Slashdot Mirror
Fiber Optic World Records Broken
Thousands of miles of existing fiber still lie dark, but as
schnucki writes, "Bell Laboratories believe they have broken two world records in the use of optical fibres to transmit information." They sent 160 gigabits/sec on one wavelength, and then in a separate experiment sent 1,022 separate wavelengths down one fiber. You do the math.
Check it out.
I am getting 21600 on my 56K modem at my new apartment? Did I win anything?
Funny and I thought Perl == Paid employment recently located
at that speed, NTL have no right limiting our cablemodems to 512k/s downstream, 128k k/s upstream.
a third record has just recently been broken. Bell Laboraties has finally succeeded in using fibre optic cabling to send pjorn at speeds until now unkown to many scientists. Guys were actually cumming before the pictures hit their screens...not to mention something else that did a second or two later!
"You do the math" is misleading. Because of interference (I don't know the actual term, but think of it like crosstalk) between the wavelengths, you can't just run 1022 separate wavelengths on a fiber and send data on it. 20TB/s would be nice though.
The article states that data can be transmitted over fibre optics at "the speed of light". I'm curious, can the speed of light be measured in Gb/s?
Bandwidth boys and girls? I knew you could!
:)
Can you say really amazing pornography? I thought so.... It seems like any technology improvement is sucked up there anyway. That's OK. Let the perverts drive the cost/unit down.
On a personal note, I was pretty pumped to get 115K a second on a download on my cable adapter (I refuse to say cable MODEM!) the other day
DO NOT DISTURB THE SE
Okay, I'm impressed.
But when does this trickle down to us home/office users? I want a multi_gigabit connection on my LAN.
Also does this mean a faster internet? I have a cable modem and many sites are just soooo slow.
[=]
Nil* Was Here
The first record promises to increase speed.
Woohoo! No more poxy 299,792,458m/s! According to relativity, this means that the data can go backwards in time. Soon we'll all be complaining about anti-latency.
Normally I wouldn't nitpick the BBC[1], but seeing as they split an infinitive in the last paragraph, I felt I had no choice.
[1]This is a lie.
Each experiment is highly impressive, but why don't they try to do both at once? Yes, logic (and the theory of fo) says that it will work, but they should still test it.
Anyhow, this is something else that still needs to be implemented over a wide area. Many of the main phone carrier lines here are fo, but most buildings still have 20 year old copper, so it doesn't do the user much good. Hopefully telco's will (are?) encouraging the use of fo in new buildings. How much of the U.S. telephone systems have been replaced with fiber?
Intolerant people should be shot.
I'd like to have that running into my home. Perhaps we could use SimCity to demonstrate the traffic surge that will generate? The subways could be the fibers, put some commerce, here, there. Cram the schools in the lower corner, Why does SimCity insist that there have to be power generators inside the internet?
Someone set us up the bomb, so shine we are!
I'm just curious what the error rate is. Can you send that much data without loosing a few bits. I'm impressed with the numbers regardless, but if you have a 50% error rate is that good? Also if you loose one bit per byte, can you clam anything? Did they only count the good bits sent?
;). If you send 160 Gb/s (20 GB/s), can you convert that to electrical data without making a bottleneck. I'm not an expert in this area, but I'm curious to know.
Also, what is the speed to transfer the light signals to electrical. I don't have (or have I ever heard of) an optical computer, would be nice though
Steven Rostedt
Steven Rostedt
-- Nevermind
160 gigabits/sec * 1022 wavelengths = 163,520 gigabits/sec on a single piece of fiber.
Which reminds me, I need more fiber in my diet, if only for the bandwidth alone.
--
Why can't I moderate something "Wrong" or at least "Grossly Misinformed"?
It's great news (and I submitted it a week ago and got rejected - grrrrrr.... :) but note that the experiment requires Lucent's new optical cables.
I doubt the mass market is going to re-lay their existing, im-pure cables anytime soon. So Doom fans take it easy.
