Help!! 10G between Custom board and zcu102 does not work

[alexkeys]

Hello,

We have implemented a 10g design for both zcu102 and our custom board. We have two of each board so we tested the designs first with the similar boards and it worked almost flawlessly. However, whenever we tried to test between our board and that of zcu102, the rx block could not receive a lock. I then checked the following registers 

1. 0404 (sta_rx_status_reg1) = 0x000000C0
2. 040c (stat_rx_block_lock) = 0x00000000

As you can see from the output of 0404, 0xC0 means 

1. stat_rx_internal_local_fault
2. stat_rx_local_fault

Now here is the clock set up for the GT

• Custom board = 322.265625MHz
• zcu102 = 156.25MHz

In the Linux kernel, I enabled Xilinx PHYs and disabled AXI_DMAs together with other settings suggested in the Xilinx wiki for Axi_ethernet (the Rx block locks when we test similar boards) 

I should also mention that the errors persisted even after I reset the gt registers. Now, I am lost for options. What could possibly be wrong? 

[nanz]

Hi @alexkeys ,

RX local fault indicates that the receiver is not up and operational.  Most likely causes for an rx local fault are:

• transceiver has not locked or the receiver is being reset
• At least one of the lanes is not synchronized
• The lanes are not properly aligned

Have you been able to track down on different stages and check where might be stuck?

[guozhenp]

1. Create 2 IBERT designs and check if they can link up with the custom board and ZCU120

2. Please make sure both designs are run at 10.3125Gbps. For example, run functional simulation and check TXP/N rate in it

3. GT locations are set correctly and P/N polarity are not inverted on board