Reorganize SFP signals
Some observations on SFP connections:
- SFP_LOS line must not share function with TX_FAULT. The LOS pin will be required for early link-down detection and so it must go directly to the FPGA and not share the function with other pins.
- LEDs now are single colored (wrs to the bicolor LEDs we have in WRSv3).
I suggest using shift registers like '595 and '166 for the slow signals (LEDs, TX_DISABLE, TX_FAULT, DETECT) and a dedicated per-SFP I2C controller. They are very easy to handle in VHDL/Verilog and can be daisy-chained. Sample/update rate of 1-2ms for each of these pins is OK, so a single chain of the expanders (using 4 or 5 PL pins) will be able to handle all SFPs and their LEDs. Moreover, we can have bicolor SFP LEDs, juts like in the V3.
Here's the proposal for SFP pin assignment:
- LEDs: '595 GPIO shift reg
- DETECT: '166 GPIO shift reg/mux
- TX_FAULT: '595 GPIO shift reg
- SCL/SDA: dedicated I2C bus per each SFP to PL
- TX_DISABLE: '595 GPIO shift reg
- LOS: dedicated PL pin (must react fast for switchover implementation)