However new corporations should look with interest. And investors may want a bet on Lucent: first a record fast switch and then a record fast cable. These guys are on to something.
Hi!
I'm most positive that bandwidth will not be a "buzzword" for future generations. I remember living in a house (as a child) with a T.V. antenna on it. Every neighbor had one too. We got Maybe 4 channels on a good day. We had a clicker on the table that actually went ~click~. Now, I have cable TV. There are over 200 channels, and everyone I know has this or a sattelite. The only place I've seen rooftop antenna's this year, is a trailer park. There's gonna be a day, in the not-to-distant future where bandwidth will be endless (in comparison to now). The Bandwidth Blues will be just another Golden Oldie.
the number of simultaneous calls that can be carried using 32 colours on a fibre to be increased from 30,000 to millions
That's a lot of calls at the same time. Maybe someday we'll be able to watch digital TV while bungee jumping. But who cares.
Everyone on earth will soon be able to call everyone at the same time (to just talk about nothing).
The world needs the new optical technology
They mean a small part of the world needs optical technology (nerds). :-|
The rest of the world needs water, electricity and food.
Solutions for a small part of the planet. a.k.a IBM's motto
Just thoughts
What they are doing here is called WDM (wave division multiplexing). There has been a standard for a while on WDM. But the thing is this only allows you to cram more information down one pipe it doesn't shrink the equipment on either end. They need to stick in and OC-48 for each wave length, so you would need about 1 foot ball field of space to stick all of the equiptment for this one piece of fiber. To sum it up don't expect to see any dramatic speedups anytime soon
160,000,000 bits / 8 (one byte) = 20,000,000 Bytes or ~ 20Mbytes * 1022 (number of channels) = ~20440 Megabytes a second. Quite nice if you ask me. NOTE: The math is probably way off since in most articles everything is rounded and nice.
They claim all 1022 wavelengths were transmitted at the same time with an ultra fast laser. I thought lasers emitted coherent light at one wavelength. I have seen adjustable lasers, but it was one wavelength at a time and depended on tube geometry and the die used. How do they tune their laser to transmit multiple frequencies at the same time? It sounds unbeleivable to me.
And boy, will that ever give us more bandwidth to support the increasing numbers of cellular phones!
(Um, oops... How do you get the fibre cables out to the phones?)
If you're not part of the solution, you're part of the precipitate.
Vacuous futurist idea: Imagine a very small machine that burrows from a central office/switch to your basement with very little operator attention. As it burrows it's dragging along a strand or two of lovely fiber.
Listen to me whine! I'm lucky enough to have a cable modem and I'm STILL not content!
License: By reading this you are agreeing that you agree with me.
I think the most impacting news is that optic "router". That is going to have the biggest effect. We dont' need 160gb/s if it has to be converted into an electronic signal everytime it switches fiber lines.
I have a similar problem..is it caused by a bad phone line or bad isp anyone know? I get a slightly better speed with one isp, but still way below what I'd expect.
------
------
poing!
This isn't really revolutionary new technology.. we've known about stuff like this for awhile. There's a nearly infinite number of ways to encode frequencies, and stack things onto each other.
I find myself wanting of the ability to insert IMG tags here. :( In short, picture a sine wave. Now along the slope of one, picture another sine wave attached to it. And so on. I suspect they're doing something like that. Actually, TVs do something like this - it's how the sync pulses and whatnot work. Very facinating technology. Also very old by today's standard, but still very useful.
--
No, this would _not_ make a way cool Beowulf cluster. Go away, get a clue.
(20440gigabytes/sec)/(1024Gigabyte/TB)=
_______________19.961TB/s_______________
Now some body add in some EC code, say we use up 10% of the traffic for EC, stil have 19TB/sec!!!
At 163 Tbps, it is capable of handling 105,496,774 T1 lines!!!
you could also calculate that since each T1 has 24 channels (right?) that's 2,531,922,576 channels available! (1 channel = 1 normal telephone line.)
I can't wait.
