Note that app. notes also specify, and this is important, that you power regulator must support inverse voltage
Imagine a case where your 3.3V network draw 50mA (for FPGA IO, 3.3
devices, ...). Then, the regulator that supply the 3.3V will outpu
50mA
Now, take a few IO that you put 5V on them, with resistors to limi
the current to 10mA. The way the FPGA is done, the IO pins hav protection diodes, which will take cut excess voltage at any IO t Vcco + 0.7V (diode at IO connected to the Vcc). Now, if you limi current to 10mA, with a resistor for example, the primary MO transistor at IO pin won't draw more current (still only some pico o nano-Amps). So, the 10mA that goes through the IO end-up feeding th
3.3V Vcco network
In the case above, having only one IO feeded by 5V (with 10mA goin
through the IO), will feed the Vcco 10mA, while the regulator wil now only provide 40mA. You will still have 3.3V on Vcco. Th problem get if you feed more current than what is needed device connected to Vcco. If you get too much current through the IO, then the 3.3V regulator will no longer supply current. All the curren will come from the FPGA through the protection diodes. The very BA thing is that the Vcco line will no longer be regulated to 3.3V, bu will get up higher (4.3V or more)
So, you must ensure that the regulator can also sink excess current
to maintain the voltage to 3.3V