DP83867IS: SGMII eye diagram issue.

Hi Chris,

I have a few suggestions:

1. Can you verify that the SGMII traces have a differential 100 ohm impedance?
2. Would you be able to perform a bit error rate test to see if the eye diagram is still acceptable?
3. Since the eye from PHY to CPU is fine, there might be an issue on the transmit side of the CPU. You can try checking the issue from the CPU side.

Regards,

Adrian Kam

Hi Adrian, thanks for quick response.

For 1/, of course, it must be 100ohm impdedance.

For 2/, I copied the eye mask below, it must fail at left side.

For 3/, we are checking the CPU internally.

Regards

Chris

What's the input impedance of SGMII rx pair? I didn't find any info about that.

Hi Chris,

The input impedance of the SGMII rx pair is 100 ohm differential.

If you are still having issues after checking the CPU, let me know.

Regards,

Adrian Kam

I have a CPU evaluation board, which is using the same CPU and PHY from other vendor, the waveform looks good. 

Another question is, for DP83867, what's the flight time between package pin and its internal die, or how long the wire is? 

Hi Chris,

Is there a reason this information is needed for your application?

Regards,

Adrian Kam

If the internal wire/flight time is too long, there will be  glitch or step in the rising edge.

Hi Chris,

We have done extensive testing on the DP83867 in SGMII mode and have not observed any glitch or step in the rising edge. The length of the internal wire should be too short to cause any issues.

Regards,

Adrian Kam

Hi Adrian,

Sorry for the long delay, we have done lots of testing recently, we even cut the wire at PHY pin side, the waveform is still nice, so we are kind of 100% sure of that there is bug in DP83867 SGMII inputs.

Regards

Chris

Hi Chris,

I have a few questions:

1. Have you tried this with another DP83867 or another PHY and see the same issue?
2. In the original scope capture, where are you measuring the signal? Is it at the SGMII pins of the DP83867 or somewhere else?
3. To clarify, in your latest capture, it is measuring the signal being transmitted by the CPU, but the CPU is not connected to the PHY, correct?

Regards,

Adrian Kam

Hi Adrian,

For 1/, I have another board with the same CPU and a different PHY, the eye diagram is much better, as the connection is pretty simply just connecting CPU and PHY by a AC couple capacitor.

For 2. All the measurement is at the SGMII pin of DP83867.

For 3/, yes, cut the wire in PHY side.

We also made an measurement with TDR, which will show the impedance curve, we found the internal impedance of DP83867 dropped a lot, almost 32 ohm.

Regards

Chris

Hi Chris,

Let me consult with my team on this. I can get back to you by Monday at the latest.

Regards,

Adrian Kam

Thanks for helping, really appreciated. 

Regards

Chris

Hi Chris,

I have the following questions/suggestions:

1. Can you set the DP83867 to loopback SGMII and see if the issue going from CPU to PHY also shows up from PHY to CPU?
2. Can you share the CPU's transmitter output voltage and jitter specifications?
3. Would it be possible to make some measurements on the CPU side and see if you are seeing the same issue? When measuring the SGMII output from the CPU, are you terminating it by 100 ohms then measuring?

Regards,

Adrian Kam