The concept of a fully software-programmable SDR is very appealing to mobile wireless handsets. To some extent, a wireless handset terminal must be able to accomplish similar functionality as an SDR. While in the current generation worldwide coverage and operation on multiple standards is matter of convenience, future generations — 3G and beyond — explicitly require multiple modes of operation (GSM/GPRS/EDGE and FDD WCDMA) with different wireless connectivity standards included in the handset (Bluetooth, wireless LAN).
An ideal multi-mode handset terminal would be a single hardware device capable of operating in any given wireless standard with the appropriate software program. And, again, in theory, such a device would be cost-effective since the same hardware would be used for all operating standards. In addition, if such a fully programmable wireless terminal were to exist, it would not only be possible to change the elements associated with air-interface standard, but it would also be possible to implement arbitrary application-level functionality by downloading executable programs, for example, security functions and complex video codecs.
In reality, however, first multimode solutions rely on parallel (selectable) single mode approaches, with a clear path to reconfigurable hardware — ultimately arriving at a SDR slam-dunk. SDR skeptics are fast to point out that software programmability may be too expensive, that cost of a hardware device increases significantly in proportion to its capability to perform more than one function. And, in an extreme case of a largely programmable device, the cost becomes prohibitive, which is a particularly severe constraint in the cost-sensitive and cost-driven wireless terminal market.
Our view is that the SDR, as it relates to wireless terminals, will come in a series of steps moving from hardware selectability to hardware reconfigurability and finally to true software reconfigurability. As such, the key question is not a purist view of whether a handset terminal can be fully SDR, but rather the extent to which it is, and is likely to be. If we examine handset terminal designs over the course of the last ten or fifteen years, to a certain degree many functions have in fact migrated from a rigid, fixed form to a more generic, programmable one. Certain functionality lends itself naturally to this transition, such as functions associated with channel and speech coding, while others have had a slower transition, such as radio functions.
Although we do not believe that in the short term a cost-effective SDR-based terminal can be successfully mass produced — judging by the way that terminals have evolved — it is very likely that in the future the degree of programmability will continue to increase. This is the reason our engineers strongly believe in the role of programmable hardware, which we have been pursuing over the course of several wireless terminal generations.
For example, ADI's SoftFone architecture is RAM-based, in contrast to earlier implementations of handset chipsets, where much of the software was hard-coded in mask-ROM. The lower cost of ROM is offset by the difficulty of changing software every time a new feature is added. The RAM-based SoftFone allows a single handset platform to be software-configured to add new features or update the control software to keep up with the evolution of the GSM standard. Time-to-market and flexibility advantages favor this approach.
The architecture is designed to allow simple substitution of the DSP and MCU cores as more powerful cores become available or necessary. It allows either processor to access either on-chip or off-chip memory. Sufficient cache memory is included to obtain the best possible performance from each core for time-critical functions. On-chip multi-clocking capabilities allow multiple standards to be supported.
The external mixed-signal chip allows different analog features to be included, from receive and transmit converters compatible with single-mode GSM/GPRS, to higher-performance devices with sufficient dynamic range for EDGE, or multi-mode converters capable of GSM/GPRS, EDGE, and WCDMA. Additional audio features such as MP3-quality audio converters can also be implemented.
There are several technical considerations that will drive the trend to programmability in the signal chain. First, radio is the toughest nut to crack. Looking at a wireless terminal, the radio is usually responsible for type approval delays, interoperability testing, and production releases. The inherent complexity of the radio design is in the effective tradeoff between performance — dynamic range, spurious emissions, sensitivity, to name the few — and product constraints such as power and cost, usually leading to a low level of flexibility. At the same time, analog radio solutions provide significant power reductions for down conversion and filtering, as compared to digital techniques, and also are the technology of choice in wireless handsets.
From an SDR perspective, fixed radio functions are the first to be removed (fixed bandwidth, frequency band, frequency plan) in order to achieve flexibility. In reality in current generation of wireless handsets even multi-band and limited multimode solutions offer little in terms of reconfigurability. However, advances in direct conversion technology have cleared the way to commercial products, and the next step is to provide reconfigurable analog hardware that supports the required performance for multiple standards, while minimizing the cost increment due to the migration from single to multimode terminals. Enabling technologies rely on a combination of analog blocks, including re-configurable analog filters, programmable frequency synthesizers, and the capability of the digital section to complement functionality via different functions such as digital pre-distortion of transmitters and adaptive radio impairment correction. The chances of manufacturing wireless terminals with reconfigurable radio functions are slim before the market really puts pressure for cost-optimized multimode solutions.
Problems carryover
Meanwhile, the same themes and problems that dominate the RF portion of SDR design emerge in the mixed-signal portion of the handset. Configurability starts with the architecture selection even before a single circuit is designed. Since all filters need to be adaptable and therefore more complicated, there is a strong desire to reduce the total number of filters, leading to the dominance of direct conversion architectures. These architectures typically bring the converter problem down to a baseband signal, but can put heavy demands on the dynamic range of the converter, and call for adaptable bandwidth approaches. Oversampling sigma delta techniques are well-suited to these requirements, and the circuits conferences have started to feature a number of papers on converters designed to reconfigure from 200kHz (GSM/GPRS/EDGE) to 5MHz (WCDMA) bandwidths with minimized power consumption in each configuration. There are similar challenges on the transmit side in supporting the different requirements of vastly different modulation schemes. To realize the required power efficiency, these schemes will ultimately require significant digital processing to compensate for analog non-idealities. One example is digital pre-distortion to compensate for non-linearities in the transmit path, allowing for greater transmit power efficiencies.
The digital base-band (DBB) is the component of the wireless handset with the highest degree of reconfiguration. Given the advances of the SDR, particularly for multi-mode operation, modern DBB DSP architectures have undergone significant changes, including:
- High computational capability;
- Memory size and bandwidth (both on-chip, and interface for off-chip memory);
- I/O bandwidth;
- Fast interrupt response times to sustain DSP system performance;
- Ability of the processor to execute typical control-oriented tasks.
Many functions found in channel coding, and in speech coding, have over the years migrated from hard-wired ASIC implementations into software. This transition has been possible mostly due to advances in DSP technology. For example, trellis-based GSM channel decoding and equalization, have until recently required some form of hardware acceleration external to the DSP. Today, DSPs generally have sufficient built-in support for these functions, and we expect this trend to continue. In EDGE, where data computation is far greater than it is in GSM/GPRS, modern DBB DSP architectures provide instruction-set support for most channel equalization functions, filtering and trellis decoding.
Software downloads are present today in wireless terminals via the SIM Application Toolkit, offering capability to download simple menus, further extended by Wireless Application Protocol (WAP) which allows more complex functionality. A further evolution path is via the Mobile Station Application Execution Environment (MExE), based on WAP and later on personal Java. In the future, full dynamic ReConFiGuration (RCFG) is expected in 3G systems.
SDR based terminals will not be present in the first generation of 3G technology to achieve instant gratification in newly launched systems, neither in commercial nor in technical elegance terms. We are looking at an uphill battle for optimized technical solution coupled with cost and power constraints. And yes, the history will repeat itself: technology will need time to mature. We may be discussing whether we can accelerate SDR migration, but working hard to establish it while it is still maturing is inevitable.
