|
Aaron Schwartz wrote an article on logicError titled Open Spectrum: A Global Pervasive Network. I'd like to address a few inaccuracies in his article, and then tackle the bigger issue.
Aaron writes:
How much information can we send over a radio station? In other words, what's the capacity of the spectrum? This would seem to be a very important question for anyone interested in radio, but the fact is we just don't know.
On the contraty: we know this quite well. Shannon proved his channel capacity theorem quite a long time ago. What we don't have, to this day, is the answer to an arguably more important question regarding the aggregate capacity of a network of N nodes.
The early models assumed that capacity was the same as the number of stations used. In other words, information rate was proportional to bandwidth.
Capacity, as defined by Shannon, has a very strict mathematical meaning. It would be more accurate to ssay that "the early model talked only on networks in which communication is end-to-end, with no intervening nodes-in-the-middle".
(This misconception explains why most people call the speed of their Internet connection their "bandwidth".)
The reason people call their Internet connection "bandwidth" is because they don't understand what bandwidth mean, and "information rate" would not sound as good. They are not far from wrong, BTW, because it *is* the phone network's bandwidth (restricted by filters intentionally put in by the phone company) that is the limiting factor. If you dialup via a moedm to your ISP, the capacity theorem applies to you. Modems today are so good they virtually achieved the limit set by the capacity theorem; that's why, while all other computer equipment gets faster, 56K modems have been with us for a few years, and won't go away until the phone network is changed (and it won't).
[...]
More research has found some other interesting results: What if we spread our communications across the spectrum? Capacity goes up. What if we spread them out across time? Capacity goes up. What happens if we have multiple paths to transmit? Capacity goes up. What happens if the transmitters move around? Capacity goes up. Every place research has looked, they've found that if they do that capacity goes up. And the research is far from done (sadly, because few people are doing it -- more on that later).
As much as I like this description of "everyone thought X, then somebody showed that Y", reality is not that simple. The article uses a few well-defined terms (capacity is cheif among them), but treats them as if they mean something other than their definition. Such language, probably chosen to make the article understandable by the general public, is bound to be inaccurate, and to lead some readers to false conclusions.
Spread spectrum (CDMA) techniques do *not* increase capacity. Their power lies in the ability, unlike other multiple access techniques (FDMA and TDMA), to dynamically achieve better quality when transmitters suddenly stop. While the older FDMA and TDMA techniques achieve efficiency by using central coordination, CDMA need to nothing to adapt.
Similarly, spreading across time gives you no extra capacity, pure and simple. In fact, one of the most striking results in information theory is that everything is connected with the relation Eb/N0, where Eb is the energy you put to transmit a bit and N0 is the noise level. It doesn't matter if you send a strong signal with 50% duty cycle, or a 50% weaker signal with a 100% duty cycle. What counts is the energy you put into sending that bit.
Spreading has other desirable attributes, such as its inherent "resistance" to duplicated shadows reaching the receiver. That's why multipath, a major source of headache, can actually be used as an advantage by smart CDMA receivers.
Capacity also does not go up when you have multiple paths. If it did, makers of microwave antenna dishes (which are built to have narrow beams) would go out of business.
It is, however, true to say that when the transmitter and the receiver have no directional communication means (that is, they are forced to use omni antennas, like your cellphone), multipath can be used to *realize* the capacity. Again, note that capacity is the absolute achievable maximum, and all the hardware and brainpower we throw at it can only help us achieve it, not increase it.
(As a sidenote, in an optimal network, you always know where the "other side" is, what spacial emission pattern is best, and have a phase array antenna to actually generate that signal. This makes multipath another channel characteristic like, say, different noise levels at different frequencies, and makes the solution "simply" another type of coding, this time spacial coding.)
[...]
The case went to the Supreme Court who decided, based on the flawed but intuitive model above, that the FCC was necessary. Spectrum is limited, they were told, if everyone tries to speak, then no one will be able to. The FCC is required simply by the way that radio waves work.
But as we have seen, that's simply not true. What if the Supreme Court had known this? Would they have declared the FCC unconstitutional?
Actually, it was absolutely necessary, or chaos would have ensued. You see, collaborative networks, spread spectrum, and all the wonderful achievements of modern digital communications, were simply not possible at the time. And even if they were, radio stations (and the like) strongly depend on the economy of "thin client" models: You have sofisticated, expensive, radio equipment at the broadcaster's location, and everyone can afford to buy cheap radio receivers. Building a radio AM receiver is a piece of cake. It costs almost nothing. Symmertric network communication equipment, however, is an altogether different thing. Only now we reach the price-performance point to even think of collaborative networks.
Don't get the impression I'm shooting down the idea. I'm not. What I don't like is the way the idea is presented. Having discussed the envelope, let us discuss the contents, and the idea is certainly worth discussion.
Aaron is entirely correct in saying that we can achieve better capacity, and the way to do it (as he correctly points out) is to use the "network effects". Node cooperation can improve aggregate capacity, somewhat analogous to partitioning a large LAN into smaller LANs to improve aggregate "bandwidth" (but only somewhat).
There are other technical issues with collaborative networks. For example, we don't have today the routing technology to make this happen. Another example would be the latency issues with such a network. I think, however, these issues are large enough to punt them to another time ;-)
I'm not at all opposed to having more bandwidth opened up for collaborative networks. I think it's a good idea with lots of potential. I also think that, when all's said and done, it won't be the basis of a global network.
The reason is social. For such a network to work, you need everyone's cooperation. Everyone must limit their transmission power. Your neighbors must agree to let you pass information through them. There must be no "bad guys" out there to jam your signal (make no mistake: it *is* possible to jam spread spectrum signals).
I don't see all this good will spontaneously happening. Is your web server being attacked today? Now suppose your internet connection is exposed to any bozo driving along with a strong-enough jammer. Who do you blame when you don't get the level of service you need? Collaboration looks like a nirvana: pay one-time for the equipment, and get Internet access for free. I don't believe in free lunches.
|
![[Valid RSS]](images/valid-rss.png "Valid RSS"){kind=small}