Bob.
technically, according to the newest published editions of a very respected book on english grammar (oxford, i think), split infinitives are now alright to use. my english teacher was a bit annoyed at this realization.
(Feel free to modify words in that sentence so as to provide bad jokes. There are ample options available...)
If you're not part of the solution, you're part of the precipitate.
The phone system is due for a major overhaul. The fact that I can plug in a *crank* telephone (not pushbutton, not rotary,... a crank phone) from 189x and *still* use it to make and receive calls on POTS lines should say something about the state of telephone tech in supposed advanced nations like the US. We need to run fiber to every home in the nation with gov't footing a large chunk of the bill (otherwise no one would do it). And before anyone says anything about the cost or the lack of necessity, recall that the US interstate highway system was rediculously expensive and totally unneeded (traffic capacity wise) at the time. However, if you build it, they will come. I mean, could Alexander Bell have ever imagined his invention being used to surf pr0n on the web?
160,000,000,000 * 1022 = 163,520,000,000,000 bits
163,520,000,000,000 / 8 = 20,440,000,000,000 bytes
20,440,000,000,000 / 1024 = 19,960,937,500 kbytes
19,960,937,500 / 1024 = 19,493,103.02734375 Mbytes
19,493,103.02734375 / 1024 = 19036.233425140381 Gbytes
19036.23342514038 / 1024 = 18.590071704239 Tbytes
That is some bandwidth on ONE FIBER....
-- iCEBaLM
can it do 160gb/s over all 1022 channels?
Just provide the link to such a pic. If we are interested we could go look. And if we really don't care we won't. Ya gotta figure it'd be abused if we were allowed to post img. Imagine the load times if first psoters were allowed to put pics of the acclaims....
-cpd
Frequency or phase modulating a signal will cause the "single wavelength" to start spanning over a range of wavelengths. Vary the single wavelength's amplitude (intensity) alone, and it's still single frequency while carrying data too.
For those of you unaware, it's often been said (though I'm not sure how true it is now) that current data transfer methods may seem fast, but they still don't beat the "data transfer rates" of simply filling a stationwagon full of data tapes, and driving it yourself to the destination ;).
Alex Bischoff
---
Alex Bischoff
HTML/CSS coder for hire
Damn your eyes, I almost got it in!
Furthermore, the light in a fiber actually zig-zags down the fiber channel and does not travel straight down it. This also reduces the signal's speed from c.
--
Advertisers: If you attach cookies to your banner ads,
Time is Nature's way of keeping everything from happening at once... the bitch.
Vary the single wavelength's amplitude (intensity) alone, and it's still single frequency while carrying data too.
It doesn't work that way either. Your amplitude-modulated signal looks like:
s(t) = a(t) . cos(2.pi.f0.t)
Its spectrum can be deduced from its Fourier transformation:
S(f) = A(f) * delta(f-f0) = A(f-f0)
(where * denotes a convolution product.)
So the spectrum is as wide as the modulation's.
Nope, this is entirely wrong. The Fourier spectrum of any non-sinusoidal signal is > 0, which means that if say you are transmitting a square pulse it will not be single frequency.
Even if you are transmitting a pure mono-chromatic (i.e. sinusoidal) signal and turning it on/off (to transmit 0/1), the fact that you are turning it on/off will generate a spectrum. You might think it is strange, but it isn't. Read a chapter on Fourier Transform in a standard math text book and you will see why.
Varying the intensity of a light source creates "sidebands", the same as it does for RF. These "sidebands" are wavelengths slightly longer and shorter than the "carrier". What you see as an amplitude variation is really the interference of the carrier and the sidebands, as they slip in (high amplitude) and out (low amplitude) of phase over time. If you have a carrier frequency of F and a modulation frequency of M, you'll create sidebands at F+M and F-M. If you have really good filters you can suppress one of the sidebands and still carry all the information, and if you have really good frequency references as well you can ditch the carrier and only bother sending one sideband (you can use the frequency reference at the receiving end to supply the "carrier" for demodulation); this is how SSB radios work.
