09-15-2014 10:55 PM
Hi, I encountered a problem about recodeerr provided by vertex 6 GTH.
As we know, if a GTH link is physically disconnected(the cable is removed), we can see rxcodderr ports from the corresponding GTH lanes would become '1' or keep jumping.
In the vertex6 chips, I configured 2 GTH as 2 independent link, which connect to 2 different chips.
The quads I used are GTHE1_QUAD_X1Y1 and GTHE1_QUAD_X1Y2.
But, if I disconnected the cable of X1Y2, the rxcodeerr of X1Y1 would be seldom '1'(kept for only 1~2 clks rarely).
I thought that that 2 links should operate independently. When X1Y2 is disconnected, the link of X1Y1 should keep its operation normally, but the recodeerr of X1Y1 shows up sometimes.
How can I solve such a problems, or should I directly omit the recodeerr of X1Y1 in this situation?
09-16-2014 01:26 AM
09-16-2014 03:53 AM
09-16-2014 03:59 AM
I am using the GTH wizard(start from scratch) from coregen.
For the both GTH quad, the corresponding gth_refclks are distributed without any other access.
The drp_clk is generated from other free-running clock.(Both of the quads are using the same drp_clk.)
09-16-2014 07:48 AM
For the vertex6 chip I used, there are 6 GTH quads, with 6 corresponding gth_refclks.
About the problem I mentioned, I also tried to use another quad to check.
If I configured X1Y1 with disconnected X0Y0, I didn't see the rxcodeerr of X1Y1.
It seems like that if X1Y1 is configured with disconnected X1Y2 or X1Y0, then rxcodeerr of X1Y1 shows up.
When the 2 quads belongs to the same column, their operation are not independent to each other.
Is there any solution to improve such a situation?