Slashdot Mirror
New Laser Data Transfer Rate Record Set At 26 Tbps
MasterPatricko writes "Scientists at the Karlsruhe Institute of Technology (KIT), Germany have published a technique to push optical data transfer rates to new levels. The article says, 'The trick is to use what is known as a "fast Fourier transform" to unpick more than 300 separate colours of light in a laser beam, each encoded with its own string of information.'"
I wish HDDs supported even close to this speed :(
So that one episode in Voyager where Seven of Nine makes a "Fourier Analysis" wasn't total bullshit?
"At those speeds, the entire Library of Congress collections could be sent down an optical fibre in 10 seconds"
This got to be a joke.
At those speeds, the entire Library of Congress collections could be sent down an optical fibre in 10 seconds.
Well played, BBC. Well played, indeed...
No one will ever need more than 1 Tbps.
The name that has everyone snickering in Germany: A shameless attempt at raising associations with the MIT by giving a German university a similar name.
Maggie Madsen: Maybe it's a precaution, because isn't that what we're doing? The signal pattern is learning, it's EVOLVING on its own, and you need to move past Fourier transforms and start thinking quantum mechanics...
Admiral Brigham: [with unintentional irony] There is nothing on Earth that complex!
Maggie Madsen: How about an organism, a living organism? Or some sort of DNA-based computer? I know that sounds crazy, but...
Is to use what is know as a "web browser" to quickly surf the "internet"
(v) To absolutely not pick. "The story submitter was unpicked from many dodgeball games growing up."
n/t
they used the American measuring system I'm familiar with: the Library-of-Congress.
At those speeds, the entire Library of Congress collections could be sent down an optical fibre in 10 seconds.
It surprised me that this came from bbc.co.uk
The best part is they were off by a factor of 8, since Tbps means terabits per second. Not so well played after all...
Sure, but what is that in Volkswagens?
Lets see what happens when we send Congress through this thing?
I said no... but I missed and it came out yes.
Serious question: It seems like it could be possible to use an infinite number of colors with interpolated laster on pulse modulation to transmit an infinite amount of information. Why won't this work?
If video games influenced behavior the Pac Man generation would be eating pills and running away from their problems.
I really wish HDD speeds could come even close to this :(
ugh.
mod me funny
I know, it sucks having to wait 1:20 minutes for my LoCs instead of the average 0.17 minutes - I'm so disappointed in this new technology, so yesterday!
OFDM is the same modulation technique used for WLAN.
Seems like the hard part is generating that wide variety of wavelengths.
The thing which is kind of critical is that you need an FFT that you can process at several trillion operations a second.
The FA says they are doing this photonically, which to me, is the cool part.
Absolute statements are never true
This comes from a lack of understanding of the field, but can someone explain to me what happens before and after, when transmitting data? The last I worked on hardware was with serial UARTs, so is someone willing to take a moment to explain better or to point me in the right direction to find more information on how it works?
The data is encoded into beams of light and transmitted from point to point. Upon "re-entry" into system at the end point, how is the information gathered and processed internally from this point? Does this rate of transmission get matched or have comparable mechanisms internal to the system? If data is moving at 26 Tbps, is it processed at anything even remotely similar, or does it fill a massive buffer?
Much appreciated!
This comes from a lack of understanding of the field, but can someone explain to me what happens before and after, when transmitting data? The last I worked on hardware was with serial UARTs, so is someone willing to take a moment to explain better or to point me in the right direction to find more information on how it works? The data is encoded into beams of light and transmitted from point to point. Upon "re-entry" into system at the end point, how is the information gathered and processed internally from this point? Does this rate of transmission get matched or have comparable mechanisms internal to the system? If data is moving at 26 Tbps, is it processed at anything even remotely similar, or does it fill a massive buffer? Much appreciated!
But why, you should have asked, is the BBC giving figures in units of LoCs?
They don't have books in Blighty?
http://www.nature.com/nphoton/journal/vaop/ncurrent/full/nphoton.2011.74.html According to the abstract, the key contribution is an optical implementation of the Fast Fourier Transform (which is pretty cool). They only tested their work using fibre, not just laser beams or w/e is implied by the headline.
wake me up when someone invents the Heisenberg compensator so we can beam the LoC in less then 10 seconds.
Boy, this will REALLY piss off the those entertainment fat cats...
They got upset because transferring a portion of their catalog no longer had a "cost" associated with it. What are they gonna do when transferring their entire catalog no longer even has a time-cost associated with it?
"The GPL is viral by design, like any good religion."
But why, you should have asked, is the BBC giving figures in units of LoCs?
Is the Library of Parliament (or whatever they call it) as big as the Library of Congress? If not, that might be part of your answer.
Why not just use a prism?
Sure, but what is that in Volkswagens?
German or American?
I work in this field so let me see what I can do to translate what's happening and which parts are actually interesting. I dont' know the details of the experiment I'm only going off the article's summary and assuming they're correct. (BTW, while I"m an expert in a tiny slice of this field, my "expert" level knowledge doesn't extend to the whole pie. So some things I say may be assumptions I've made and can be wrong. If you see a mis-statement please correct me.)
First off, we do fiber optics all day long for internet backbone communications. We even do "multi-mode" optical (different color/wavelength lasers) all day long but only for short cable lengths. Neither of which is article worthy. This thing called a "fast Fourier transform" is just math that is taught in school and is nothing even close to revolutionary, it is simply a fundamental mathematical tool of everything in this field.
First off let me give you the basic framework. When you're talking about sending data at these speeds and over these lengths, you can forget the idea that you're sending lots of data down the line in nice waveforms. The data is so distorted that significant energy is put into compensating and un-distorting the waveforms. Fiber optics at these speeds just doesn't work at all without heavy duty data recovery techniques. So we send down the line data, get back garbled gibberish, apply techniques for removing errors and you can recover your data stream.