What does this mean for optical fiber? It limits how close together your "colors" can be based on how fast each one is modulated. The sidebands get farther and farther from the carrier as the modulation gets faster, and if the sidebands start clashing you get crosstalk and data errors.
--
Advertisers: If you attach cookies to your banner ads,
Time is Nature's way of keeping everything from happening at once... the bitch.
This has to make it through slashdot, so I thought I'd post it here.
link
"The 50-nanometer transistor - roughly 2,000 times smaller than the width of a human hair - is known as a ``vertical'' transistor because all of its components are built on top of a silicon wafer and its current flows vertically. In today's conventional transistors, which typically measure 180 nanometers, the current flows horizontally and the transistors are formed within the wafer itself."
"Although many researchers have tried to build vertical transistors during the last 25 years, the Bell Labs approach has several advantages over previous designs; it can accommodate ultra-thin insulating layers, and the channel and gate are closely aligned. Because the Bell Labs vertical transistor provides a solid foundation, it may be possible to add additional layers of transistors to silicon chips, resulting in so-called high-rise chips -- one of the holy grails of semiconductor manufacturing."
..we need a Lucent logo for articles.
/me
it's often been said (though I'm not sure how true it is now) that current data transfer methods may seem fast, but they still don't beat the "data transfer rates" of simply filling a stationwagon full of data tapes, and driving it yourself to the destination.
:o)
:o)
True, but the thing that this approch doesn't take into account is latency. (which, depending on the length of the drive, and how long it takes to load/unload the stationwagon, can range from ~5 minutes, to several days..)
If I'm playing a game of quake, I'll get my ass whopped if I rely on the stationwagon method
I think that (although the bandwith is considerably smaller) this is a step in the right direction.
Of course, they don't have a working product.
Still, neither is this fibre.
www.annovation.com
-- perl -e'print pack"H*","6e656d6f406d38792e6f7267"'
as well as Gowers and others, the fear of splitting infinitives has always been a superstition and rarely justified. Yes, you should try to avoid splitting your infinitives if you can convey the same meaning by restructuring your sentence, but avoiding the splitting of infinitives at all costs leads to much greater sins like splitting direct objects from transitive verbs, as well as even sillier superstitions like fearing to split helping verbs from their infinitives.
In any event, rules of grammar are not beholden to any single authority; either something is correct or it isn't, and incorrect things eventually become correct when enough of the population is using them. Of course, I will likely continue to assert the difference between "owing to" and "due to" until I die.
"If one is really a superior person, the fact is likely to leak out without too much assistance" -- John Andrew Holmes
www.nanovation.com
= 3) was in a frame. *sigh*
The article I wanted (http://www.nanovation.com/news_story.cfm?article
-- perl -e'print pack"H*","6e656d6f406d38792e6f7267"'
It sounds better if you write ".05 microns".
Check out Project Upper/Mute, an all-around awesome compiler fra
so now windowsNT and 2000 can crash at the speed of light?
- "If you tell the truth, you don't have to remember anything" -Mark Twain
Bzzzt.... try again. The UART's speed setting has nothing to do with the actual data rate on the wire. And don't say "but I have compression turned on." Doesn't help where you need it most -- downloading huge files. Also, 115kbps is far below even ISA's bandwidth.
Next time, do some research before you post.
And to make this post marginally on-topic: It's doubtful that these optical developments will speed up POTS modems at all, except in the cases that they cause old, crufty analog lines to be bulk-replaced with new digital lines. These optical developments are more important for backbones and long-haul connections, not connections between you and your CO.
--Joe--
Program Intellivision!
The same confused AC here again. I just thought of something. Let me guess my own answer. Amplifying a pure sine wave "exactly at the zero crossings" to eliminate sidebands is not possible because the amplification or deamplification cannot happen instantaneously without an infinite amount of power and since the amplitude change would always take some non-zero amount of time, the sine wave will deform over that period. The deformarions show up as other frequencies. That's gotta be it. Thanks everyone!
