Aussie Claims Copper Broadband now 200x Faster

SkiifGeek writes "Winner of Melbourne University's Chancellor's Prize for Excellence, Dr John Papandriopoulos could soon find himself the focus of a number of networking companies and government agencies interested in wringing more performance from existing network infrastructure. Dr John developed a set of algorithms (US and Aussie patents pending) that reduce the impact of cross talk on data streams sharing the same physical copper line, taking less than a year to achieve the breakthrough. It is claimed that the algorithms can produce up to 200x improvement over existing copper broadband performance (quoted as being between one and 25 mbit/sec), with up to 200 mbit/sec apparently being deliverable. If the mathematical theories are within even an order of magnitude of the actual gains achieved, Dr John's work is likely to have widespread implications for future bandwidth availability across the globe."

The limit has been exceeded....

[mks113]

And we learned, in Electrical Engineering, that the theoretical maximum bandwidth for a phone line was 2400 bps.

Using basic bandwidth calcs for voice (500 to 4000hz?) and imposing a modulated signal inside that, the distortion created by the physical arrangement of the wires would cause the limit.

I'm glad that some people aren't scared off by theoretical physical limits.

(That was in about 1986, A Hayes 1200 baud modem was an amazing piece of equipment and cost about $700)

Static vs. Dynamic correction

[G4from128k]

I suspect that his algorithms require very very careful analysis of the cross-talk environment to remove its effects. The result is a very high-gain function on the high-frequencies to correct for crosstalk and modulation effects at high bandwidths. That's fine in a controlled environment, but won't work if the amount of crosstalk varies dynamically. Temperature, wind, rain, ice, humidity, and squirrels all change the crosstalk characteristics.

Re:Metaphor please

[Omnifarious]

Your post is labeled informative, but it is so filled with jargon that is missing any nice links to references that explain it that I find it quite unhelpful.

Not for distance

[Russ Nelson]

Not for distance. You're still subject to the 18Kfeet (max) limitation imposed by the resistance (gauge) of the wire.

Re:Across the globe == developed nations

[CastrTroy]

But the wireless spectrum is very limited. For the first 100,000 people or so on the wireless network, it could probably remain pretty fast. But try running all the computers in New York City on a wireless network, and see what kind of speed you can achieve at each node. So as a starter point, to get the first few people in the country on a network, or to connect a small village, wireless networks could prove extremely useful. However, if you want to take all the network communication in a large city and try to accomplish that without wires, you'd probably fail very quickly. Don't even start to mention current cellular networks, because there's still a lot of wires involved.

Re:Metaphor please

[arivanov]

Quote from the article: one wire is wirelessly pushing its signal on to another wire (a phenomenon known as crosstalk), a microprocessor could use the noise from the crosstalk to do error correction on original signal...

Err... That is exactly what I described (without even reading the article).

IMHO not patentable due to being bleeding obvious. The sole reason it is not being done at present is that till recently it was impractical. You just about handled one wire with one chip. Handling a bundle and running a "cool" algo on them was simply beyond what the electronics could do.

As far as the likelihood with 3G: 3G does something quite similar using the signal in a feedback loop. As a result echoes from buildings and reflections from earth (aka multipath) which in other technologies decrease your signal to noise ratio are used to increase the signal to noise ratio.

For example you have the following sequence of bits: 1 1 1 0. Once you get past the first 1 you get the same sequence arriving reflected from a different source. As a result you get slightly better signal to noise on the next 1 1. After that you have a 0. It overlaps with a reflected 1. As a result you get garbled input. If you use a delay shift register and optimise where do you need to add your signal from 1,2,3,4 units of time before that to yourself you can actually eliminate this and improve your signal to noise based on reflections instead of garbling the signal. In addition to that the output of the filter is used also in guess what - power control: telling the mobile to adjust its power.

What this chap is doing is doing the same by applying signal from wire N to the signal from wire Y as a digital filter. Which means exactly what I said - in order for this to be of any use all wires in the same bundle should be handled by the same ASIC. I should probably do the math but they should probably also run the same line protocol. If you have a third party provider running an ADSL in the middle of your "precious" DSL2 bundle this nice scheme fails.

Pity actually, while not particularly original this is a cool way of using a well known existing way of improving signal to noise ratio (including the power control part of it).

Re:Metaphor please

[skarphace]

I remember the days of telephone crosstalk, but I never got the impression that it was RF-related; I always assumed it had something to do with ground loops (are there ground loops in balanced telco?) or improper balancing or things like that.

Crosstalk [wikipedia.org] is much broader then what people are insinuating here. Crosstalk can be RF, groundloops, bare wires douching each other, etc. Simplest definition is when one signal interferes with each other.

I also have developed the impression that the biggest speed barrier in copper is reflections, not crosstalk, although I suppose that's more true of CAT-5/6 than of untwisted telco wiring, which is what this invention is supposed to work with.

Reflection(in data networks atleast) is rarely a problem anymore. It was in the 70s and early 90s, during hub/patch days but not much anymore. This essentially occurs on an unterminated line when it is left connected to the network. It would travel to the break in the line, hit the end, and travel back towards the source destroying everything in it's path.

Today though, most switches should not allow an unterminated line access to the rest of the network. Should just ground out everything coming in from that line. Probably simply because it doesn't have a source/destination MAC address that makes any sense. Those handy little Fluke handsets use reflection to find the break in a faulty line. Hopefully someone can correct me if I'm wrong.

Even more confusing is TFA's use of "the same physical copper line"; if they really mean that (a continuous strand of copper), then they can't be talking about crosstalk. But maybe they mean "the same cable (set of wires), and cables are made of copper".

Crosstalk can be on one wire, or a cable, air, or whatever. I believe they are talking about multiple signals on the wire(multiplexing [wikipedia.org]) which could interfere with each other.

Re:200 mbit/sec

[Kingrames]

Reminds me of my own solution to the Gabriel's Horn problem.
That's the "infinite surface area, finite volume" problem, if you needed to jog your memory.

My teacher explained the paradox by saying that it would be like something that would take an infinite amount of paint to paint the inside of it, but it would be able to hold a finite amount of paint.

I quickly pointed out that this was only true if paint were not made out of molecules. At some point, you can no longer put any more paint on the surface, because the molecules are too big to place there.

Of course, if you were to ignore that fact, and only look at the mathematical side, I also pointed out that you could paint the surface of the horn with a single very thinly sliced proton.