Madden: Performance-driven or cost-driven? Now we have to decide.

Joe Madden

     Joe Madden

Consumer femtocells are cool. Active antenna systems are cool. HetNets are cool. But the mobile infrastructure market is not a consumer market, where "cool" devices go viral. In the mobile infrastructure game, it's important to remember that all of the technology we invent has one purpose: Get the bits from Point A to Point B, faster and cheaper than before.

Most people in the mobile industry look back at the major innovations in our market in terms of the technology advancement. We achieved a 10x improvement in spectral efficiency with 2G and a 5x improvement in 3G. In our industry, we tend to believe that we introduce new technology in order to satisfy end-user desire for fatter pipes. It's a performance-centered point of view.

Here's another way to look at it: Each generation of technology is, in fact, introduced as a cost-reduction measure. Increasing throughput is a way of spreading the network's cost over a higher number of users. This is the cost-centered point of view.

For the past 15 years, the performance point of view and the cost point of view have agreed remarkably well. We have more mobile data to process every year, and we have essentially redesigned the mobile network every five to 10 years to push more data through the pipe. Base station equipment and operating costs have come down but not very quickly. For many years, the dramatic cost reduction in our industry has come from pushing more data through a given bandwidth of spectrum. So the "green eyeshade guys" in Accounting get along with the "pocket protector gang" in Engineering.

Here's the problem: Now we are reaching practical physical limits on spectral efficiency. In the future, we can't depend on the same kinds of improvements for dramatic cost improvement. We can't cram more bits through the same spectrum, so instead we need to increase capacity (and reduce cost) through improved spatial efficiency. It's important to note that increasing single-user throughput may be different than increasing system capacity, and, in fact, we may need to distinguish between these dimensions to choose the right technologies for the future.

Looking forward, we have a lot of technology choices in the hopper. Instead of drooling over these sexy new technologies, we should evaluate each one from a cost/benefit point of view:

  Active Antenna Systems: Using active arrays can increase capacity in a macrocell configuration as much as 30-50 percent in a dense urban network. Unfortunately, the AAS products that we have seen so far also add at least 30-50 percent to the cost of the base station. This technology will be limited to specific sites where density is very high.

  Cloud RAN: Centralized baseband processing can save duplication of resources in the mobile network, if the fiber is available to route traffic with low latency. In the end, the operators with fiber will take this approach, and operators without fiber will use it in combination with small cells. The impact on cost: Don't expect a 10x improvement here. Cloud RAN will improve total cost more incrementally.


►    Carrier Aggregation can possibly lead to higher single-user data rates, and will give operators more flexibility in how they use spectrum. However, CA will not have much impact on data capacity or cost per Mbps, because separate transceivers are likely and system complexity increases with the same spectral efficiency. UE economy of scale will be a problem which could increase cost. Mobile Experts predicts some adoption of CA but not the widespread adoption assumed by the industry.

►    Coordinated MultiPoint (CoMP) will increase uplink throughput and capacity roughly 20-25 percent for an incremental increase in baseband processing horsepower. This one could be a winner in the age of social networking and heavy uplink data demand, but it needs fiber backhaul to make it work. Expect adoption in uplink-limited LTE networks with fiber.

►    Inter-Cell Interference Coordination (ICIC) will increase throughput and capacity roughly 20-40 percent depending on the network density and terrain, again for an incremental increase in baseband processing. Moore's Law will enable this kind of savings to take place in the next generation of chipsets. Again, expect adoption to be limited to dense networks with low-latency (fiber) backhaul.

  Femtocells: We've been predicting for many years that residential femtocells will be a flop. This prediction has come true, with our quarterly shipment survey tracking extremely weak numbers for 2012. The reason: Femtocells address coverage, not capacity...and the operators are no longer motivated by the performance available to individual users. Operators are concerned with overall network capacity nowadays, and any infrastructure that only helps occasionally for single users is considered a distraction.

  Small Cells: There's a wide range of outcomes for small cells depending on the existing spectrum, backhaul options, and network. The cost improvement from Small Cell deployment can be minimal if the available backhaul options are expensive. On the other hand, operators that have access to fiber inside buildings can achieve a factor of 10x cost improvement over new macro base stations. Operators that don't have access to cheap fiber will find that half of their Total Cost of Ownership will be attributed to backhaul initially, and many will wait until TCO comes down before rolling out large numbers of units. It's important to note that LTE small cells will not give users a 10x improvement in throughput. Instead, the data rate will be comparable to the LTE macro network, but capacity will increase and the cost of delivery will drop.

Overall, despite the rich mix of new technologies available, the clear winners will be the technologies that bring cost down. Our industry has, for many years, reduced cost through leaps in spectral efficiency. Now that we're reaching physical limits, we need to bring cost down in other ways. Over the next few years, watch as the sober businesspeople tell the engineers to fold up their centerfold photos of a few complex technologies and push for the ideas that drive cost down.

Joe Madden is principal analyst at Mobile Experts LLC. He is a Silicon Valley veteran, with 23 years in mobile communications, navigating through IPOs, LBOs, divestitures and M&A. He leads the analysts at Mobile Experts, focusing on nitty-gritty analysis of mobile communications and semiconductor markets. Madden graduated, cum laude, from UCLA in 1989.