How long is it?
I think I'm finding mentions of up to 9000KM long, with example of Japan to/from west coast US. At 100Gbps? That _is_ long and fat. I'm jealous, um, errr....
Submarine cables and other long distance data connections may now be simpler and cheaper to deploy and operate, thanks to a new set of products from Ciena. Ciena is the heir to Nortel’s Optical Multi Service Edge 6500 products and has wrapped them up, with its own kit, into the new WaveeLogic 3 suite. The headline act for the …
You don't need to convert the signal to amplify it, there are optical amplifiers now.
Are they talking about dispersion? I'm not sure if that's even a problem at dispersion compensated fibers (you take one fiber with "positive" dispersion, and a matching fiber with "negative" dispersion and put them in series so the effects cancel each other out).
So what have they actually achieved?
If you amplify a signal you also amplify any noise on that signal, above a certain amount and the signal is no longer recognisable from the noise (poor signal to noise ratio)
So, on a short cable (less than say 250 KM you can probably (just about) get away with amplification (espescially if you are using Raman amplifiers)
On longer cables (1000 KM plus) you need regeneration, that's where you take the optical signal incoming, convert it back to an electrical signal and then convert it back into optical for the next leg of it's journey hence the rise in latency because the electrical - optical - electrical process takes time and on a VERY long cable (transatlantic say) you would have MANY regenerators.
What Ciena have done is remove the need (maybe completley in some cases) for ANY form of regeneration. It's very clever stuff, I'm currently working on 100 Gbps circuits that travel long distance (800 KM +) with no inline amplification or regeneration and it is quite complex (and quite scary, some of our fibre systems have 88 wavelengths running at between 10 and 100 Gigabits per second (PER FIBRE!)
It's very simple really but way to complex to explain in a small comment, does that make sense?
"So, on a short cable (less than say 250 KM you can probably (just about) get away with amplification (espescially if you are using Raman amplifiers)"
No, on a short cable of 250km you can get away with no submerged amplifiers, just using contra-pumping Raman units at each receive terminal - I am heavily involved in the commissioning and maintenance of two such cable systems.
"On longer cables (1000 KM plus) you need regeneration,"
No, undersea amplifiers are more than adequate for long span work - as an example, the Southern Cross cable's longest span is some 8000km from NZ to Hawaii, with no regeneration, just amplifiers.
"It's very clever stuff, I'm currently working on 100 Gbps circuits that travel long distance (800 KM +) with no inline amplification or regeneration and it is quite complex (and quite scary, some of our fibre systems have 88 wavelengths running at between 10 and 100 Gigabits per second (PER FIBRE!)"
Let me guess - 1830PSS?
Eh? According to wikipedia (yeah, I know...):
"The current laboratory fiber optic data rate record, held by Bell Labs in Villarceaux, France, is multiplexing 155 channels, each carrying 100 Gbit/s over a 7000 km fiber.[26] Nippon Telegraph and Telephone Corporation has also managed 69.1 Tbit/s over a single 240 km fiber (multiplexing 432 channels, equating to 171 Gbit/s per channel).[27] Bell Labs also broke a 100 Petabit per second kilometer barrier (15.5 Tbit/s over a single 7000 km fiber)."
http://en.wikipedia.org/wiki/Optical_fiber
So okay that was in a lab but where does the need for repeaters come in /at all/?
It's in the lab, under perfect conditions (it's still very impresive) and may come to fruition but the point about the Ciena kit is that it's doing it today, in the real world on fibres that were never specifically designed to acheive these feats.
My God, I sound like a real Ciena Fanboi, I do think the Ciena kit has an attrocious UI compared to the old Nortel 6500 Kit but the Ciena core director stuff is much simpler to operate.
I only hope they keep the best of the UI from Nortel and the best of the transmission kit from Ciena (but you know what they normally do!)
Attenuation can be overcome with amplifiers - the noise introduced by the amplifiers and the fibre itself is a limiting factor though - as are Chromatic Dispertion (can be compensated for at terminals), Polarisation Mode Dispertion (can be compensated for at terminals), non-linear effects as channel count (and hence launch power) increase eg., SBS, inter-channel modulation effects, Raman Scattering.
