The majority of today’s cellular networks are now on their third generation (3G). Based on Universal Mobile Telecommunications System (UMTS) or Code Division Multiple Access (CDMA) technologies, they support data rates of a few megabits per second (Mbps) under low-mobility conditions. During the last few years, these networks have been overwhelmed as the number of cell phones has dramatically increased. Meanwhile, most cellular service subscribers now also use the mobile Internet and, over the past three years, many new broadband mobile devices have been commercialized including smart phones, tablets and laptops with 3G capabilities. As a result, the amount of wireless data that cellular networks must support is exploding.
This avalanche of mobile data makes it increasingly difficult for operators to ensure sufficient network performance. Yet the only way they can remain competitive and continue offering unlimited data plans is to reduce the price/performance ratio for mobile data. In response, the industry has introduced the LTE standard as the first step toward a new 4G cellular network, which has recently entered the deployment phase. The latest releases of the LTE specifications also include the concept of the Heterogeneous Network (HetNet), which will rely heavily on System-on-Chip (SoC) integration to handle the computational complexity of many different types of small cell Evolved NodeB (eNodeB) LTE base stations.
In particular, a HetNet requires a special variation of application-specific SoCs that enable more robust, multi-level parallelism while delivering all the required eNodeB Layer 1-3 protocol stack software and managing its execution across multiple programmable processing cores.
Standardized by the 3G Partnership Project (3GPP), the first, LTE Release 8 specification was frozen in December 2008 before being published in March 2009. Release 8 offers downlink data rates of up to 150 Mbps -- a very promising solution for operators seeking an efficient way to evolve their network.
Because LTE technology is based on an evolution of the existing 3G network, operators will be able to deploy is more quickly than alternative wireless standards such as WiMAX. The first LTE deployment was executed during 2010 by TeliaSonera in three Norwegian cities and 25 Swedish cities. Since then, many other commercial deployments have been launched and are now underway around the world.
LTE introduces two major innovations beyond previous 3G techniques. First, in order to reduce network latencies, a flat, all-IP core network architecture is proposed. This new core network, called evolved packet core (EPC), introduces new interfaces that directly connect enhanced base stations (eNodeBs, or eNBs) together, and removes 3G’s hierarchical approaches.
The second major enhancement proposed in LTE is a new radio access network called the evolved universal terrestrial radio access network (E-UTRAN). E-UTRAN is based on the orthogonal frequency division multiple access (OFDMA) technology for the downlink and single-carrier frequency division multiple access (SC-FDMA) technology for the uplink. LTE supports bandwidth of up to 20MHz making it a suitable standard for high data rates. An advantage of OFDMA and SC-FDMA is the orthogonality of neighboring subcarriers, which avoids interference between adjacent narrowband frequencies. SC-FDMA was chosen in the uplink (unlike WiMAX, for instance) because its transmit signal has a lower peak-to-average ratio.
A new frame structure is also proposed in LTE, in which resource elements called physical resource blocks (PRBs) are allocated in both the time and frequency domains, thus allowing the network to serve more users at higher data rates. Like all other new wireless technologies, LTE also generalizes the use of multiple antennas at both receiver and emitter side. Such systems, called multiple input multiple output (MIMO), also greatly increase performance, while making it possible for eNodeBs to perform beamforming and focus their signal in the optimal direction required by users, thus reducing interference between neighboring cells and users.
Release 9 of the LTE specification proposed several additional enhancements including architectural features for home eNBs (HeNBs). HeNBs are a new kind of small base station, also called femtocell, that are aimed at increasing radio coverage indoors.
In 3GPP LTE Release 10, it has been proposed that all techniques from previous releases be enhanced and extended. There are three major enhancements, including advanced antenna techniques, carrier aggregation, relay mode, enhanced self-organization network (SON) techniques. SON is a key tool for deploying small cells, as part of the HetNet, which is made up of a variety of basestations from femtocells to macrocells. The HetNet concept is particularly important because it reduces the distance between users and their serving cells, which correspondingly increases coverage and, therefore, capacity.
