications.

Also check Table 102, section 15.0, specially the Placement of the 82574 section.

In the Schematics and Layout Specification (doc# 361392) states:

"If you are using discrete magnetics the RSET trace length should be less than 0.5 of an inch"

You could also be having troubles with the power of the chip, please make sure that you are complying with the Electrical Specifications for the Power Rails (Section 12.2) and the Power Sequencing (Sections 12.3 and 12.4) of the datasheet.

Also you should check for compliance of the instructions of the Section 2.6, Table 6 of the Design Guide.

I hope this information is useful to you.

Best Regards,

Adolfo

[BMatt4]

Hello AdolfoS,

I have some trouble to download the design guide, so I'll check all your remarks once the download issue is resolved.

Nevertheless, I forgot to precise that the controller isn't dead. On a board in failure, if I change only the resistor, all is working correctly afterwards.

So, do you think that power supplies or magnetic placement can break only the RSET resistor without damaging the chip itself?

Regards

Matthieu Baque

[BMatt4]

Hi, I have again some trouble to download the design guide but it seems related with our proxy or firewall.

Nevertheless, could you confirm that the power rating of 63 mW is correct for this resistor?

If power is correct, do you have any idea of which could break the RSET resistor (spikes...)?

I assume that this resistor have connections with both PCIe and Ethernet part regarding the datasheet, could you confirm that also?

Thanks for your support.

Matthieu

[Adolfo_S_Intel]

Hello baquematthieu

I assume that this resistor have connections with both PCIe and Ethernet part regarding the datasheet, could you confirm that also?

On the datasheet page 485, Table 103. Schematic Checklist related to RSET it states: This is required by the PCIe and MDI interfaces

Regarding your questions about the power dissipation of RSET I haven't been able to find any documentation that states a minimum or required capacity. This case is being consulted with our engineers and I will let you know as soon as I have more meaningful information for you.

Best Regards,

Adolfo

[BMatt4]

Hello,

First, let me wish you a happy new year.

This issue is escalating and we have now stopped our production of computers.

So, your answer about the power is very important for us.

We are searching why the value of the resistor changes during the life of our product.

Do you know if any event on the PCIe, Ethernet part, power supplies... could affect the RSET resistor?

We could provide our schematic if needed.

Thanks for your help.

Best Regards

Matthieu

[Adolfo_S_Intel]

Hello baquematthieu

Have you been able to download the Design Guide and contrast its recommendations against your design?

Regarding the power issue, I haven't been able to find any specific guideline, RSET shouldn't need to dissipate more than a few miliwatts

Have you considered the possibility that the batch of resistors that you are using for RSET were defective?

I will probe our engineer for further recommendations regarding your issue.

And happy new year to you too!

Best Regards,

Adolfo

[Adolfo_S_Intel]

Hello baquematthieu

I think this service might help you to rule out any possible error on the design: https://edc.intel.com/Tools/Design-Review/Default.aspx?language=en Design Review Services

I hope this is useful to you.

Best Regards,

Adolfo.

[BMatt4]

Hello,

Sorry for this delayed answer.

For the update, I have tested the controller in failure on a new board and the controller works. I solder a new one on the board in failure and the result is the same (controller in failure).

So, it seems that the issue isn't related with the controller but with our board (PCIe link, host PCIe or PCIe switch....).

I'll continu investigation on the PCIe part.

Thanks for your help.

Best Regards

Matthieu Baque

[Natalie_Z_Intel]

Hi, you should now be able to access Intel confidential content!

[BMatt4]

Thanks, I'm still not able to download the design guide but I'll check with my IT because the issue could be related with our firewall or proxy.

Matthieu

[BMatt4 · Accepted solution]

Hi All,

We have progressed in the search of the root cause of our issue. It seems not related to the resistor itself nor the ethernet controller. Our issue is related with corrosion inside the resistor.

So, thanks at all for your help and support.

Regards

Matthieu

[Natalie_Z_Intel]

You are welcome. Thanks for using the Embedded Community. I am glad the issue is resolved! LynnZ