BPI-R4: Add notes about GPON modules / 3.3V

Author: avollmerhaus

_sfp_cage/banana-pi.md

@@ -18,5 +18,17 @@ layout: default
 | SGMII        | ✅                   | ✅                   | ✅                   |
 | Type         | Router               | Router               | Router               |
 
-
+# Notes
+## SFP power behavior on the Banana Pi R4  
+Unlike some devices, the Banana Pi R4 does not apply 3.3 V to the SFP cage immediately when a module is inserted.  
+Instead, the board uses a load-switch MOSFET which only enables power once the module asserts the MOD_DEF0 pin (the presence/ID signal defined in the SFP specification).  
+This design can cause problems with “smart” GPON sticks such as the Zyxel PMG3000-D20B.  
+The stick’s internal SoC requires a long time to boot before, and only then it pulls MOD_DEF0 low.  
+Because the R4 waits for MOD_DEF0 before supplying power, the stick never receives power at all, resulting in a deadlock where the module stays unpowered.
+  
+Some users in the BPI-R4 forums discussed a [workaround](https://forum.banana-pi.org/t/bpi-r4-zyxel-pmg3000-d20b-sfp-module-not-detected/18413/2) which bypasses the load-switch and feeds 3.3 V directly to the cage.  
+This allows modules like the PMG3000-D20B to start, but at the cost of losing proper hot-plug behavior.  
+  
+FS.com also offers GPON ONU sticks advertised as “with MAC,” which reportedly assert MOD_DEF0 early and may therefore avoid the issue without hardware modification.  
+However, evaluation of these modules on the R4 is still outstanding.