You responded to yourself before I could respond. I beleive that you are correct. Although I have to say that the idea of only changing Amplutude on zero crossings was quite smart.
We'll run fiber to the homes shortly, if something isn't developed soon that's even better. (Take a look at the recent Scientific American articles on the current candidates.)
Individual fibers to the home are a lot of bux. But a multiwavelengh fiber to the neighborhood and a passive wavelength divider (think prisim) and a bunch of short fibers to the house look like a good cost-tradeoff.
But having the government pay for it means you get to pay for whichever solution they chose at least three times - once for the install, twice more for the administrators. And the government will chose the wrong one. And the government won't even chose the best price/performance combo for the data rate they do chose.
Sure the government built the Interstates (kinda). And then they installed a 55 MPH speed limit - city, prarie, or deserted desert. Let them wire your home (or your kid's school) and they'll do it badly, expensively, and use it as a wedge to control the content.
The fact that I can plug in a *crank* telephone (not pushbutton, not rotary,... a crank phone) from 189x and *still* use it to make and receive calls on POTS lines should say something about the state of telephone tech in supposed advanced nations like the US.
Actually, it says more about good standards lasting a long time. Just like the Roman's choice of wheel spacing affecting cars, trains, and spacecraft components (that are shipped on trains) to this day.
The POTS standard is about getting audio from the switch in the city to the houses in the city and to the farms around it. The last mile of the audio part of that job hasn't changed materially since Bell and Strowager. A cheap low-tech solution does it, so why pay a bunch of bux to replace it with something that doesn't interact? Especially when doing so creates an administrative nightmare for no advantage.
Data is now hitting the wall on the capacity of the infrastructure designed for voice, so you need to replace part or all of it to go beyond.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
That press release sure implied that the two methods can be multiplied together, but I'd bet good money that it just ain't so. The 160Gbps test used a laser that was tuned for sending very quickly on a single frequency. The 1022 channel test used a DIFFERENT laser that was tuned for shifting rapidly between frequencies.
Also notice that Bell didn't mention a damn thing about the bit rate on the multichannel test. For all we know, those 1022 signals could be running at 2400 baud. And both tests were based on single lasers. Can they convince hundreds of these lasers to play nice and merge signals onto a single fiber? What the press release DIDN'T say leaves much to be desired, so far.
I think that they called them feelies in Brave New World...
If you're not part of the solution, you're part of the precipitate.
But funny! so please don't mark this down :-)
Winston Churchill is reputed to have answered, when accused of ending a sentence with a preposition, "That is something up with which I shall not put".
--
Infuriate left and right
Well, that may sound plausible to you but that's not what happens.
:-))
When you change the amplitude of a sine wave at zero crossing, you'll create a kink there since the slope of a sine at zero is proportional to it's amplitude. This discontinuity in the slope put energy in all frequency range.
But since this is all put together from sinewaves, which are nice and continuous, how the heck do you get a corner? In practice you can't, because it requires frequency components out to infinity (and infinite bandwidth!). But even an approximation of that "corner" requires other frequency components which add during the high-amplitude section, and then they all come together at the "corner" and suddenly they all subtract from the "carrier" for the low-amplitude section. All of these different frequency components mean lots and LOTS of sidebands. You can hear this in Morse code communications; if a keying network isn't set correctly and it cuts the signal off abruptly, you can hear the sidebands (key clicks) far away from the sender's carrier frequency. An over-modulated AM signal that "flat-tops" (clips)or cuts off completely causes broadband "splatter" which can be heard well off the channel too.
The way to limit bandwidth is to vary things smoothly, with no edges or corners. It may not be intuitive, but this is an area where your intuition doesn't get much experience.
--
Advertisers: If you attach cookies to your banner ads,
Time is Nature's way of keeping everything from happening at once... the bitch.
