[FAQ] DP83869-EP: Ethernet compliance Testing failure when interfaced to AM64x

Hi Sreenivasa,

Hope you're doing well! 

To debug compliance failures, I see that if the mask was readjusted to fit the waveform (shift waveform down and left), there is a good chance this would fit. I would advise customer to check with TekExpress SW for this functionality as this was evident on their predecessor (TDSET3). For distortion, I would like to check what register configuration is being used for this. Specifically, single channel test mode outputs as these are extended registers.

Regarding document, this is actually in revision and will include support for 869. This register is also needed for that device. This register increases VoD swing to output to compliance voltage levels, where the default configuration of the PHY was for maximum power savings.

Does the magnetic follow datasheet recommendation? And is layout as optimal as it can be as recorded in SNLA387?

Sincerely,

Gerome

Hello Gerome,

 i am doing good and i guess things are good on your side.

Thank you for the inputs.

Do you recommend the 10K + 22 pF cap on Rbias change done for the DP83867 to be done on DP83869.

Gerome Cacho said:

Does the magnetic follow datasheet recommendation? And is layout as optimal as it can be as recorded in SNLA387?

Do we have recommendations for integrated magnetics similar to the AM64x EVM.

Is there a compliance test report for DP83869?

Gerome Cacho said:

Regarding document, this is actually in revision and will include support for 869.

I assume the same register configuration is valid for DP83869. 

Regards,

Sreenivasa

Hi Sreenivasa,

We don't have any data on this scheme with regards to compliance testing. I know that was more for ESD if I recall correctly. However, I don't believe the issue customer is facing requires such a rework at this point in time.

The recommendations are the same for discrete and integrated magnetics.

UNH test report is available on Ethernet/ sharepoint under Industrial Ethernet Collateral. Please note if trying to share with customer that this is only available to them under NDA.

Register 0x1D5 is the same between 86x parts.

Sincerely,

Gerome

Hello Gerome,

Thank you.

Gerome Cacho said:

We don't have any data on this scheme with regards to compliance testing. I know that was more for ESD if I recall correctly. However, I don't believe the issue customer is facing requires such a rework at this point in time.

On the DP83867 We had observed improvements during compliance testing. Could you please confirm. 

I

Gerome Cacho said:

UNH test report is available on Ethernet/ sharepoint under Industrial Ethernet Collateral. Please note if trying to share with customer that this is only available to them under NDA.

Thank you and noted.

Regards,

Sreenivasa

Hello Gerome,

Please refer below inputs from customer and let me know your thoughts.

Finally I found the solution to pass DP83869 MDI test (partially tested) :

Following your advice I have to use https://www.ti.com/lit/an/snla239b/snla239b.pdf, which implies to set some undocumented registers (reg 0x1D5 for instance)

+

According to this E2E thread :

https://e2e.ti.com/support/interface-group/interface/f/interface-forum/1265530/dp83869hm-one-of-two-phys-not-passing-gigabit-complaince/4870370#4870370

I have to disable ‘ALL channels’ for Test Mode 1:

I share with you my personal feeling :

• I have to go fishing the right information just to pass the Ethernet MDI compliance test
• It makes me feel unconfident with the fact that we have to change buffer MDI output strength just to pass MDI compliance test
• What about same test with multiple channel activated?

Regards,

Sreenivasa

Hello Gerome, 

Thank you.

Gerome Cacho said:

DP8386x was tuned for lowest power performance at 100m reach. While this does translate to lower swing than compliance requirement, we do not expect any issues to arise from this. However, if compliance is something that customer needs to pass, this register can be set for compliance and end use case applications with no other expected side effects other than increased power consumption.

I assume you meant that this register can always be programmed. please clarify and this is the customer question here.

Gerome Cacho said:

With single channel activated, this may affect results as now there is no opportunity for cross talk from other channels.

Did you mean all channels activated? Are you suggestion to activate one channel at a time ?

Regards,

Sreenivasa

Hi Sreenivasa,

Yes, this register can always be programmed if customer desires.

For distortion test (TM4) as crosstalk plays a significant factor, we recommend doing single channel for this test. 
Sincerely,

Gerome