The cell edge is a convenient concept for simplifying the optimization and analysis of wireless networks, but it’s almost completely without real-world significance.
The rapid move toward OFDM-based 4G wireless network technologies – particularly LTE – has given rise to a lot of scholarly research aimed at devising useful optimization strategies. A frequent theme of these studies is development of rules-based processes for spectrum resource allocations. Unlike 3G networks based on CDMA, LTE networks require some sort of “frequency planning” to manage co-channel (or, more accurately, “co-subcarrier group”) interference.
Unfortunately, in their efforts to devise frequency allocation strategies for LTE, all too many researchers rely on the expedient of using a very simple model for the physical network. Articles in peer review literature often show networks as a field of circles representing propagation footprints of individual cells, with predictable and well-behaved zones of overlap. Of course, such representations bear little resemblance to real-world conditions, wherein RF propagation is neither predictable nor orderly. However, perhaps the most misleading simplification of network morphology is use of the concept usually referred to as the “cell edge.”
As commonly used for purposes of interference management or performance prediction, the “cell edge” refers to the region within the service area of a cell nearest to a boundary of service areas of neighboring cells. Mobile stations within this region are assumed to have relatively higher path loss for signals from their serving cells and relatively lower path loss – and thus potentially greater interference – from other nearby cells. In order to maintain data speed performance for “cell edge” mobiles, it is assumed that they require more interference protection, typically in the form of more conservative frequency reuse. Alternatively, it may be recognized and accepted that higher interference levels in cell edge regions will result in lower data speeds.
Unfortunately, while the cell edge is a handy concept for simplifying the optimization and analysis of wireless networks, it is, for two important reasons, almost completely without real-world significance. First, there is simple geometry: By any reasonable definition, the cell edge will often contain at least half of the mobiles operating in that cell. Consider, for example, a circle of some radius that represents the service area of a cell. (Yes, I understand that I am using the same simplification here that I denigrated a couple of paragraphs ago, but please bear with me.)
The edge of the circle represents some RF path loss from the base station located at its center. Now within that circle, let’s draw a concentric ring that represents a 6 dB smaller path loss from the center of the circle. That’s not all that much path loss margin to define the “cell edge,” but let’s go with it for now. Where would that ring lie? Well, assuming freespace attenuation (generally meaning that the mobile is in line-of-sight with the base station), the ring would have exactly half the radius of the entire circle. But the area of the portion of the circle outside the ring would be three times the area of the portion inside! Assuming geographic distribution of mobiles is homogeneous, that means that according to our model, a full 75 percent of mobiles are within the “cell edge.”
“But wait,” you say, “this model uses the unlikely assumption of free-space path loss. In a typical urban environment, attenuation is likely to be much steeper.” Very true. So let’s assume that, instead of attenuation varying as the square of distance, as is the case for free space, it varies as the fourth order of distance. In that case, the ring defining the “cell edge” would be at about 70 percent of the circle’s radius. The area of the “cell edge” would then still be a full half of the cell’s total footprint. Put another way, the geometry is very simple: There is a lot more area near the edge of a circle (or a cell’s service footprint, regardless of shape) than there is near its center.
The second factor that makes the “cell edge” concept pretty useless is that there is little if any correlation of path losses between serving and interfering cells. In other words, given the chaotic nature of RF propagation, particularly in urban settings, a mobile is just about as likely to be subjected to strong signals from an interfering cell when operating close to its serving base station as it is operating in the “cell edge.” Furthermore, the variability of that interference is enormous. Path loss between a mobile and an interfering base station can change substantially with movement of the mobile of only a few meters.
None of these urban propagation characteristics or laws of geometry are new, nor are they unique to LTE. Very early on in the evolution of analog cellular systems, network engineers discovered that frequency planning with rigid, rules-based frequency reuse patterns, to say nothing of modeling of networks as nice even hexagonal cells, did not work very well. So to me, at least it seems odd that, in developing supposedly sophisticated approaches to LTE spectrum resource optimization, today’s researchers often choose to start with obviously flawed geometric constructs, in particular, the “cell edge.”
In my opinion, a much better approach would be to assume that RF propagation in urban settings is both chaotic and unpredictable, so that interference cannot be managed on the basis of simple geometry. Instead what is needed is a system that allocates spectrum resources in real time based upon actual path loss conditions for mobiles operating in the network. The required information is there, mainly in the form of neighbor cell signal level measurements taken and reported by active mobiles. What is required is a system that can take that information and use it to manage subcarrier group assignments in both time and frequency domains so that each link operates with the lowest possible interference levels. Such a system will undoubtedly be enormously complex and expensive, but time and time again, network operators have found that, compared to meeting growing demand with more spectrum or more infrastructure, network optimization is almost always a better deal.
Drucker is president of Drucker Associates. He may be contacted at edrucker@drucker-associates.com.