By that argument, any interface which converts digital signals into an analog transport and back with some form of modulation should be called a modem. Like your Ethernet card, for instance. We all call them Ethernet Modems don't we? Didn't think so. That said, the term "cable modem" is a reasonably accurate name for what it is/does, and since it replaces a traditional POTS modem, its name matches its intended purpose pretty well. --Joe
--
Program Intellivision!
You can derive the result of amplitude modulation from the trignometric identity:
sin(a) * sin(b) = sin(a+b) + sin(a-b)
By treating the carrier as a sine wave and the modulation as a sum of sine waves, and using the normal properties of real-number multiplication and addition you can work out the spectrum that results from amplitude modulating a carrier with any periodic waveform. It gets slightly more complicated for aperiodic waveforms, but the basic result is the same: A pair of sidebands, on either side of the carrier, that reproduce the spectrum of the modulating signal.
To send more bits on an AM carrier you essentially have to either modulate the signal faster (spreading out the spectrum of the modulating signal and thus that of the modulated signal) or modulate it more finely (using more bits to control the amplitude). The number of bits you can cram into the second is limited by the signal-to-noise ration of the channel (i.e. when you get near the noise your least significant bits get corrupted).
The Nyquist sampling criterion gives you a quantitative limit on this: If you have a band-limited signal, you can encode it with a number of bits-per-second equal to 2 times the bandwidth times the base-2 log of the signal-to-noise ratio, and reproduce it to within the the noise threshold. So that's the absolute maximum number of bits the signal can carry.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Someone recheck my math, this sounds way to high...
c=299792458 m/s
d=300000 m (Article stated using a 300km fiber)
bps=160000000000 bits/s (160 gigabits)
The Number of bits "stored" in the wire.
bit=(d/c)*bps
299792458/300000 * 160000000000 = 160110765 bits
or aprox 19 Mbytes worth of data is stuck in the wire. Talk about latency!
Read bell labs blurb on their web site here they do get a /bit/ more technical...not much though i'm afraid...
Sort of a tangent to what you were saying, but somewhat on-topic too... :)
/that/ much about fibre installations, but I haven't found any.
:)
There is actually another way of installing fibre that is a bit more future-proof than the traditional way that they are installing it in most situations today. I'm not sure of the name for the technique, but the company(ies?) that markets it calls it 'Air Blown Fiber'.
Installers bury conduits that have sub-conduits inside (looks like honeycomb when you look at the cross-section). Each internal conduit is independent of the other. After the conduit is laid out and in-place (buried etc), the installer shoots a 'BB' through each mini-conduit to make sure it is clear (using a big air tank and some special equipment).
Once it is determined clear and safe, the installer places a strand of fibre (protected but very light-weight - no kevlar!) into one of the conduits and 'blows' it through. Air pushes the strand(s) of fibre to the other side. It does this *very* fast, somewhere between 50-100FPS if I remember correctly. I was amazed to find that damage to the fibre strand is not an issue when they do this... it's actually quite rare. Installers I talked to said it was much more rare than the damage that occurs to the fibre in a conventional installation.
The idea behind it is: yes, you have to dig the initial hole and bury the conduit. But you'll never have to do it again. The conduit has many extra mini-sections for later expansion if you ever want to add more fibre (you can buy conduits with 2-72 mini-conduits, and each holds up to 6 strands of fibre), and if the fibre becomes outdated or destroyed for whatever reason, you just yank it out (using the same method in reverse if you want) and blow different fibre through. You can even blow a fibre through, remove it, then reinstall the same one multiple times with no problems.
I worked with a project at the National Department of Energy (Fermi) where this technique was used, as well as at the J. Paul Getty Museum in Southern California. Some universities are also adopting this method.
The amazing thing is that it's actually considerably less expensive than the conventional installation technique. Where's the catch? I don't know
Pretty neat stuff, sorry no URL. I'm sure a Google search on 'air blown fiber' would return some good results.