Wavefront dispersion (blurring) is a far bigger problem on longhaul cables than attenuation.
Of course none of this helps endusers much when there's a local monopoly. We're stuck with 1Gb/s at 40k+ per year for the forseeable future, no matter that 10Gb/s for 1/3 that price is available just down the road where there's a competitive supplier.
Nothing against these achievements (on the contrary), but worth noting anyhow, even if plenty of readers already know this:
Those ever faster streams are doing what CPUs run into a cycles-ceiling are doing, too: Split into multiple threads, or streams, or wavelengths, or what-have-you. Meaning that bandwidth increase no longer automatically implies latency reduction. The pipe's getting fatter, but not faster.
Hi Terranesia,
attenuation sure is basic, but can be reduced by using better fibre quality, and (the fibre permitting) the initial signal can be strengthened. Remember those old telegraph lines across the Atlantic used several hundred Volts. Not sure what optical signal strength this Ciena kit can operate with (really not my field), but if they really can lay a cable that's just a cable and has no added repeaters, it'll doubtless save some money. Also remember those repeaters have to be powered, which adds running cost.
I watched the video because this did sound on first reading something rather special. Removing the need for regeneration? It would be amazing if it were true. But I can only see they now have the flexibility over the transmission format to reduce the need for regeneration (at the expense of reduced bandwidth), not remove it entirely. Use of DSP in signal regeneration is nothing new, just they're now using a faster one (based on a smaller feature size 32nm CMOS process).
Hi Alan Brown, Rustident Spaceniak
Attenuation is the problem that limits how long the cable can be. Multiple vendors now offer the DSP that means the dispersion of a cable to be compensated (equalised) allowing thousands of kilometers with no inline dispersion compensation. Loss on the other hand limits you at best to several hundreds of kilometers before an optical amplifier is needed if you wanted to do something like a cable link from an island to a mainland. Crossing something like the Atlantic requires an optical amplifier every 50km or so due to noise considerations. There's no way around it.
Since the Atlantic was mentioned - if you assume something like 6500km of a cable route between Cornwall and New York and a state of the art loss of about 0.15dB/km you still have 1000dB of attenuation which is 100 orders of magnitude. To convert from optical to electrical I need to detect a signal of about 10 microWatts in NY. Therefore I would need to launch something like 10 to the power of 95 Watts to overcome the fibre attenuation. Give that our sun is outputting about about 10 to the power of 26 Watts you can see how big the attenuation problem is and how the DSP can't address that problem. We can probably argue about an order of magnitude or two in my above hand waving, but we're not going to find the missing 70 or so orders.
If we are just confining ourselves to shorter island hopping systems, then the cable is on the continental shelf and and still needs plenty of steel armouring to protect against drag fishing nets, anchors and so forth. This completely dominates the weight of the cable compared to any in-line repeater housings, so again the article is misleading.
Erbium doped fibre amplifier
magically turns green LED light, powered by simple auxiliary DC feed in the cable, into noiseless laser amplification of even densely packed IR carrier beams (wavelength division multiplexing), independently of the modulation scheme - upgrade both ends to 200GB/s, whatever, and you're done. .
You've never worked on any system that uses EDFAs, have you?
The amplifier pump lasers are slightly more sophisticated than a "green LED light" and the amplification is by no means "noiseless" - or so says the Alcatel-Lucent DWDM technical manuals I'm looking at, but hey, what would they know about it compared to you?
Am I really pretending to know more than an EDFA service manual? No, I am posting a short summary of a relevant technology. In essence, the EDFA is simply driven by green light and DC, rather than an entire demodulate/decode/error-correct/re-encode/remodulate chain for each of the 256 wavelengths. Sure there is a little more to it than that, the interested reader will know where to look.
Similarly, the noise argument is abbreviated, booby-trapped, even. It is an engineers conundrum - am I claiming an impossible result, a noiseless amplifier?. Yes it is certainly true that laser amplification is noiseless, the stimulated photon is identical in phase and frequency. The resolution of this puzzle offers insight to an enquiring mind, yes the amplification is noiseless, which is never normally the case, but hold on, there is an inescapable spontaneous emission in there also. Aha, seemingly fundamental constraints are observed then. Next week, the parametric amplifier, is it noiseless as claimed?