Reed's Talk. Reed is the radical mad scientist of open spectrum, who maintains that spectrum is not scarce, except due to a policy framework that is obsolete in the current technological reality. He's the appropriate opening keynoter for this conference.
Does spectrum have a "capacity?" This is the key question. If spectrum is limited, then there's a reason for apportioning it carefully. If it doesn't, then making spectrum scarce is scary First Amendment country.
The radio tradition developed from 1900-1950. In the beginning, all radios received all frequencies. Resonant systems allowed users to divide spectrum for different apps. Different frequencies had different properties — low-frequency would go around the world, high frequency would bounce off the ionosphere. Increasing power lets you go farther.
Shannon invented information theory in the 50s, and invented the bit — a measurement of info regardless of the form it takes.
C = W log (1+(P/N0W)), where C = Capacity in bits/sec; W = bandwidth in Hz, P = power in Watts and N0 = Noise power in Watts/Hz.
Channel capacity is roughly porportional to bandwidth and log of power. Capacity is analogous to bandwidth, but bandwidth is not the same as capacity.
This is only part of the story, though. The original theorem is a simple model consisting of a sender, a receiver and noise, with no consideration of geography and other transmitters (we treat other transmitters as noise).
Interference: this is the other key question. If interference exists, we need strong regulation to limit it. If it is an artifact of technology, then we're doing it all wrong. Regulators describe interference as damage or rivalry. From a physicist's perspective, radio interference is superposition — two radio waves floating through space don't harm each other, but they do add to each other.
He's showing a beatuful physics model of a wave tank, showing how transposition works. The simulation is strangely violent, like the proverbial storm-toss'd sea crossed with a 1980s VR snakeoil conception of "cyberspace." WELCOM TO CYBER SPACE.
A distant and strong transmitter can be received, even if there is a closer, but low-powered transmitter on the same channel — the low-powered signal doesn't stomp on the high-powered signal. Directional receivers make it even easier to disambiguoate overlapping signals.
The point: no information was lost, even though there was superposition. Receivers may be confused, but that's a systems-design/architectural issue, not a physical inevitability.
The big policy question: Where does "interference" occur, and who causes it. When a new radio is added to the system, does it displace capacity — do your 802.11b speakers fundamentally reduce my capacity to use mine? Even if the answer is no, does this impose costs on others, who will have to use more effort to accomplish the same amount of communication?
The problem is static partitioning is that demand is dynamic, so the regulatory framework creates wastefulness in space, frequency in time. We leave gaps between TV channels because receivers may not be able to disambiguoate the signals.
The Slepian-Wolf theorem: Frequency partitioning is optimal only when the bandwidth of each band is proportional to its power at each receiver.
Transport capacity is an important measure of radio-network capacity. Add of all the useful bits of received by all the receivers in a system and that's the capacity of the chunk of spectrum they have. Under static partitioning the best you can do is assume a fixed capacity and divide it by the number of stations.
An architectural improvement is hop-by-hop repeating. Many paths can operate concurrently, and energy/bit is reduced by 1/(number of hops). What happens to a repeater network's capacity as radios are added? The more stations you have, the less power and so the less receiver confusion.
The capacity of this network is not constant as you add radios — rather, you add capacity with every radio you add, which partially offsets the drain that the radios create.
But there are other ways to think about this. Spatial organization takes into account the spatial relations of stations, and as the power of any antenna drops off exponentially over distance, you get more capacity from the same spectrum. Directional antennas provice fixed allocation with much greater spatial multiplexing.
Smart antennas (phased-array?) provide dynamic allocation — a single smart anternna can receive two differnt signals in two directions at once. They can dynamically select direction and frequency — blends the antenna and modulation.
Another approach is spatially organized waveforms, such as Bell Labs's BLAST, which exploits multipath in the environment to increase capacity, using a technique analogous to ghost elimination in television. Related ideas: MIMO systems, cooperative signal regeneration.
Signal is assumed to decrease at 1/(distance^2). But that's a Platonic ideal. In the real world, walls, trees, buildings and other chazzerai create much worse propagation characteristics, which makes capacity go up! If radios can adapt to these circumstances (say, by routing within a room when the signal only needs to be in that room, and by routing around the wall when they need to), the reduced propagation of competing signals is good news!
It's counterintuitive:
- Adding stations increases capacity
- Multipath increases capacity
- Repeating increases capacity
- Motion increases capacity
- Networks reduce total energy to acheive the same capacity (safety, battery life)
- Dynamic sharing decreases latency and jitter
But! Does adding new radios impose other costs? Well, sure. It makes legacy systems obsolete — that's a lot of TVs to throw out. But legacy should never preempt innovation. There are three ways forward: throw out the old stuff; uppward comaptible evolution, where newer systems compensate for older ones and upgrade existing systems.
Software-deinfed and cognitive radio are systems where a computer and a DSP generate and recognize waveforms — hook a computer to an antenna and that's it. This can simulate all the old kinds of radio, and generate new kinds of waveforms that can "dance between the raindrops," communicating in the spaces left by legacy systems. MEMS and nanotech promise dynamically reconfigurable antennas that can adapt to their environments. Taken together, this is a set of technologies for enabling evolutionary progress.
Also worth considering: ultra-wideband (UWB). A coded sequence of extremely short high energy pulses to achieve high-rate comms — very low average energy, and can coexist invisibly with many radio services.
Related to this is the infotech of security — how do you authenticate who's speaking, how do you ensure integrity, etc — this is all well-understood stuff from the crypto world. Dynamic and adaptive reconfiguration enchances security against attack and robustness against failure — you can spread your comms over spectrum, space and time.
