Hacker News Reader

Comment by fulafel | Hacker News Reader

"Researchers will this week demonstrate a newly-refined data-transmission technique that can deliver one terabit per second (Tbps) over optical fiber"

We already had 10 Tbps per fiber pair: http://www.extremetech.com/internet/231074-googles-faster-un...

View on HN

There's supposedly a 255Tbps multi-mode cable[0] as well.

[0]: http://www.extremetech.com/extreme/192929-255tbps-worlds-fas...

Just to clarify this is not regular multi-mode fiber, but a single fiber with seven cores.

A singlemode fiber is limited to about 100 Tbps per fiber at 10 bits/s/Hz. It is not obvious to me how much bandwidth usage has to grow until it becomes more economical to deploy this new kind of fiber rather than multiple (existing) singlemode fibers. For example, for high bandwidth applications MPO connectors with 12 or 24 fibers per connectors are used.

I'm curious as to where you came up with your symbol rate numbers. The theoretical limit[1] is 2 bits/Hz which for infrared light (~400THz) would put the limit at ~800Tbps.

[1] https://en.wikipedia.org/wiki/Nyquist_rate

The Nyquist signaling rate only applies to binary/two-level signals. Most optical communications systems use many levels to improve the data rate. StackExchange discussion: http://electronics.stackexchange.com/questions/21854/why-is-...

Ahhh, that makes sense! I wasn't aware that the Nyquist rate only applied to binary signal. Thanks for the SE link.

The usable parts of infrared light (amplifiable C and L bands in single mode fiber) is about 10 Thz. According to Shannon's theorem the theoretical upper limit for symbol rate in a linear channel with 30db SNR is 10 bits/s/Hz, independent of modulation format. This is how I derived my rough estimate of the capacity of single mode fiber.

The Nyquist rate is modulation dependent. Optical fiber can certainly achieve spectral efficiency over 2 bits/s/Hz.

Ok, Throndor provided a link which has a better explanation about this. I have some reading to do!

What technology determines how cable bandwidth is segmented to different tenants on fiber optic trunk connections? In theory 10 Tbps would divide out to something like 1,000 dedicated 10Gbps L2 connections but I'm guessing there's a lot more to it than that.

Either way it's pretty cool to see the increases in trunk capacity. This is enabling for a huge range of applications.

(Edit: Bad math on bandwidth, must.have.coffee!)

Bandwidth can be segmented off using multiple technologies. Basically you can either dedicate wavelengths per tenant (typically 96 wavelengths on a DWDM system) or dedicate lower bitrates per customer (for example 1 Gbps) and mux them together with other traffic until you fill up a 10G or 100G wavelength.

Thanks, that got me started. :)

the main difference is, this demonstration uses cables within Telekoms Network - so it's no laboratory but a real-world-test. (german source: http://www.golem.de/news/saser-muenchner-forscher-uebertrage...)

According to the article, FASTER has already entered service. It's the name of a deployed submarine cable, not a lab demo of new fiber technology.

uh, the parent is talking about an in-operation undersea cable. So it's no laboratory, it's a cable deployed in really harsh conditions.

Reply on news.ycombinator.com