The concept behind the HetNet is to deploy multiple small cells in specific locations where high capacity is needed (see Fig. 1). In this way, rather than relying exclusively on large macro base stations throughout the desired coverage area, mobile operators will be able to deploy them primarily in rural and other locations where capacity is not an issue. Elsewhere, smaller cells can be either by the mobile operators or end customers to deliver additional capacity in those locations where it is needed.
During the last several years, a number of HetNet small cell devices have emerged to join earlier macrocells, including relay nodes, picocells and femtocells (which can be further sub-divided into metrocell, enterprise and home platforms. Table 1 summarizes the main characteristics of the proposed HetNet small cell devices.
Compared to traditional homogeneous networks made from a single layer of macrocells, HetNet small cell devices will be deployed in high numbers and in an uncoordinated way. This presents new challenges including interference management, mobility management, security, backhaul, and synchronization.
Interference is a particularly difficult challenge that must be solved. This includes interference between both the macrocell layer and the small cell layer, as well as interference between the small cells, themselves. As a result, 3GPP is exploring new techniques for incorporation into HetNet standards. These techniques include:
* Range Expansion: This applied to small cells deployed by the operator, since it requires accurate inter-cell synchronization. In order to reduce the load on over-burdened macrocells, it has been proposed that mobile users be authorized to connect to a neighboring picocell even if its radiated power is lower than the macrocell (which usually is not possible). This technique, as illustrated in Fig. 2, is similar to a virtual extension of the picocell coverage area, and can be executed using advanced receivers that work well in a low SNR environment. Picocell range expansion is a good technique for reducing interference at the cell edges.
* Enhanced Inter Cell Interference Coordination (eICIC): this is the advanced version of ICIC proposed in earlier 3GPP releases. The eICIC technique uses almost blank subframes (ABSFs) which mean that portions of the LTE frame are left empty for one of the layers. This is done in a coordinated way so that there is no overlap between the data channels of both layers, and can be performed either in the time or frequency domain.
* Advanced receivers: These receivers operate at lower SNR and cancel the signals from neighboring cells. They also help to avoid overlap between cells and improve performance as compared to previous techniques.
The Solution: The SON
Solving the aforementioned challenges is not an easy task, especially when the main goal for operators is to reduce CAPEX and OPEX. Cutting these expenses requires a reduction in human involvement, which can be accomplished by including SON capabilities in small cell equipment. SON technology is the key solution to the challenges described earlier, and to the success of small cells.
Indeed, small cells must have plug-and-play simplicity, and they must be capable of automatically adapting their parameters (i.e., radio, mobility, security, etc.) depending on conditions. Therefore, in LTE Release 10 and beyond, SON will continue to be a very important technique.
SON works as follows, as described in the 3GPP specification:
1: Gathering data from neighbors: Each small cell gathers data from neighboring cells. This can be done through a combination of:
- Measurement reports generated by the mobile users, which include a list of their serving cell’s neighboring cells;
- Network Monitoring Mode detection of downlink signals from neighboring cells, which is performed by small cells that behave as simple user equipment (UE) devices.
- Exchange of information directly between cells, which is difficult in the case of femtocells because the LTE X2 interface is not yet defined. This could be done either via copper links or, preferably, wirelessly in order to reduce costs.
2: Optimizing: Based on the collected data, parameter configuration is then performed. This can be either done in a centralized way, whereby a SON server optimizes a group of small cells, or it can be done in a distributed way, whereby each small cell optimizes its own parameters, as illustrated in Figure 3. In practice, a combination of both is generally used, depending on the parameters, and is referred to as a hybrid cell. However, fully distributed SON will be the preferred approach in the future because it will reduce complexity and cost.
