More metro Wi-Fi projects are in planning and trial phases than deployed in the United States. But those actually in production, such as Chaska, Minn., St. Cloud, Fla. and Mountain View, Calif., were installed with approximately 16, 20 and 33 Tropos nodes per square mile, respectively. It's important to note that this list is ordered from the oldest to the most recent deployment. Although it's perhaps risky to tease out a pattern from such a short list as this, apparently cities are installing more nodes per square mile than before.
A few months ago, the Chicago Tribune wrote a story on the Chaska deployment. According to a former member of that project, performance and coverage issues plagued the system until more nodes were brought in to boost link stability and fill the gaps. The node count was bumped up from 16 to almost 23 mesh nodes per square mile. In St. Cloud, the news was the lack of good indoor coverage. Mountain View may have unique terrain that requires an even higher density, but its high watermark count of 33 nodes per square mile may be indicative that cities have learned from past coverage and service problems and are now designing denser deployments.
Beyond these specific examples, the generally accepted number is roughly 25 nodes per square mile to provide the necessary 95-percent outdoor 802.11 coverage and 85 percent of indoor rooms that have an outside-facing wall. With nodes at roughly $3,000 a pop, $75,000 per square mile adds up quickly. Even if you apply volume discounts to the mesh hardware, there's still wireless backhaul equipment, pole rental and installation. Currently, with low penetration and standard access rates that border sub-DSL speeds, metro Wi-Fi deployments will be able to focus on minimum coverage requirements for quite some time. In fact, if the link budget (that is, the difference in signal strength between what is transmitted and what is received) can be increased, then lower densities will likely be architected to drive down deployment costs.
Link budgets are not automatically symmetrical in nature. FCC regulations state that access points or nodes in certain configurations are allowed to transmit at significantly higher levels than the clients (downstream). This means that even if the clients can interpret the transmitted signal, the nodes--even with a well-developed receiver and good sensitivity--may not be able to demodulate the response (upstream). Because the majority of consumer traffic is downstream, optimizing that aspect of the link is valuable; higher link connection rates result in a better subscriber experience and a more efficient use of the air time, which in turn increases overall capacity. Of course, if the access point or node can't hear the return signal, it's all for naught.