Geez, that was way over a year ago! Nonetheless, I do appreciate your reply. Since my original message, I have implemented a complete ethernet fiber optic system by directly programming an STM32F427 (ARM Cortex M4) chip’s ethernet interface. I’ve connected the ARM chip to a Micrel (now MicroChip) KSZ8041FTL PHY driving a Planet SFP fiber-optic transceiver. I can shove many thousands of ethernet packets per second with this system, and that’s with just an RMII interface, not MII.
It turns out that there is very little difference between CAT5 and SFP fiber-optic, from the point-of-view of a PHY. The KSZ8041FTL has one pin dedicated to selecting CAT5 versus fiber-optic, and all that does inside the PHY is some very minor
stuff like enabling or disabling a scrambler circuit. I have a connector with RMII signals for the input to the PHY, and another connector (this is a prototype) on the output of the PHY that holds the usual TX+,TX-,RX+.RX- signals. Into this PHY output connector I can plug EITHER a CAT5 Magjack adapter board OR an SFP fiber-optic adapter board, with one extra signal line telling the PHY whether to drive the output as CAT5 or FO.
Yes, I must use fiber-optic, not CAT5. This is an industrial application where the customer demands extreme galvanic isolation as well as long cable runs (CAT5 is only good for 100 meters, while FO is good for kilometers).
I would suggest that the WizNet folks would be smart to incorporate fiber-optic PHY capability the next time they update their line of ethernet chips.
Thanks again for your reply.
E. Nicholas Cupery