Friday, August 06, 2004

But it also reminds me (and hopefully in a tangible way reminds you) of several quite important things about radio receiver evolution:

• Receivers evolve to fit markets (even hackers are a market);
• "Shannon's limit"* at a point in space is obviously not independent of receiver/antenna design;
• useful systems designs are not closed or isolated - new applications get invented, and sometimes are bad, but sometimes good ('toothing as a social phenomenon for meeting desirable partners has grown from a security bug in phones to a feature for users - it's not too hard to find people who prefer Nokia phones to Motorola phones, because the latter are so secure they don't allow 'toothing).

Don't get me wrong - I don't think that yard-long mobile microwave yagis are likely to be a common direction of evolution, at least not in my probable future prediction. But I wouldn't bet a lot.

* [why the "quotes"?] Note that Shannon himself never claimed he was describing a physical limit. He was describing a limit to the amount of information capacity that could be realized via an abstract channel, not a physical channel, given a prior probability distribution on noise. I use the term in quotes above in the sense that engineers commonly misuse it today, not in the sense of Shannon's actual theory - hence the quotes, since I try to use the term correctly at all times. The yagi extracts a huge amount of information from temporal/spatial correlations within the electromagnetic field, while decorrelating randomly generated energy in the field. Thus the yagi is an essential part of the system that must be considered when measuring "noise" not a separable component. The noise energy and entropy are still there (the temptation to think that a bigger antenna works because it collects more signal energy is not at all correct - it also collects more noise energy, so the ratio remains constant...). The yagi doesn't make the noise go away.

Spatial correlation of electromagnetic noise is one of the reasons that it's foolish to speak of "noise temperature at a point in space" as a limit on capacity. If there is even such a real thing as noise temperature, it is too abstract to be useful as a limit on systems capacity. Noise at a point is not independent of noise nearby - and correlated noise can be removed (just as the yagi does in its simple way, and the Slepian-Wolf theorem does in a more subtle way). Capacity has everything to do with the transport system's architecture, starting with the antennas.