DSP permeates the wireless infrastructure
The size of cellular and PCS base stations continues to shrink in direct proportion to the amount of digital signal processing (DSP) employed to perform the functions traditionally handled by analog components.
The “bunker” has been a permanent fixture at base stations and cell sites of most types of wireless communications service, from SMR to paging and cellular. The familiar tower and equipment shelter contain all of the wireless infrastructure components, from baseband to switching, RF receive and transmit subsystems, transmission lines and antennas. While hardly lavish, the equipment shelter affords the service technician a place to set up shop, and a well-controlled environment in which to operate electronic systems. Now, all of this is changing.
The feverish pace of wireless network build-outs and intense competition between cellular, personal communications services (PCS) and enhanced specialized mobile radio (ESMR) has produced a common goal: to develop, implement, maintain and upgrade base station equipment at the least possible cost. Inherent in this mission are the following wireless infrastructure design goals:
* Reduce base station size.
* Reduce the number of components required to achieve functionality.
* Make the station upgradeable in place by means of software.
* Reduce station power consumption.
* Ease the nightmare of siting a base station by making its physical appearance more palatable to the community.
The architecture of PCS simplifies this task considerably; these networks transmit at much lower RF power levels than their large-cell cellular counterparts. This eliminates the need for high-power RF components that consume large amounts of dc input power and inherently reduces the size of the base station. Unfortunately, PCS networks require more base stations to serve a prescribed geographic area, which increases overall network cost.
Consequently, base station manufacturers are looking for all possible ways to reduce costs_right down to the component level. In this task, they are aided immeasurably by digital signal proceessing (DSP).
What was once analog_ By performing signal processing functions in the digital domain, DSP has the ability to affect all of the aforementioned wireless infrastructure subsystem design goals. Not surprisingly, DSP is replacing some analog components, and it is driving base station cost to levels more than an order of magnitude below those of their least expensive cellular counterparts as shown in Figure 1 on page 16.
It is essential to remember that once a signal has been converted to the digital domain, an entire new world of opportunity emerges to manipulate it. Large numbers of analog components are no longer needed because a few DSP-integrated circuits (ICs) and related components can perform all of those analog functions for less money, with less power consumption, in a fraction of the size and weight.
Technicians can remotely command DSPs to modify their functions and implement new functions via software, without replacing a single component. For service providers, this means fewer trips to the field by their growing armies of technicians.
DSP design flexibility Only in the past five years has DSP technology progressed to the level needed to perform some of the most demanding base station functions, such as speech encryption and orchestration of complex higher-order modulation techniques. Previously, the application-specific integrated circuit (ASIC) was the obvious choice because it could be designed to perform many functions within a single chip, and it could be tailored to meet the needs of specific applications and even specific products.
For this reason, the ASIC has become the basis of many electronic products. However, the ASIC is not a programmable device. Changing functions requires redesigning the ASIC and retrofitting it within platforms in the field. Even under the best circumstances, this requires months of design and prototype effort and at least one pass through the foundry.
Consequently, the ASIC design alone may not give designers the flexibility needed for future applications in the wireless infrastructure system. Nevertheless, the first code-division multiple-access (CDMA) PCS base stations to be installed (most of them in 1997) rely on ASICs because of the complex air interface and high data rate specified in the IS-95 standard. However, the rapid progress of DSP technology should make it applicable for CDMA sometime in 1998.
The digital wireless environment The air interface standards with which base station manufacturers must comply rely on one of two access methods and their variants. The first is time-division multiple-access (TDMA), and the second is CDMA. Each method relies on a higher-order modulation scheme and a protocol that are the medium and format by which information is communicated.
Each of these standards has different specifications. Their data rates vary considerably_a fact that directly affects the ability of DSP to handle their signal processing functions. For example, the IS-54 (TDMA) standard specifies a data rate of 48.6kbps, while GSM (a TDMA variant) has a much higher 270.8kbps rate. CDMA transmits data at a lofty 1.288Mbps, as shown in Table 1 on page 18.
As a result, DSP solutions for wireless infrastructure applications have begun with lower-data-rate TDMA applications and are progressing through development of demodulator and equalizer subsystems for GSM. As noted, CDMA DSP will soon be available.
