29.04.2014
08:44:23
Phase-1
compatible links to backplane, according to Dave Sankey scheme. That’s 2*8
pairs from the backplane (fabric interface). Mezzanine. Clock, TTC data, ROD
links. Since in this case the ROD is external, the links should come from processor
FPGAs. Caution! That would require additional clocks that we haven’t thought about
and might not be possible! But in this case a separate FPGA might have to sit
on the mezzanine as well!
Check Ethernet
backplane link. Base interface. Future use of something different from 1G Ethernet
would be possible only if running onto mezzanine
We discussed
widening phase0 ROD/control path by using MGTs out of processors. Same argument
as above, due to limitations on clocks that might be possible running at 40.08
multiple only. What is the current MGT budget on the Kintex device?
Not to forget:
Kintex clocking
We are
probably free to decide where the scsi
connector goes but we need to inform CTP on consequences:
Fixed routing
16+18 or 17+17, also remind them that the source is 2 FPGAs and that has
consequences on clock and parity!