DP83867E: The trace length and signal quality

Hi Amelie,

1. RGMII trace length should be ok, recommended is less than 6 inches. 

2. RX_CLK waveform suggests there are a lot of reflections. Have RGMII traces been controlled to 50 ohm single-ended? Are there any stubs in the trace path?

Thank you,

Nikhil

Hi Nikhil,

Thanks for your reply.

As customer feedback, they make 9 pces board.

3 of them could reach both TX and RX 900Mbps but 6 of them could only reach TX 900Mbps but RX only 100kbps. 

Within the 6 board, 2 of them could reach  both TX and RX 900Mbos after cutting RX_D0 stub res. However, 4 of them couldn't even after cutting all the RX_D's stub res.

Their stubs are very short, less than 110mils.

Customer want to know how could they judge whether the RX_D[3:0] are ok? By eye pattern？

Customer is worried whether it is because of the PCB manufacturer. Do you have some advice about how to make sure the RX_D from PHY to MAC is ok？

Hi Amelie,

There may be an impedance mismatch between the PHY and the MAC. Can the customer replace the 22 ohm series resistors with 0 ohm resistors? Depending on the internal impedance of the MAC, series resistor values of 0-50 ohms may show improved performance.

Additionally, can you provide a schematic/layout?

Thank you,

Nikhil

we replace the 22ohm series resistors with 0 ohm resistors, the receive speed is OK now, it can be about 900Mbps.

schematic:

pcb:

sch:

Hi Amelie,

I am unable to view the first set of schematic and layout images, but it sounds like the speed is ok now with a 0 ohm series resistor. Is there still any issue remaining?

Thank you,

Nikhil