Most people now agree that the next wave of performance gains for mobile networks will have to come from ‘small cells’. That is, instead of using the large ‘macro’ base stations that already connect mobile devices to the network, carriers are going to need to install a large number of much smaller stations in already covered territory. But beyond that basic agreement, little else is settled.

The term small cells itself is already fraught, meaning different things to different people. But the notion is simple.  There is no more space to put up big base stations. By loading already covered areas with more cells, or ‘densifying’ them,  carriers can increase overall capacity, which will in turn yield greater bandwidth.

So much for the ideal. Most people we spoke with at the show agree that we are still several years away from deploying small cells in large quantities. Several carriers are engaged in trials now. A few, notably Sprint, have deployed femto-cells to plug holes in their coverage network, but this is a small subset of the problem. Most carriers are still in wait-and-see mode. At root, we think the industry still needs to address three major problems before we see large-scale deployments:

1 – How will small cells interact with each other and the wider network?

2 – How will we physically connect all these cells to the core network?

3 – How will carriers get approval to add all these sites?

The first two are technical problems, and we will share what insights we picked up from the show about them. The third one is the thorniest, but since it is a political question we are going to totally dodge it here. Suffice it to say, there is a lot of lobbying and political process that still needs to be worked out. Also, a lot of negotiation with city councils and property owners. This should be a big opportunity for the tower vendors like Crown Castle, but our basic checks indicate that they are yet to fully come to terms with it. Another note for a future date.

Going back to the technical.  The first question is probably the most complex. And this is where we start to run into trouble with the term ‘small cells’. Every carrier tends to have their own, unique set of circumstances when it comes to densification. Some carriers lack spectrum, some carriers lack locations, some carriers have complex regulatory regimes. The list goes on.

We think some definitions may help. We would categorize small cells into three categories that hinge as much on business terms as on the technology under the hood. Small cells can be defined based on where they are used (indoor versus outdoor), who pays to connect them to the core network (the carrier or the user), and what frequency they utilize (licensed versus unlicensed).  So the first box is the ‘pico cell’ which we will define here as a low-powered base station in which the carrier pays for the connection to the core network. These are typically used outdoors. At this stage, every major equipment vendor offers some variant of the product. We sometimes see these used indoors at convention centers or other large arenas. Most importantly, they use licensed spectrum, already owned by the carrier. It is also important to note that these are much more complicated than the ‘repeaters’ already in widespread usage. Repeaters can extend the signal of an existing base station (e.g. into subway stations), but do not add capacity in their own right.

The next box is the femto cell.  For this note, we are referring to low-powered base stations where the end-user pays for the backhaul. We see these in some homes today, notably the “Microcell” product that AT&T marketed for a few years. Often, these are used indoors, but importantly they use carrier-licensed spectrum.

There is a growing category of enterprise femtos, being championed by privately-held Spider Cloud. These use licensed spectrum, and backhaul from either the carrier or the enterprise installing it. Notably, Spider Cloud’s femto-cells do some very clever work to switch traffic and manage access points in traditionally hard to reach indoor locations.

The third category is Wi-Fi. This can be connected to the core network by anyone, but uses unlicensed spectrum. The trick with using Wi-Fi is determining who gets to access a network, and importantly how carriers will bill for it.

These definitions are only going to get more complicated in the future as equipment vendors roll out different sized boxes and merge different radios into one box. Nonetheless, the key axis of comparison remains clear – spectrum, backhaul and location.

The problem that all of these boxes share is how to coordinate them with the core network. Existing cellular infrastructure is built to control a few hundred base stations at most in a single city, but as small cells proliferate we are going to start seeing thousands, then probably tens of thousands deployed. This will swamp existing control mechanisms, notably the carrier switches (MSC) and aggregation points (RNC, BTSC). The industry is currently debating exactly how to re-design the network to handle this. Solutions vary widely. Some propose off-loading traffic to the public Internet immediately at the base station. Others would divide traffic at some point between the carrier network and the Internet. Others want to see the ability for small cells to switch traffic locally. While still an other approach is to massively upgrade all those existing core nodes. We imagine all of these and more will be deployed over time.

However, the real problem here is how to coordinate all these small cells. A big role of cellular networks today is determining which ‘cell’ any given phone should connect to.  This actually consumes a large portion of the base station infrastructure’s processing and switching functions. The coordination problem gets bigger the more cells there are, and these decisions have a huge impact on bandwidth, capacity and battery life. Currently, the hot topic in the space is “Self-organized networks” (SON). The idea here is that the small cells make decisions at a local level, either at the cell itself or one element above them. The space is hot now because of Cisco’s recent purchase of Intuacell. Cisco probably made the acquisition at the behest of a major US carrier who likes Intuacell’s approach.

Our checks indicate that there is still a lot of development needed for SONs to reach their full potential. We are not disparaging Intuacell’s work, but only want to point out that we are still in early innings.

For us, the interesting feature of SONs is the key importance of software, especially the software that will be used to determine the ‘best fit’ for coordinating all these cells. Eventually this may get burned into silicon, but the interesting thing about SONs will not be the performance of its radios, it will be the algorithms behind it.

Before we stumble into the rabbit’s hole of Game Theory, Nash Equilibrium’s and Russell Crowe’s future Oscar prospects, we should move on. SONs are just getting started.

This brings us to backhaul. For reasons not entirely clear even to ourselves, we spent much of our time at Mobile World Congress meeting with small cell backhaul vendors. If you tell this to the average Silicon Valley denizen or even the average MWC attendee, you will be met with eyes glazing over. Many people asked what evil deed we had committed to be subject to such a fate. Backhaul is not sexy. It is the mechanism by which a base station connects to the core network. This is the point where wireless starts to look a lot more wired. Put simply there are three ways to supply backhaul: wireless (usually microwave); fiber optics; older wired technologies.

The older wired technologies – notably T1s, DSL and cable, are the most commonly used today, but are not-so-gradually reaching their maximum capacities. So carriers are increasingly looking to fiber connections. However, in many places neither solution works (think old buildings) so carriers use microwaves – high capacity, short range radios to link up.

The prevailing view among most carriers is that small cells will need a lot of wireless connections to work. If the goal is to put a small cell on top of every municipal lightpost there will often be no practical way to link them all with fiber.

Which leads us to wireless solutions. Cellular carriers already use a lot of wireless technologies to backhaul existing base stations. Typically these are microwaves (i.e. high frequencies than used in most cellular radios). And traditional microwave vendors are starting to roll out their ‘small cell’ solution.

We had the chance to sit down with Ceragon at MWC, and they have recently introduced a high capacity system that could work well here.

The key point to keep in mind is that most microwave systems are ‘point-to-point’, connecting a single base station to the core. Small cell systems will have to connect point to multi-point, that is multiple small cells back to the core. In addition to the existing vendors, we also spoke to another dozen of such systems including Exalt, Ubiqam, Siklu, Taqua and Tarana, among many others. We can provide more detailed descriptions of these and others if you are interested.

So far, we have discovered no magic. All of these companies have interesting offerings, but none can defy the laws of physics. We did identify the key metrics of comparison: bandwidth, spectral efficiency, compression schemes and number of nodes each can support. Again, one or two companies stood out on some of these metrics, but there will probably not be a silver bullet to solve everyone’s problem. At least not any time soon.