Stardate 20030825.0200

(On Screen): TMLutas was inspired by my post earlier today about cell phones, and proposes what he thinks is an obvious solution to the problem of dead spots.

I'm afraid that he is completely deceived about where the difficulty lies. He seems to think that the bottleneck is communications between cells and the central switch, and his idea is to get volunteers with persistent net connections to put "a limited cell antenna on the roof and provide the ability to route calls over your broadband connection".

Getting a data link to the cell is trivial. That's not the problem. The problem is that his "limited cell antenna" has to be connected to a rather non-limited cell.

One of the reasons why CDMA took so long to appear is that it is extraordinarily sophisticated. We're not just talking about a simple FM transceiver here; even in the phone you have to implement things like rake receivers, and that's with the system designed to put as much of the work in the cell as possible so as to keep the phones cheap. The receive electronics in the cell is vastly more complicated than in the phone. It costs more per channel and the cell has a lot of channels.

Which, unfortunately, means that there's no such thing as a "cheap cell". Even a small one (e.g. one unidirectional sector) costs several couple hundred thousand dollars. Add more sectors and the price goes up, up, up.

And the antenna isn't just a whip mast; at those frequencies it has to be a tuned antenna with certain physical characteristics. A cell with a single sector still has three banks of antennas because there isn't any way to make an adequately efficient omnidirectional antenna at those frequencies. (When you're working with microwaves, everything changes.)

Some of the work isn't being done even in the cell. The cell sends its data to the central office, and the central switch does even more processing on it before it's possible to reproduce the call; it isn't rendered directly to voice at the cell. Cells do not stand alone; they are part of a distributed computing system, and the central switch does a lot of essential work which can't be dispense with (and can't be duplicated locally for any reasonable price).

Some of the error correction can't be done at the cell. That's because of a feature called "soft handoff" which comes into play when a phone is talking to two (or three or four) cells simultaneously. The spec permits up to six cells at once, but no phone in existence is capable of that. However, most of the IS-95 phones out there are capable of talking to three cells simultaneously. Some of the newer ones can handle four.

You as a user are not aware of when your phone goes into soft handoff; it's all handled automatically, and you don't get told. But it happens a lot; it's an essential part of the system. If you're using your phone while driving down the highway, you're probably in soft-handoff about a quarter of the time. CDMA pretty much can't work without it. And when a phone is in soft-handoff, no single cell actually is capable of reproducing the call. All of them send their data to the central switch, and it does the work. Each of the individual data streams has an unacceptably high error rate, but by processing all of them at one time, the data stream can still be retrieved. (The errors don't coincide; the central switch picks the best data from each of the cells.)

By the way, that also happens when you move from one sector of a cell to another sector in the same cell. Both data streams are sent to the central switch, and reassembly of the call data happens there. The equipment and software to do that isn't present at the cell, mostly to keep costs down.

I'm afraid his suggestion is totally impractical. The communication link between the cell and the central office isn't the hard part; the problem is that cell equipment in a CDMA system (or, in fact, in any cell system, though CDMA is much worse) is inherently expensive. There is no shortcut; there's no way to make a "small one" that's really cheap, unless you think that a hundred thousand dollars is cheap, which is about as low as I think you could get it by cutting down on redundant parts (and cutting down on the number of simultaneous channels it could carry).

And there's no way to make it operate independently; it has to be part of the system and it has to communicate with and cooperate with the central office. You can forget about it routing its voice directly onto the internet as a VOIP call. I just hit a few of the high points; there are dozens of reasons why the Lutas plan can't work.

I'm not sure that what he's describing could even be done for AMPS. In the case of CDMA the problem is that the electronics in the cell is inherently complicated and expensive, but in the case of AMPS the problem is frequency reuse. When they lay out AMPS systems, they have to divide their overall spectrum into sections, and assign sections of spectrum to cells in such a way that no cells which are near one another use the same sections. If they do, the people with the two phones on the same frequency will hear one another. (Or rather, each one will hear the other guys remote.) Even if you could create a mini-AMPS cell (and even that isn't all that easy, and it also has to cooperate with the central switch) you would have the problem of spectrum assignment. AMPS is so gawdawfully inefficie