Sounds good. Let's hope it's not another of those amazing inventions that are going to be ready in "two or three years" and then nothing more is heard of them.

The problem may be that this would open the door to new, bandwidth hungry Internet applications that just slow the whole thing down again or am I talking nonsense?

Hi hope bill doesn't read this topic. Is he on 1 GBps or is it 2 GBps? ;-)

Anything that speeds up the speeding up of our connections is a good thing. We all still lag along way behind Japan and Korea. Almost makes you want to emigrate.

<sigh> ;) One of the major mobile phone companies in Japan was running tests on 4G network speeds and achieved [nttdocomo.com] a 2.5Gbps download speed while traveling at 20km/h. But that's last year's news. This year they claimed they hit 5Gbps in tests...and this is wireless phone connectivity we're talking about here.

My landline is fiber optic, and they only offer 1Gbps connections for home users at this time.

The technology may be useful but is unlikely to be revolutionary in its effect. Channel capacity is strictly limited by the formula C = W Log₂(1 + P/N) where W = bandwidth and P/N is the signal to noise ratio.

No matter what you do, this law cannot be broken - it is mathematical fact so, to increase the channel capacity it is either necessary to increase the bandwidth (which is not what is being proposed here) or it is necessary to increase the signal to noise ratio. However, to increase channel capacity by 100 would require signal to noise ratio to increased by 2¹⁰⁰. I doubt that this new technology can achieve this.

To be blunt, this story sounds like hogwash to me. If speed can be doubled (in the real world not the lab) that will be a good effort.

Kaled.

C = W Log2(1 + P/N) where W = bandwidth and P/N is the signal to noise ratio.

Ok, I'm patenting this in it's application as a method and device for ranking blog relevance!

I think the original citation addressed the improvement of S/N using some kind of predictive method. So, some improvement in C is possible. That's all I got figured out.

If he is attempting to predict noise, he isn't the first. A post grad in my dept at university started his thesis on this very subject - but being a smart chap, he didn't ever write up (because he figured out noise can't be predicted).

I suspect the proposed system attempts to compensate for crosstalk by predicting its effect based on knowledge of all the driven signals in a bundle. It's a nice idea but likely to be massively expensive in terms of cpu power and could only work when downloading data (not uploading) and any uploading of data in the bundle would tend to negate possible benefits by throwing in wholly unpredictable crosstalk noise.

By "downloading data" I mean data is transmitted by the exchange.
By "uploading data" I mean data is transmitted by the modem.

Kaled.

And of what use is this?

I have DSL 16000 but average download speeds are usually at 300 kbs because usually the bottleneck is the server not the dsl connection.