Mobile infrastructure continues to grow, but at a slower rate than in recent years, paralleling the slower sales of tablets and smartphones. Even so, network operators are struggling to keep up with bandwidth demand. Worse, from a business standpoint, operators are still having trouble increasing their average revenue per user. To add capacity, they must use the available radio-frequency spectrum more efficiently or find new RF spectrum.
Small-cell base stations are one solution. The smallest examples are residential base stations, sometimes deployed by operators to address customer complaints about poor cellular coverage. Small cells also include enterprise base stations that cover a workplace, outdoor cells that cover a campus, and cells that cover a large venue (such as a sports stadium). These base stations cost much less to build and deploy than the large macrocells that form the cellular network’s backbone. A cellular network that combines macrocells with small cells is called a heterogeneous network (hetnet). Although hetnets provide better coverage, cost-conscious operators are increasingly reluctant to deploy small cells in the vast numbers once predicted.
To make small base stations easier and less expensive to build, chip vendors have introduced integrated base-station processors that consolidate most of the system’s vital functions in a single chip. These functions include radio-baseband processing, which is performed by digital signal processors (DSPs), and user scheduling, which is performed by general-purpose CPU cores. By adding various accelerators, I/O interfaces, and memory controllers, such an integrated processor is almost a base station on a chip. A few of these products are powerful enough for even macro-cells, especially in multiprocessor configurations. The newest chips implement LTE and LTE-Advanced, capitalizing on the global deployments of 4G networks.
Several vendors have introduced these processors in recent years, but the number of vendors is now dwindling. Lower-than-expected demand and corporate acquisitions are taking their toll. In early 2016, Avago acquired Broadcom (while keeping the Broadcom name) and immediately began major cutbacks. Among the casualties were Broadcom’s integrated base-station processors, which are still available to existing customers but are no longer marketed for new designs. After acquiring Mindspeed’s wireless business, Intel eventually terminated its development projects.
Most other base-station processors have killed 4G product development, although a few new 4G variants have shipped in the past year. These vendors are focusing their development efforts on the upcoming 5G transition, which will debut new technologies. The 5G standard is still in flux, however, and commercial base-station deployments aren’t expected to begin until 2019 or later, although a few operators plan to deploy pre-standard devices sooner.
One new technology stirring controversy involves offloading some cellular traffic onto unlicensed RF spectrum that is currently used for Wi-Fi networking, Bluetooth links, and numerous other consumer applications ranging from garage-door openers to baby monitors. Adding cellular traffic could interfere with those signals. Cellular operators and chip vendors are promising peaceful coexistence, but so far, none have deployed such a service commercially. Any such deployments would boost sales of new base stations that support this technology.
The most powerful integrated base-station processor is Cavium’s Octeon Fusion-M CNF7500, which can handle up to 3,600 LTE users in a 12-sector macrocell. NXP’s QorIQ Qonverge B4860 can handle up to 1,200 LTE users, enough for a smaller macrocell (or a larger cell when operating in a multiprocessor configuration). All other integrated base-station processors are designed for small cells that serve from about a dozen to a few hundred users. Almost all 3G-only processors have left the market, given operators’ emphasis on 4G deployments. Most existing processors are dual-mode 3G/4G products.
Chip architectures vary widely, so base-station software isn’t portable from one chip vendor to another. Each company uses a different combination of CPU and DSP cores. This incompatibility discourages customers from switching vendors. Over time, we expect more vendors to use ARM CPUs, but the industry is unlikely to standardize on a DSP architecture, so vital baseband software will remain incompatible.
Intel is more interested in selling general-purpose Xeon x86 processors for cloud radio access networks (C-RANs) and other new elements of the wireless infrastructure. Qualcomm offers the industry’s most complete hardware/software solution, especially since Broadcom’s withdrawal. Its pending acquisition of NXP will widen its product line and create an opportunity to establish leadership in this market. We expect TI’s product line to remain relatively static; unrelated analog products are becoming its main focus. Cavium has strong base-station processors and an opportunity to exploit niches left vacant by other vendors, but it’s a relatively small company with several product families to support. If industry consolidation continues, it’s one of the few remaining takeover targets.
In summary, integrated base-station processors, small cells, hetnets, 4G networks, and future 5G technologies are rapidly changing wireless networks and are realigning the market players. We expect to see few new 4G products as the surviving vendors focus on emerging 5G technology for their next-generation products.