0
\$\begingroup\$

In the past, I have dealt with fiber optic network communication devices that utilize two fibers, RX and TX, each being dedicated to one direction. I was under the impression that two fibers are always required for bidirectional communication.

However, recently I have encountered several devices that utilize a single fiber while providing bidirectional communication. These devices are present in telephone and intercom systems. An example is this device which provides two zero-latency analog audio channels plus a 10/100 Ethernet port over a single fiber.

From this document, I understand that single fiber bidirectional mode uses different wavelengths for send and receive modes, and filters on each end of the circuit to pass only the desired wavelengths.

I am not sure if the aforementioned devices use this method, or if they use some form of half-duplex interleaving to send and receive in separate time windows. Since the spec says a maximum of 50Mbps Ethernet transmission is possible, I am assuming a half-duplex method is being used, but this is not clearly stated anywhere.

My questions:

  1. How does the bidirectional communication of these devices work? does it use either of the above methods, or something else?

  2. Why would anyone want to use two fibers when one can accomplish the task?

Is there some sort of performance or other benefit to having two fibers? The aforementioned unit even provides a 10/100 Ethernet jack in addition to two zero-latency voice channels, so it appears bandwidth isn't an issue. I can understand if this method employs interleaving, the bandwidth would be drastically reduced, but if two different wavelengths are used, it should permit full-duplex transmission.

Thank you for any explanations of this, references etc.

\$\endgroup\$
4
  • 2
    \$\begingroup\$ networkengineering.stackexchange.com/questions/34306/… \$\endgroup\$
    – Eugene Sh.
    Commented Nov 18, 2022 at 18:16
  • 1
    \$\begingroup\$ 2. less hardware complexity \$\endgroup\$
    – jsotola
    Commented Nov 18, 2022 at 18:30
  • \$\begingroup\$ @jsotola I understand that it is less complex, but is it of enough difference to matter? For example, saving 50% of your fibers seems like a much better option than saving a few dollars on manufacturing the transceivers. Is there another reason, like performance, that would also push for dual-fiber solutions vs single? \$\endgroup\$ Commented Nov 18, 2022 at 18:38
  • \$\begingroup\$ For higher end devices, you can use a circulator, which splits incoming and outgoing photons into different fibers with almost no loss. Too expensive for cheap electronics though. \$\endgroup\$ Commented Nov 18, 2022 at 19:06

1 Answer 1

1
\$\begingroup\$

At 100 Mb/s you have so plenty of capacity left in the fiber that you can do a very basic half-duplex scheme without needing to worry much. 1Gb/s interfaces for these short-range cables literally are 12€ including distribution and marketing cost, new. You can guess how cheap the actual device is to manufacture! So, yeah, I'm surprised to even see a device that does less bits per second over fiber.

So, you get one tenth of the bidirectional bandwidth one of these interfaces could do unidirectionally. That's the hefty bandwidth reduction you'd incur if your half-duplex scheme is built on this relatively old technique and didn't even try to be smart about negotiating access to the shared medium.

So, my guess is really that they they basically use Aloha with a bit of a preamble to establish timing at the receiver. That's simple, needs no new hardware, and would work over arbitrary length cabling.

However, I can't look inside your consumer-grade device. It might be making use of a more rule-based medium access scheme (always replying to a packet you see, either with your own data or with an empty packet, so that there's a steady ping-pong, and each side can always train their clock recovery on something).

It's a bit of a mystery.

However, this single-fiber comms method (also, these days at much higher speeds) is widely deployed in Passive Optical Networks (PON); here, wavelength multiplex (WDM, as mentioned by you) is used. For that to work, you need to know which side is which – PONs as used to connect e.g. households to the internet have a notion of "upstream" and "downstream", so they define which device uses the shorter and which the longer wavelengths.

There's a hint your device does that as well:

Product picture, showing stickers

Although a bidirectional device in an ethernet topology, one is labeled RX (communicaitons engineer slang for "receiver") and the other TX ("transmitter"). That means these are different devices, and you can't use two TX together, or two RX together. That would perfectly fit the restrictions a cheap WDM system would impose (because you'd only put one receiver diode and one laser diode into each end, and not both).

If you want to learn about these PONs, check out wikipedia. If you're more interested in high-end optical transmission, you'll not find them in PONs; these are usually limited by the multi-user access and device cost, so they're (superficially) "old and boring" in tech (just like the device you referred to certainly isn't bleeding edge). WDM is important – all the modern speed records for single-medium transports are done on optical fibers using WDM and very intense techniques to combat not only the dispersive (think: different for different frequencies within the signal of a single wavelength), but also the nonlinear effects (where wavelength and signals in themselves suffer drastic phase distortion).

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.