In communications class in the 90s, I was taught this standard hexagonal model for cellular telephone systems. 
For those not familiar with it, it’s a way to model frequency reuse. No one imagines that radio signals from a given base station drop abruptly at the cell boundaries. Certainly no one imagines the coverage area for a base station is a hexagon. This is just a model for reusing frequencies.
As the number of cellular users increased, the size of the cells has decreased. It was increasingly difficult to place the base stations near the center of little hexagons. They are placed wherever there is room near places with a high density of people. The recent increase in demand for mobile data has hastened this process to the point where the hexagonal cell model no longer makes sense. A better model for their location is a random distribution. The diagram below from Dr. Andrew's article shows nested randomply placed cells, with macro bases in red, pico bases in green, and femto bases in black.
l
This failure of the traditional cell model raises the question of if congestion on the mobile cellular network could be drastically reduced by some new model of frequency allocation or coordination among base stations.
An article in the latest issue of IEEE Communications by Dr. Jeffrey G. Andrews highlights seven ways in which a modern HetNet is different from the traditional cellular network:
- Performance Metrics - In a densely populated area you may easily be able to get a strong signal but not be able to get much data through. You cannot judge network “coverage” by the signal strength of the base stations. Also, spectral efficiency of a transmission method in b/s/Hz is no longer a good figure of merit for spectral efficiency. A better measure takes into account the area covered by the signal. Spectral inefficiency is not important the signal takes up a very small area, e.g. a few rooms of a building.
- Topology - Pico and femto cells are now placed almost randomly, not according to a hexagonal cellular plan.
- Cell Association - It may be better to connect to a weaker cell that has more available data throughput than a congested one with a stronger signal.
- Downlink/Uplink - The pico and femto base stations have less power than macro stations, but the mobile phone transmits with the same signal strength to all of them. A user may be on the edge of a nearby low-power femtocell, too far to receive a good signal from it. It may make sense for the mobile to upload data to the femto cell but use a macro cell for download. There is no need to connect to the same base for upload and download.
- Mobility - It takes some time to handoff to a new base station. A mobile connected to a large congested cell may not benefit from briefly handing off to a picocell if it is only within range of it for a few seconds. This is a problem for vehicles passing through many small networks.
- Backhaul - Just like Wi-Fi Internet connections in small coffee shops, small BSs do not necessarily have a high-speed connection. Any protocol needs to take into account the amount of throughput available on the backhaul, not just the wireless side.
- Interference Management - Separating reused frequencies geographically, as with traditional cellular, is becoming impractical. One approach is to put small cells in different bands (or different ODFM sub-channels) according to their size. This is not optimal because it wastes spectrum not covered by certain sizes of cell. More complicated approaches require all base stations to coordinate and share the spectrum in their area. Dr. Andrews believes increasing the sheer number of cells is more important than a good algorithm to avoid collisions.
It seems counter-intuitive to me that just adding more pico and femto cells without coordination is an efficient use of bandwidth. I have never studied the problem, so I accept Dr. Andrews’ expert opinion. I expect to see some other service like Wi-Fi or perhaps a service allocated in the 60GHz range (perhaps intended as “cable replacement”) to replace the very low-area femto cells.