Flit

Cell phone musings

Being a glutton for punishment, Steven Den Beste's provoking some musings again. B-)

He's back from vacation and talking about cell phones among other things and giving very good information about why the system doesn't drop certain traffic and expand basic capacity in case of emergency. One thing that isn't clear is whether he meant bits (Kbps) or bytes (KBps) when he referred to kbps. I'll assume he meant bytes. Update: in email, SDB clarified he meant bits. That means you'd be able to pump 8x more calls in the same bandwidth I originally calculated.

His article reminded me about an old idea I had kicked around for some time, the thought that voice over IP (VoIP) could be mated to a limited cell antenna on the roof and provide the ability to route calls over your broadband connection. There are all sorts of dead zones for cellular coverage, from rural zones that might be able to get broadband but don't have a cell tower handy to the middle of the big city where building reflections create oddly shaped dead zones, there is plenty of opportunity for somebody who is willing to share some of his bandwidth to cell phone customer in need. The number of calls handled by each individual station would not be large. At 8KBps someone willing to share 256Kbps up/down (this would have to be symmetric) would be able to handle 4 simultaneous voice calls. Why would he share it? Two reasons come to mind, he gets some income from it so that his bandwidth expenses go down (perhaps he even makes a buck or two) and being in such a dead zone, he wants his own cell phone calls to go through easily (which is how I got the idea in the first place).

Obviously, you can't be stomping all over these calls with other traffic. You'd need to be on an IPv6 network and have Quality of Service (QoS) turned on so that these voice packets have priority and your Quake slugfest gets a bit of extra lag but that's not theoretically hard and IPv6 is now a sure thing now that the DoD has started the changeover process.

Living in a world where cell reception sites can overlap might prove more of a challenge and require some improvement to current standards so that automatic negotiation can occur with overlapping sites properly negotiating who handles which call but no doubt the cellular carriers will enforce fairness among their affiliate receivers.

The relevant side effect to SDB's post is that such a distributed receiver network would be more resilient. In case of emergency, people would likely be willing to switch more of their bandwidth over to phone conversations and the larger variety of receiver maintainers would ensure that some capacity would be preserved by techno-geeks with hefty backup generators even if the regular towers quickly drain their batteries.

I don't want to cause any confusion that such a system is practical today. IPv6 isn't supported by *any* mainstream ISP that I know of (in N. America at least) and the equipment to throw up such a affiliate tower in an apartment window or a house's rooftop is just not in existence as a do it yourself prebuilt kit. But a decade from now IPv6 will be a mainstream reality, bandwidth will likely be cheaper, and the need for a more robust cell phone network will still be with us.

Update: SDB's riposte was swift and characteristically pessimistic. One thing he didn't address fully was the time factor.

There are three basic ingredients in creating a cell receiver once you take away the problem of finding a location, the human time necessary to design the thing, the intelligence necessary to calculate and process the signals and the physical bits of plastic and metal needed to make the system. Is antenna design a science or an art? I suspect it's science, and very precise science at that. Given the proper software, could the process be automated and take out the human engineering that contributes to the cost? I suspect that if it's not true today, it'll become true during the decade timeframe I established as a minimum necessary for such an innovation to come to pass (remember, IPv6 and QoS needs to happen first). You'd just get shipped a box that you'd place where the antenna will go and turn it on. The box would gather the requisite data of where nearby cells are, etc. and make it ready for processing into a decent antenna design (remember, we don't have to be super-efficient, we're going for antenna volume).

The processing power at the central office needed to process a call is no different if the call is received through a big central tower or through an IPv6 Internet gateway so we can take that out of the equation. Processing power is supposed to follow Moore's law for about the next decade at least so by the time the Internet infrastructure can support this idea, we'll have doubled processing power at a given cost and electricity draw 6 times and be two thirds of the way on to our 7th doubling. That means that whatever the current silicon costs, it'll be somewhere between 1/64 and 1/128th as expensive as today's versions. How much this deflates the actual cost depends on how much of the current cost is electronics and how much is the antenna.

So we have an automated system for determining the best design for your new custom antenna, we have incredibly cheap and powerful chips to power the thing, but how does the actual antenna get made?

If people are currently custom hand tooling the antenna assembly it's likely that a decade from now, the same results could be printed out by the same program that calculates how the antenna should to look. It's not such a big stretch to see 3d printers available in either Home Depot (a do it yourself home products store), Kinkos (a printing and office service bureau), or both.

So, human input, processor power, and antenna fabrication all can have their costs drastically reduced over the next decade. I can certainly see a $200k receiver station being ground down to $1k-$3k in a decade if most of the cost is subject to Moore's law. In two decades it'll be in the hundreds and at a price point where the middle class can routinely afford to put it into place.

The right to be unhumane

The City Comforts Blog links back to my observations about conservative/libertarian urban planning. Unfortunately, there's a little gotcha in his response. He writes "I would be delighted to hear practical, conservative alternatives which do not deny the goal of humane urban form as a way of achieving it". The problem is that a pro-freedom solution to the problem will always deny an end result as the only possible solution to a problem. That's the very definition of freedom, the ability to choose to pick the optimal solution or not without outside forces coercing you. So a libertarian solution, even if it creates a great deal of net social benefit over all other solutions and greatly improves the present state of land use planning, is ultimately always going to be less efficient in creating new urbanist paradises then an imposed solution because a libertarian solution would always seek to minimize coercion and coercion, taken in isolation, is very efficient.