I have a question for the group, where I have some experience with ethernet and such... I haven't the experience designing devices which use it as much.. So I'm wondering if anyone knows of a solution to use fiber optics instead of cat5/6 cables.....However, not in the traditional fiber manner... More what I mean is there a product that allows a fiber optic transceiver of some sort to be connected to the mbed for example in place of the normal magjack one would use for a copper cable? An extreme longshot I'm sure, especially since I don't recall any fiber switches, etc. that would accommodate that, or would they? Any hacks come to mind that might allow this?

Anyway, I'd love to use fiber in a project instead of copper for so many reasons... Any word is appreciated!

BTW- I am aware of media converters and the like but I would like to also keep the cost and part count down.... Also, this would be on a closed network of nothing but a few mbeds talking to each other.....

Can you list any of the "so many reasons"?

Why can't you use a media converter? You probably wouldn't need a magjack if its right next to the mbed... The media converter probably already has magnetics in it so a magjack would be somewhat redundant.

My reasons would include the weight savings (every ounce counts), bulk, along with some benefits of using as many non metallic components as possible, and a few others.....

I'm avoiding media converters due to part count, cost, weight, size restrictions, etc...

It just would seem there would be something that could be done sort of like a toslink type, even if it were a few smaller components..

Something along these lines? http://www.electronickits.com/kit/complete/fibe/ck1500.htm

Very close to what I was thinking Igor..! I had considered some sort of hack using toslink (http://www.sparkfun.com/products/10205) (http://www.sparkfun.com/products/10203) I would imagine you might need one cable for transmit and one for receive? I'm just wondering about the signal conversion between where you'd connect the magjack and the toslink transmitter/receiver... And you'd probably need a modified ethernet switch as well.. =)

Ethernet signaling uses differential pairs for RX and TX. The signal is also modulated. It would require some effort to interface the toslink RX and TX to work with 10base-T, although it's certainly possible. See http://ronja.twibright.com/.

Serial, however, would be fairly simple to interface with a pair of TX/RX toslink thingys. But I get the feeling that you want Ethernet..

EDIT: Actually on second thought, Ethernet might not be that hard to acheive. Have a look at these:

ML4667 transceiver (http://www.datasheetcatalog.org/datasheet/MicroLinearCorporation/mXrusss.pdf)

HFBR-4663 (http://www.datasheetcatalog.org/datasheet/hp/HFBR-4663.pdf)

Ahh, very cool Igor.. I would like ethernet. The two transceiver chips look pretty good, I was kinda hoping to be able to use the mBeds tx but I wouldn't rule something like this out... Good point on the ethernet, I had not thought about the fact there is a tx+ and tx- as well as rx+ and rx- not just tx and rx with a common...

So, after a little reading on differential signalling (http://en.wikipedia.org/wiki/Differential_signaling) and 10-baseT ethernet (http://en.wikipedia.org/wiki/10BASE-T )..........

They state A 10BASE-T transmitter sends two differential voltages, +2.5 V or −2.5 V.... And that signal goes into a subtractor (XOR and NAND gate - http://en.wikipedia.org/wiki/Subtractor)...

Doesn't it make sense that you could have a subtractor connected straight to the mbeds tx+/- and drive something like a toslink (given the bandwidth of the device can handle the 30mhz or whatever 10bt uses) connector from that?

That subtractor is a digital system (i.e. made using logic gates..) The physical ethernet layer is purely analog. The subtractor you have referenced is not relevant to what you want to do :)

To keep it simple, the TX+ and TX- voltages are subtracted from each other to yield the actual ethernet TX signal. To do this, you can use a differential amplifier.

Getting differential amplifiers to work at 30MHz isn't very trivial. You would have to use very fast op-amps (gain bandwidth product >> 30MHz) which make it difficult if not impossible to do on a breadboard.

That being said, you might be able to use the TX/RX pairs from the mbed without the need of a differential amp. It could be possible to use TD- and RD+ (referenced to GND) to interface with the toslink transceivers. I'm not too familiar with the PHY implementation of the mbed so I would not recommend going this route as you may damage the PHY chip on the mbed.

I think it might be easier to use the transceiver chips that I mentioned above rather than building your own interface circuitry.

I hope that helps :)

It is a lot of help Igor.. Thank you.. I'm obviously not as familiar with ethernet as I should be to take this on... ;)

Considering the need for reliability as well as the other factors I've mentioned... I might have found a middle ground.. Half of the devices are grouped in a relatively small area several feet from the other half which are also grouped.. It may be an acceptable compromise to use a switch at both ends with a fiber uplink between them. That actually reduces the long cabling required in between in the long run.. The challenge there is to find a switch that's small and lightweight enough... =)

Thanks so much for the input! I might actually play around with your idea of connecting the TD- and RD+ referenced to gnd on a device I don't care about just to play around with the idea anyway... It might be easier/better to use the transceiver chips you mentioned, but I'm wondering if the current IP stack for the mbed would be usable?

No problem. It sounds like an interesting project. As long as the physical layer does what its supposed to do, you should be able to use any stack available for the mbed.