W.r.t. DIDO,

Phased arrays and GPSs get very close to the accuracy req. I would imagine that the computing power required would be tremendous.

However, it does show that there are ways around today's supposed limits.

My original point though was that spending all this money without the desire in the market to pay for it is reckless and would not be tolerated in any private company. Labor's solution is to try and eliminated the competition. Given the strides that wireless is taking, this might only be partially possible.

SM:"I would imagine that the computing power required would be tremendous. "

Undoubtedly, which is why he proposes having it done at the centre. I really like the idea of encoding separate transforms on the signal for each transmitter. Rstuart seems to have missed the way in this makes it possible to break Shannon's law: the modulation is effectively "virtual", in that it isn't encoded on any one carrier, but is an emergent property of the signal that is processed within the receiver. The wireless itself never gets close to encoding the amount of data that was first encrypted, just a small fraction. It would even be possible, I imagine, to have the signal built up from individual components rather than having to coordinate all the phase-shifted signals arriving at once.

The more I think about it, the more I reckon he's really onto something important.

@Shadow Minister: Phased arrays and GPSs get very close to the accuracy req.

Nah, it would have to be in the cm range, GPS is 1000 times worse. But I imagine the towers sending out a sync signal and the receivers reporting what they saw could be made to work. The towers would have to synchronise their transmissions within the picosecond range. Effectively what you are doing is turning the mobile network into a sub-centimetre terrestrial GPS network.

For someone walking the tower's would need to accurately and collectively decide on their position every 100 milliseconds or so to keep within the required fraction of a wavelength range, which I imagine is impossible. So this is fixed wireless only. But since it is fixed the convolution matrices can be prepared ahead of time, and if you can distribute the effort across the towers you might keep the computational effort tractable.

Bad multipath interference would cause havoc with DIDO just like as does with GPS because the changing signal strengths on the various path's makes fixed look like mobile.

The data links between the towers would have to be enormous because every tower if sending at 1Gbit/s every tower would have to know what every other tower is sending, meaning if 6 towers are visible to a receiver all those towers would have to see each others traffic. Then they have to negotiate sending times and strengths. 100 Gbit/s would cover it the typical 6 towers that see each others signals. Doing this for 100's of towers is obviously a fool's errand.

So given the computational requirements and bandwidth requirements, doing this in a single data centre even for a city like Brisbane as suggested in the DIDO paper isn't possible. It only scales if you distribute it across the towers, but my guess he currently doesn't have have the algorithms to do that.

If he does manage to pull it off in a decades or two's time the pay back is a 6 fold improvement in the system spectral efficiency of the downlink, and no improvement on the uplink.

@Antiseptic: Rstuart seems to have missed the way in this makes it possible to break Shannon's law: the modulation is effectively "virtual"

Once again Anti: you can't break the Shannon-Hartley theorem. It isn't a law. A law is a physical model for the Universe that has been confirmed through repeated observation. This is a mathematical theorem. Assuming the workings are correct it can't be wrong.

Think of it like this. The area of a piece of land is Length x Width. This is mathematical statement of fact, just like the Shannon-Hartley theorem. An architect can then come along and build a multi storey block of flats on that block of land. That doesn't mean Length x Width has been broken.

So Shannon tells us how much you can send within a given Signal To Noise ratio. The peak link spectral efficiency of 4G (ie where they get 1Gbit/s/hz) assumes a SNR of 20db, which is pretty high for wireless. In fact its pretty good in general. Even on HFC a SNR of 30db is considered acceptable, on DSL the 20db is considered OK. HFC is shielded from background noise. Wireless isn't.

So - Shannon's law can't be broken, and single 4G link is already assumes an SNR close to the limits of what background noise will allow. That doesn't leave a lot of wriggle room. I guess reducing the cell size effectively is the equivalent of building multi story blocks of flats. The current 6 fold loss for system spectral efficiency does look juicy and that is what DIDO is targeting. But to put it into perspective, reducing the cell radius from say 5km to 200m would achieve a factor of 600.

The theorem is being broken because the data is being encoded on multiple streams and decoded within the receiver. No individual signal is carrying all of the data.

Further, because the data is encoded as a custom waveform at the central processor then reduced to components than can be modulated satisfactorily the total data rate is not dependent on the modulation, but on the initial pre-processing. It's effectively a very clever form of compression. The S/N doesn't really come into the equation at all, since the final data is a composite of lots of different sources, meaning error correction is trivial.

As for the location business, it's simple triangulation. Any body with a Yagi has been able to do it forever. It's trivial for a distributed phased-array with lots of elements and even with only a few elements the error would be down to the limits of the timing system.

I'm sure there are lots more clever tricks yet to come. I have no idea what can't be done.

Rstuart,

As a matter of interest GPS is capable of sub cm accuracy. However, as most of the GPS satellites are US, there is a random signal mixed in to prevent them being used for cheap missile guidance for non friendly countries.

Survey systems can get accurate readings, they just need to average over a few minutes.

US weapons such as cruise missiles have access to the clean signal and can fly a missile through an open window.