Slashdot Mirror
Germans Reach 360 Mbps in Mobile Network Tests
povvell writes "German telecomms giant Siemens has managed to hit speeds of an astonishing 360 Mbps in field tests in the centre of Munich using 'orthogonal frequency division multiplexing (OFDM) and the so-called multi-hop technology'. This is not the only demonstration of OFDM producing super fast wireless speeds, as other companies are also working on variants of the technology. It surely can't be long now before we're all streaming the latest blockbuster movies to our laptops on the commuter train home?"
Okay, let's get some raw movie stats. Assume plain RGB pixmap flipping at 24 frames per sec, movie size. That's 720x480, three bytes per pixel.
That's about 1 Megabyte a second, or 8 Megabit. Add another 256 Kilobits/sec for audio (Mp3, Vorbis, or AAC, anyone?) and that's 10 Megabit and change.
Isn't Divx compression good?
--
# Canmephians for a better Linux Kernel
$Stalag99{"URL"}="http://stalag99.net";
I remember Cisco offering a product six or seven years ago that did vectored orthogonal frequency division multiplexing. It could do 45mbps non-line of sight as point-to-point or as a unidirectional 28 channel T1'looking setup al la Cisco 2600 WIC Cards. I wonder how this is different.
OFDM is a modulation technique (i.e. it turns 1s and 0s into radio waves) and CMDA is a multiple-access technique (i.e. it lets multiple radios share a channel), so they're really orthogonal.
Well, Claude Shannon showed that, with a perfect modulation and error correction scheme, you could only push so much data over a given communications media, with a given amount of bandwidth and SNR. If you want more, you have to either
Since background noise is not controllable, they would have to be doing one of the first two (effectively increasing radio pollution), or overcoming inefficiencies in a previous modulation scheme.
Anyone know how close the various 802.11 standards are to the shannon limit?
-jim
It's even smarter than that! Your little rect in time domain is an inf. sinc in the frequency domain. Of course, after a certain length it dies down to a negligible strength (call this point B). If you wanted to modulate another pulse, to guarantee they will both be exactly recoverable you need to modulate the new pulse up to 2B.
OFDM basically take advantage of the fact that the signal is digital. Instead of modulating the next pulse at 2B, you modulate it so that the next pulse is centered over the first zero crossing of the first pulse. Normally, this would cause horrible aliasing, but since you know the shape of your input data, all you care about are the values at the origin and the zero crossing. You don't care about recovering the two original signals exactly, the value at the origin and the first zero crossing give you enough information to reconstruct them. Aliasing be damned!
This takes advantage of the fact that simpler data is more resistant to noise. If you know what you're sending is a 1 or 0, then the waveform can be horribly degraded before it makes a difference. Contrast this with simple voice data, where a deformation in the wave can't be repaired (you don't know what it should look like). In this case, your encoding scheme introduces noise it knows doesn't matter in order to save bandwidth.
Of course, this is also a form of quadrature multiplexing, which lets you send two signals at the same carrier frequency but differentiate them based on the phase. So that gives you twice the transfer amount you'd normally get above.
(yes, I'm in a communications theory course right now)