The vocoder example DSP-based solutions were first realized in wireless base stations within the base station controller. Early circuits were basic and required one processor per channel of compressed, digitized speech. When a call was set up, the mobile user was assigned a single vocoder resource, much like a modem is assigned.
The processing power of DSPs has grown, and vocoder algorithms have increased in sophistication. Service providers eagerly adopt each incremental advance in speed processing power because each advance allows better quality communications for their customers. However, to accommodate the many handsets in the field using early vocoders, service providers have adopted a flexible vocoder architecture that uses a pool of general-purpose DSPs to provide multiple vocoder implementations as shown in Figure 2 on page 20.
The software associated with a particular service provider’s vocoder options is stored in the base station controller database along with information about the amount of processing resources, specified in millions of instructions per second (MIPS) and the available memory required by each vocoder.
Rather than dedicating a DSP resource on single-channel-per-resource basis, a pool of DSP resources is created. This pool consists of an array of RAM-based DSPs such as Texas Instrument’s TMS320C6X series. Each DSP can support at least a single implementation of the most complex vocoder contained in the vocoder database.
When a call is set up, the proper vocoder is selected from storage to respond to the mobile user’s vocoder and is downloaded into an unused portion of the transcoder’s DSP-based resource. The process is reversed in call breakdown, which frees up DSP resources. The base station controller keeps a running total of available DSP resources and manages them accordingly.
As more powerful DSPs become available, they can increase base station controller DSP resources by the simple act of swapping out chips. The advantage of this approach over assigning one vocoder element per caller is best seen as system requirements change. When a new vocoder is adopted, the software and DSP resource consumption of this new option is simply added to the base station controller’s database. There are no more changes to make. Because new vocoders appear in current access methods once every two years or so, the advantages of this programmable DSP-based approach are formidable.
More power on the way The power of DSPs is being realized in more and more areas of digital communications systems, but their capabilities have barely been exploited. Higher-speed DSPs, combined with higher levels of ASIC integration, will make an even greater impact on the modem subsystem of the base transceiver station (BTS). This will allow base stations to employ software upgrades, even active noise cancellation in the communications link, without major hardware replacement. The increased performance and higher integration of DSP ICs will also enable multiple vocoder algorithms to run on a single DSP of the transcoder, giving service providers further flexibility.
Software radio The “software radio concept” promises to improve every critical-performance wireless base station parameter. In the ideal implementation of this technique, the signal is captured at its input frequency and converted at that frequency to the digital domain. Beyond this point all of the signal processing is digital, which eliminates most analog components and enables simultaneous processing of many signals to be conducted by a few DSP-based circuits.
Software radio advantages are almost entirely realized through DSP. At present, DSP technology cannot process signals directly at the input frequency of PCS systems (1.9GHz). However, it is likely that within two to three years this obstacle will be hurdled. When it occurs, the size, cost, and power consumption of PCS base stations will decline precipitously. A single base station will be configurable to transmit and receive according to the specifications of multiple air interface standards and will implement new features through changes in software alone.
A growing number of DSP ICs are emerging to serve the demands of wireless infrastructure. Base stations require DSPs to have a high level of functional integration, large amounts of on-board memory, low power consumption, a wide array of I/O such as buffered serial ports, and a comprehensive set of development tools. Texas Instruments’ TMS320C6x series of DSPs is an example of how DSP can be tailored for base station requirements.
The TMS320C6x design is based on the company’s very long instruction word (VLIW) architecture, which increases the parallel execution of instructions by packing as many as eight 32-bit instructions into a single cycle. The advanced C language compiler generates code that executes on highly independent functional units with triple the efficiency of previous fixed-point DSP compilers. The architecture is scalable to allow future enhancements.
This development path allows customized, programmable solutions to be created for applications such as CDMA base stations, which were previously the exclusive domain of ASICs. Using the TMS320C6201 architecture, it is possible to implement a base with 30 extended, full-rate (EFR) GSM channels for $3 per channel, compared to five channels at $7 per channel with previous solutions.
Street is wireless infrastructure marketing manager for the Wireless Communications Business Unit of Texas Instruments, Dallas.