Editor's note: To view a PDF version of this article, click here.
For roughly 70 percent of the global market, the cornerstone of ubiquitous broadband wireless service is the newly standardized Universal Mobile Telecommunications System (UMTS) extension known as High Speed Downlink Packet Access (HSDPA). The issues surrounding the introduction of HSDPA, which offers peak data rates of up to 7 Mbits/ second, are representative of the major hurdles that carriers and OEMs face as cellular networks evolve from the voice-centric 2G to data-centric 3G. In particular, carriers find themselves faced with the problem of creating market demand for new data-oriented services while ensuring that they can adequately support those services with minimal capital expenditures. While HSDPA would seem to supply the answer, it involves some front-end basestation modifications, the details of which must be well-understood by the network designer.
UMTS fundamentals
The UMTS network architecture consists of several components: the Node-B, or basestation; the radio network controller (RNC); the serving GPRS serving node and gateway GPRS serving node; and the mobile switching center (Figure 1). For the purposes of this article, it suffices to view a Node-B as handling the physical-layer air-interface functions, the RNC as managing radio resources and most mobile-call functions, and the serving and gateway nodes as handling and routing data traffic. The mobile switching center, of course, provides the interface to the wireline network.
The primary objective behind HSDPA was the establishment of a cost-effective high-bandwidth, low-delay packet-oriented service within UMTS. Backward compatibility was critical, so the HSDPA architects adhered to an evolutionary philosophy. From an architecture perspective, HSDPA is a straightforward enhancement of the UMTS Release '99 (R99) architecture, with the addition of a repetition/scheduling entity within the Node-B that resides below the R99 media-access control (MAC) layer. From a cellular-network perspective, all R99 techniques can be supported in a network supporting HSDPA, since HSDPA mobile terminals (called UEs) are designed to coexist with R99 UEs.
Technically, the basic operational principles behind HSDPA are relatively easy to understand. The RNC routes data packets destined for a particular UE to the appropriate Node-B. The Node-B takes the data packets and schedules their transmission to the mobile terminal over the air interface by matching the user's priority and estimated channel operating environment with an appropriately chosen coding and modulation scheme (that is, 16QAM vs. QPSK).
The UE is responsible for acknowledging receipt of the data packet and providing the Node-B with information regarding channel condition, power control and so on. Once it sends the data packet to the UE, the Node-B waits for an acknowledgement. If it does not receive one within a prescribed time, it assumes that the data packet was lost and retransmits it.
In short, HSDPA continuously strives, with some modest constraints, to give the maximal bandwidth to the user with the best channel conditions. The data rates achievable with HSDPA (Figure 2) are more than adequate for supporting multimedia-streaming services.
Although conceptually simple, HS-DPA's implementation within the context of a Node-B does raise some architectural issues for the designer.
In a typical network deployment, the Node-B radio cabinet sits in proximity to the radio tower and the power cabinet. For indoor deployments the radio cabinet may be a simple rack, while in outdoor deployments it may be an environmental-control unit. The guts of the radio cabinet are an antenna interface section (filters, power amplifiers and the like), core processing chassis (RF transceivers, combiner, high-performance channel cards, network interface and system controller card, timing card, backplane and so on), plus mechatronics (power supply, fans, cables, etc.) and other, miscellaneous elements.
The core processing chassis (CPC) is the cornerstone of the Node-B and bears most of the cost. It contains the RF transceiver, combiner, network interface and system controller, timing card, channel card and backplane. Of these CPC elements, only the channel card needs to be modified to support HSDPA.
The typical UMTS channel card comprises a general-purpose processor that handles the miscellaneous control tasks, a pool of DSP resources to handle symbol-rate processing and chip-rate assist functions, and a pool of specialized ASIC or ASSP devices to handle intensive chip-rate operations such as spreading, scrambling, modulation, rake receiving and preamble detection.
To support HSDPA, two changes must be made to the channel card. First, the downlink chip-rate ASIC (or ASSP) must be modified to support the new 16QAM modulation schemes and new downlink slot formats associated with HSDPA. Some vendors anticipated HSDPA in their downlink chip-rate ASIC designs and thus are in good shape. But others designed their ASICs well before the emergence of HSDPA and are now exploring respins or are looking to integrate newly introduced ASSPs that support HSDPA. In addition, the downlink symbol-rate processing section must be modified to support HSDPA extensions.
The next change requires a new processing section, called the MAC-hs, which must be added to the channel card to support the scheduling, buffering, transmission and retransmission of data blocks that are received from the RNC. This is the most intrusive augmentation to the channel card because it requires the introduction of a programmable processing entity together with a retransmission buffer.
Since the channel card already contains both a general-purpose processor and DSP, one can make convincing arguments that the MAC-hs could be effectively realized using either of the two types of devices. Nonetheless, many designers are finding that, because of the close ties between the MAC-hs function and the lower-layer symbol and chip-rate functions, the DSP is the more practical choice. Simulations have shown that a retransmission buffer of approximately 2.5 Mbits in size is adequate to handle the buffering requirement of a standard cell with 75 or so users.
Future enhancements
Despite the importance of voice for revenue, carriers realize that their networks must be capable of handling accelerated demand for more sophisticated data services as time goes on. To this end, carriers are examining technologies such as remote radio units, widespread fiber-based transport, processing server farms, tower-mounted power amplifiers and software-defined radio. Also under investigation are efficiency-enhancing algorithms such as power amplifier linearization, adaptive antenna arrays and interference cancellation.
Related Article
"Monitoring Data Throughput in Wireless Apps"; www.commsdesign.com/design_corner/OEG20030805S0009.
Author's Note: The author would like to thank Sam Macmullan and Alex Flaig for their help with this article.
About the Author
Tandhoni (Tan) Rao (trao@ti.com) is a systems architecture manager for Texas Instruments Inc.'s wireless infrastructure business unit. He received his PhD in electrical engineering from Georgia Tech, an MSEE and BSEE from the University of Delaware, and an MBA from the University of Texas, Dallas.