So typically when we do "long haul" fiber (> 1k or so) we do single mode fiber, this means a single frequency or color (remember your physics, each color is a different frequency of light). This is because different frequencies of light travel at very slightly different speeds down the fiber and if you have long enough fiber this difference in speeds becomes significant and starts to harm your ability to regenerate the information. Additionally one frequency can cause noise in another frequency band so keeping things to a single frequency makes things more stable at long haul lengths. This is why "traditional methods to separate the different colours will not work".
So Professor Freude and the article:
There are two steps forward here:
1. He's using a single laser to create different frequencies of light. I don't know if this is a common technique or not. I've typically hear of different colors of light being generated by different lasers but I am not an optics guy so I'm not sure.
2. He's using an optical method in place of a Fast Fourier Transform (FFT) instead of silicon that somehow helps him decode the data. An FFT mathematically converts from frequency domain to time domain so maybe he's just using a prism or something to separate the different frequencies as a pre-processor step and then pumping this into his processor, but I can't tell.
So Professor Freude and team and working on making "multi-mode" work at long haul. This is typically not done today so that's the step forward and since you can pack more information into your data stream if you include multiple frequencies, that's a nice win. but of course research success does not necessarily equal a marketable product.
(Again, I am not a guru here, so if you are, please politely correct any mis-statements I've made.)
d
all language nazi's will burne in heil!
If spending profit to improve the worth of your company actually resulted in higher short term stock prices, maybe... Currently it does the opposite.
Want faster Internet? Too bad, that kills our stock prices -- the investors won't allow it, we have to charge the customers more without letting the investors latch on to that profit so that we can spend it on improving our services. Too bad it's illegal to do that once you go public.
If it costs anything beyond standard maintenance, it's not coming to the giant ISPs. That's what was so great about the smaller ISPs, they could risk more to provide better service and eventually end up with better stock prices in the end. The big ISPs of today are so immobile, the transition would take too long and cause the short-sighted investors jump ship.
Seen it happen many times, don't take my word for it, do your own damn research if you actually care -- but you don't; so let's just bitch about not having fiber to the home, and make excuses for why our speeds suck in the US compared to smaller places, while marveling at new advances that will never reach us.
AHHAHA That's the standard data density unit, LoC/Football field
Here's a fast furry transform: http://media.fukung.net/images/5390/furries_internets_go.jpg
And how many coconuts?
Ughh. way to gut the summary I had written.
Anyway here's the original paper in nature photonics
http://dx.doi.org/10.1038/nphoton.2011.74
Basically the advancement in this technique is to take several incoming tributaries of data and use an optical FT method (OFDM) to encode into a single laser for transmission. Apparently the encoding/decoding is simple and low-power enough to fit on a silicon chip so this technology seems very implementable, just by upgrading the hardware on either end of existing fiber optic cables.
The BBC article with their 0.1 Libraries of Congress per second was great too ;)
I'd tell a UDP joke, but you may not get it. I'd tell a TCP joke, but I'd have to keep repeating it until you got it.
The British version of the Library of Congress would be the combination of the National Archives and the British Library. Given the BBC partly functions in the manner of the FCC, you may need to include the BBC archives in there as well. The sum total of these three collections probably exceeds the Library of Congress but there's no collective name for all of this information.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
What's the point if ISP caps are 4 - 5 GB per month?
Wow, the summary is completely misleading and tells you nothing about the actual news. It reads like they just figured out that they can use multiple frequencies to send additional bands of data. Everybody with a slightest clue about how optical data transmission works knows this already. The news in TFA however is that they have figured out how to encode this data with a single laser instead of one laser per band making it enormously more economic. A car analogy would be if someone had figured out how to make cars fly without wings, and the summary would be "a new vehicle pushes transportation to its limits. by sitting in a container with an engine one can travel much faster than walking." Facepalm.
The BBC doesn't function anything like the FCC. The body you're looking for there is OFCOM. They handle frequency allocation*, what little broadcast content regulation we have**, the sort of thing the FCC would handle in the US. The US doesn't really have a counterpart to the BBC. The closest would be PBS, but they differ in some very important ways.
*Much like the FCC, they'll occasionally say something about public benefit before auctioning it all off to the highest bidder.
**As we don't have that many radio broadcast channels, and cable/sat are outside OFCOM's direct oversight, they don't have much to do here. They occasionally send a strongly-worded letter.
Conventionally, there's 10 libraries of congress that form parts of the Library of Congress, so in this case you would look at a speed of 1lcs or 0.1 LCs.
26 Tuna Boats per Shark?
...Who read it as 26 tablespoons?
And AT&T will buy this technology and lock it away for the next 40 years then complain that they have to raise internet user rates to cover their yearly costs, a la the oil industry to alternative fuel sources.
No it doesn't, not even close to exceeding it.
You can derive a set of sine waves that can reconstruct the original, but FFT will not break it down to the original sine waves. Only an orthogonal set. Two sine waves as you describe may be simple to pull apart reliably. Two sine waves is like factoring a prime.
The article even calls the data "orthogonal frequency-division multiplexing (OFDM) data streams". The authors are aware of this problem. That's why I pointed out this isn't a simple FFT solution.
Optical FFT means you don't have to decide which windowing function to use, or the size of the window. If you have any experience with FFT the implications will be obvious. If not, try any sound source and any FFT display software with different sizes and window functions, and you'll see you get different results depending on the settings. Then imagine not having to choose one.
I further speculated that selecting a specific set of input frequencies made it easier to separate. This part seems pretty obvious, but if you don't know FFT results are orthogonal, the reason won't make any sense.