Software defined radio (SDR) is a technology that allows the functionality of a radio transceiver system to be defined by software, thus offering the flexibility to support multiple radio standards and multiple interfaces on a common hardware platform. Changes to the functionality or the interfaces of the SDR can then be made by developing the appropriate software. Upgrades to existing modules that provide new features and new radio standards need not require changes to the hardware, thus simplifying the roll out of new services. The flexibility at the interfaces of the SDR allow the base station manufacturer to select from a wider range of peripheral modules, thus opening up their selection and driving down costs. The earlier deployment of new services allows network operators to increase their ARPU, whilst end users also benefit from the earlier deployment of advanced mobile services at competitive prices.

However, this implies the need for a very different design approach to that traditionally used in hardware design. In particular, the techniques used to minimise lifetime cost need to be based around an understanding of the costs and benefits of SDR. These benefits are primarily associated with the flexibility we talked about above, and the resultant improvement in time-to-market. Critically, flexibility must be designed into the SDR architecture from the start in order to reap the full benefits.

Amongst the key issues are;

•  Using the DSP processing resources efficiently, particularly when supporting variable bandwidth services such as those provided by WCDMA standards.

•  Providing the correct interfaces to work with a range of sub-systems, which may or may not comply with interface standards proposed by bodies such as OBSAI or CPRI

•  Adopting a layered software design appropriate to the complexity and range of radio standards required of the product

This article describes our views on how to handle these issues, based on a practical solution we have implemented.

  Adaptive DSP resource allocation maximises channel card capacity

The first major consideration is to reduce the cost through intelligent design, by optimising the usage of each DSP device. A static allocation policy, in which a channel is allocated to a specific processing resource such as an ASIC for its duration, requires that sufficient margin be allowed at the initial allocation to accommodate peak loading of that function, leading to inefficient use of the processing resource for the majority of the time. In contrast, a dynamic allocation policy, in which channels may be re-allocated to different flexible processing resources such as DSPs for the duration of the call, can result in greater capacity for a set number of devices.

For example, a 3G channel requires a significant amount of processing power to perform the baseband processing. One DSP may be powerful enough to handle a small number of channels as might be handled by a pico-station, whereas more demanding applications such as base stations will require a number of DSPs collected together to form a cluster. Typically an SDR to support 128 simultaneous voice users would require a cluster of 6 DSPs, though this number will fall as more powerful DSPs are brought to the market.

With dynamic resource allocation, the process to determine which DSP in the cluster will receive a new channel involves comparing the processing requirements of the new channel with the remaining processing power available to each DSP. The processing requirements for each type of channel are obtained from metrics that are established during development and may be used when the channel is created. However, the processing requirements of a channel are likely to change during the lifetime of the call, for example, during handover between cells, and the dynamic resource management may then re-allocate resources as required to maximise the capacity of the system.

The allocation of resources to the requirements can be made on a per-user or per-function basis. In the former a whole channel is processed by one DSP, whereas in the latter the processing of one channel may be partitioned between multiple DSPs. In both cases, the success of the system depends critically on the availability and accuracy of the processing metrics, and in the case of dynamic allocation, on the ability to re-allocate resources seamlessly.

Software-defined interfaces allow commoditised multi-module support

Base station manufacturers have traditionally developed the interfaces between modules independently and, as a result, similar modules or sub-systems produced by different manufacturers are not interchangeable. However, the base station industry is increasingly looking for ‘open' suppliers, who can provide competitive modules to a common specification. There is an initiative in place to standardise these interfaces between these modules, as promoted by the body OBSAI , but until such standardised interfaces become widely adopted and rolled out there is still a requirement to support multiple interfaces.

For maximum flexibility, the logical interfaces of any SDR module should therefore be designed to support alternative requirements such that different software builds can be used depending on the particular requirements of the interface. Thus a baseband module can be produced that will work with a variety of transport, RF and control modules within a base station architecture. This is of significant advantage to network operators who will benefit from the ease of upgrading elements of the base station as more attractive commoditised modules appear on the market.

tructured design simplifies support of multiple radio standards

The different 3G standards such as UMTS and cdma2000 can be conveniently broken down into similar layers of functions. The lower layer handles processes primarily concerned with chip-rate and some symbol-rate processing, such as spreading and de-spreading, path search, modulation and demodulation, coding and decoding, interleaving, and physical and transport channel implementation. The upper layer handles processes such as data and control path interfacing to higher layers, control of the DSP resources, interfacing to the transceiver and provision of a clean interface to higher layers to hide the complexity of the DSP processing.

With the core functional elements of each layer determined, a framework of control software can be custom written for each radio standard required. Extensive re-use of the software is feasible between WCDMA standards since the basic principles underlying each are very similar, so with suitable interfaces to the key functional blocks, future adaptation to support other similar radio standards is very much simplified. Similarly the timescales for adding new services to existing standards are reduced with a well-structured architecture.

In summary, we propose that by designing SDR systems from the onset with flexible processing resource allocation and inter-module interfaces, and by supplementing this with a well structured, layered software architecture, real benefits may be realised by manufacturers and network operators alike.

This article is based on PA's experience of designing both SDR and hardware-defined radio systems. For example, one of our recent implementations of an SDR system uses an array of four ADI TS201 TigerSHARC DSPs and supports the W-CDMA and cdma2000 (1xRTT and EV-DO) radio standards. PA has expertise in the design of radio products supporting multiple standards and offers a full range of services to assist in the design of SDR systems.

Tel +44(0)1763 267492 email: innovation@paconsulting.com www.paconsulting.com/wireless

Dr. Andy Thurston is a consultant in the Wireless Technology Practice of the PA Consulting Group. He has been involved with a variety of WCDMA projects over the last six years, working mainly on physical-layer software development and on the integration and testing of modules in mixed-manufacturer WCDMA products.