--SONET
Any fool can criticize, condemn and complain and most fools do. --Benjamin Franklin
This technology will enable 1/3 of the end to end system. If we look at the flow of information, it goes:
Content Server -> BackBone -> User
This technology will enable the backbone to move the bits between the server and the user. The day of evaluating your backbone provider by how many wavelengths they have in use, and how many are spare are coming fast. Optical switches, and one day optical routers will push this to speeds we can only dream of today.
The other two thirds are not keeping up though. Servers tend to be overloaded. If you work for an ISP where you have 100Mbps or 1000Mbps desktop connections to the network you know this already. Most servers, particularly big sites, couldn't fill a 10Mbps connection sending content. We're at a point where faster backbones carry more connections not faster connections .
The end user is even worse. Last mile problems persist, and will for a long time. The installed base of low quality copper is huge, and preventing even DSL from reaching many areas. New developments and business parks are not buring fiber during construction, even though the investment would be minimal. We should all be worried about how we are going to get high bandwidth to the home or small office.
Looking forward to the "killer apps" of the next 10 years, like Video on Demand I just don't see how it can happen. Care to dream of a server, or server complex that can deliver 2-5 million streams of video on demand? Will there be 20 million households with >= 10Mbit/sec access to watch those 5 million streams? I have my doubts. Will be Internet backbones be able to carry that traffic? I think so, that I'm not worried about.
Photons can and do leak out of fibers. And they can leak in. At the moment, crosstalk isn't an issue, but remember when they ran the first AC powerlines it wasn't either. Look what that stuck us with.
Briefly: The longer a path light takes, the longer it takes to reach its destination. So if some photons go right down the middle of the fiber, while others take a bouncy route off the sides, the edges of each pulse get stretched and blurred. Over short distances (under a kilofoot?) or at low speeds (under a gigabit/s) this isn't an issue, but it quickly becomes a problem when you move from the pansy-ass glass campus Ethernet into the serious bit-pumping that today's telcos do.
Each path that photons can take is called a "mode". I don't know why they picked that word. The fibers that a typical datamonkey is familiar with are "multimode", which means a wide core (125 microns?) of glass, which accomodates lots of bouncing. The fibers that your friendly telcos are burying are "singlemode", which is a much narrower core surrounded by more glass of different refractive indices. The idea is to straighten the light out and get it all traveling along the same path, so that pulse stretching is minimized. Still, the quality of your glass has a lot to do with it.
Now think about the construction of a typical optical cable. The stuff has to be somewhat flexible, to allow for installtion. In order to allow for some curving, the fibers/bundles aren't laid straight in the cable, but they spiral a little bit. This relieves stress when the cable is curved.
The problem here is that the photons are being asked to travel down this several mile long helix of glass, while remaining in as tight a group as possible. You can have one or the other, folks. Every time a fiber bends, a few photons wander off and do mischevious things.
Want to see something nifty? Go see EXFO's "live fiber detector" tool. Next time your phone service actually works, thank these folks for making sure the wrong fiber wasn't cut/disconnected by some technician working from outdated paperwork.
Errors happen for all sorts of reasons. Stray photons are actually the least of them. Imagine the sensitivity of the poor receiver that gets to turn all those bursts back into electricity. Can we say "sensitive to interference"? I thought we could. Most SONET systems achieve an in-service bit error rate well under 10E-10, and some are guaranteed better than 10E-13. How's that stack up against the consumer market, eh?
I need to go learn more about the funky laser those Bell folks used, it sounds like it's more of a color generating toy than a data pumping tool, and I bet the actual bitrates it achieves are laughably slow on each individual wavelength. I don't see DWDM disappearing any time soon.
Besides, none of it matters if the world ends next month. Tee-hee! You should see the disaster plans the telcos are putting together. Stock tip: Buy Dunkin Donuts shares, because telco techs are gonna buy 3 days worth of munchies on the 30th, because they can't leave their appointed sites after that.
Andrew S. Tanenbaum: Computer Networks Prentice/Hall, 1981 ISBN 0-13-164699-0.
-- Abigail