System-on-Chip (SoC) Advantages for Small Cell Deployment
Small cells require a multi-core, SoC approach in order to improve processing performance without power and cost increases, and to deliver and manage all of the production-grade, carrier-class eNodeB software -- including the L1-L3 protocol stack -- that is needed for a complete, scalable dual-mode solution across a variety of device platforms.
The optimal SoC configuration for small cell development is an application-specific solution that provides the re-programmability of a general-purpose processor without the cumbersome software development/updating, and the performance of a fixed-function processor without its inflexibility. After all, small cell SoCs don’t have the rapid obsolescence model of handsets, and must have relatively long lifecycles in the typical operator infrastructure. They must be easy to update so they can meet new feature requirements and as LTE specifications continue to evolve after they are deployed. Additionally, small cell SoCs need to be optimized for parallelism not only between cores, but also between clusters of cores and even clusters of network devices, in order to handle the large and complex eNodeB L1-3 protocol software stack. Finally, small cell SoCs also must be capable of dividing up all of that protocol processing across multiple cores.
Application-specific SoCs cut their teeth in the wireline infrastructure, and have steadily grown in processing performance and capabilities over the past decade. As can be seen in Figure 5, the number of transistors in the typical communications SoC has grown from under 3 million for dial-up modems in the mid-1990s through approximately 300 million for VoIP equipment in the mid-200s, and now approaches 1 billion for eNodeB devices. Similarly, what once required less than 10,000 lines of code for dial-up modems progressed to 2 million lines for wireline VoIP infrastructure equipment in the late 1990s and early 2000s, and has now grown to more than 20 million lines for executing the LTE baseband processing tasks required for 4G networks (see Fig. 4). This growth in software complexity reflects the parallel growth in computational loads of for such tasks as fast Fourier transform (FFT)/inverse FFT (iFFT). Even the transition from 3G to 4G has resulted in an order of magnitude higher baseband processing complexity.
Fig 4: Evolution of System on Chip solutions
Mindspeed first began working with high-performance communications SoCs in their first incarnation for carrier-class wireline VoIP, data and mixed-media processing. More recently, this combination of multiple, task-oriented processing engines has been adapted for triple-play customer premise equipment (CPE) gateways. These devices perform the entire processing function from time division multiplexing (TDM) samples to IP packet processing, which enables the same device to support applications ranging from single-chip access gateways (as a Class V switch replacement for PSTN) to very-high-capacity media gateways (as a Class IV switch replacement).
The application-specific SoC approach maps extremely well to solving LTE eNodeB baseband processing needs. Using the same multi-core approach in its Transcede LTE baseband processing family, Mindspeed has partitioned complex LTE tasks in such a way that each core performs the jobs it does best. General-purpose DSP cores are used to implement complex algorithms such as speech compression, channel estimation, etc. Reduced Instruction Set Computer (RISC) processors are used to implement control or protocol layer processing. And finally, ASICs or co-processors are used to perform computationally intensive yet algorithmically simple or fixed applications. This architecture enables designers to accommodate 4G’s increased computation complexity along with all of the other 3G/4G and legacy system needs, while also simplifying the task of incorporating compact and advanced radio interfaces, fulfilling all the SON requirements.
Each of these parallel processing elements is managed using a single-threaded, simplified programming model. The use of a hardware abstraction mechanism eliminates the need for partitioning the hardware resources in a traditional static manner, resulting in ease of design and faster time to maturity for wireless baseband products, as well as future software maintenance. This approach also facilitates optimal concurrent access and backhaul functionality at full throughput. Finally, it provides an ideal foundation for future converged fixed/mobile solutions as the industry continues to demand more and more functionality in a single UE box.
HetNets is a key technology for enabling LTE-Advanced and future wireless networks to reach the higher capacities that are required to serve a growing subscriber base consuming a rapidly expanding volume of mobile data. It is expected that HetNet small cell devices will proliferate extremely quickly. Implementing SON capabilities using highly integrated, application-specific SoCs provides an optimal solution for increasing performance at the lowest possible cost.