The silicon technology companies continue to stay ahead of the curve with ever improving chip technologies and architectures. The common Digital Signal Processor (DSP) design has worked in the past to satisfy modem design requirements. The release of 3G release99 puts the modem implementation beyond the capabilities of the DSP architecture. The big question is not whether faster processors are possible, but can the wireless networks keep up and deliver these higher speeds at reasonable cost.
I just upgraded to fiber optic service at 25/25mbps. Downloads are faster but the website response times have not improved. Will this be the same experience when faster wireless speeds are offered? I question whether faster wireless service will help. Faster wireless devices may be best demonstrated in advertisements but not in real life.
An article posted at EDN discusses cellular communications data speeds and their evolution; it’s entitled, "Why modems are going soft: A new approach to modem development separates the modem-specific software from the hardware and, therefore, has a profound impact on the platform design flow". Figure 1 in the article shows how transmission speed, which can be exponentially increased with new technology, is directly related to mega and giga operations per second of the processor. The DSP on the market can support about 1G operations per second. The newer standards are requiring processor speeds of 50G to eventually 300G operations per second.
The advent of High-Speed Downlink Packet Access (HSDPA), an enhanced 3G (third generation) protocol also called 3.5G, 3G+ or turbo 3G, allows networks based on Universal Mobile Telecommunications System (UMTS) to have higher data speeds. The current HSDPA deployments support down-link speeds of 1.8, 3.6, 7.2 and 14.0 Mbps. Higher speeds are possible with HSPA+, which supports downlink speeds up to 42 Mbps and 84 Mbps with Release 9 of the 3GPP standards. The increase in speed places an ever increasing burden on the wireless device processor. New processor architecture is required to continue the supporting the higher transmission speeds. The article anticipates transmission speeds up to 1Gbps for LTE-A by about 2015.
The article presents a technical examination of the evolution of Software Defined Modems (SDM). SDM will have to continually evolve to solve the processor requirements of the future. The article describes a new development process that calls for new integrated development tools and system design methodology. For example, the new wireless standards call for large amounts of instruction level parallelism, where many functions can be run concurrently rather than sequentially. The development tools that will be required for the evolving SDM are:
* A parallelizing compiler
* Instruction set simulator and profiler
* System design and modeling
* System debug and real time trace
This points out that a faster DSP will not be the solution for the future. Moving to a software designed solution that will make hardware centered solutions obsolete.
The article conclusions include:
Baseband architecture has evolved around hardware-centric solutions as a result of significant limitations in DSP, compiler, and silicon technology. The lack of functional re-use between successive mobile wireless standards has led to a large, disparate set of hardware functions, each often designed with its own uniquely dimensioned memory subsystem, resulting in large areas of silicon being unused for any one mode of operation.
The emergence of highly efficient parallel processors and compilers has demonstrated that the silicon area and power consumption of some vector signal processors are now sufficiently competitive for use in multimode LTE/3G-HSPA/GSM modems....
A competitive wireless platform can be achieved only if it is designed as a system. This involves correct dimensioning of the hardware resources for all modes of operation and an efficient software environment for code development and execution. The processor-instruction set and architecture are key to enabling the software approach and must be designed specifically to meet the power and performance requirements of the target application: battery-powered consumer wireless terminals....
SDM is highly disruptive; players that embrace the technology early will quickly win market share, as the combination of reduced time to market, reduced silicon cost, and faster response to requirements will place them in the pole position.
The emergence of highly efficient parallel processors and compilers has demonstrated that the silicon area and power consumption of some vector signal processors are now sufficiently competitive for use in multimode LTE/3G-HSPA/GSM modems....
A competitive wireless platform can be achieved only if it is designed as a system. This involves correct dimensioning of the hardware resources for all modes of operation and an efficient software environment for code development and execution. The processor-instruction set and architecture are key to enabling the software approach and must be designed specifically to meet the power and performance requirements of the target application: battery-powered consumer wireless terminals....
SDM is highly disruptive; players that embrace the technology early will quickly win market share, as the combination of reduced time to market, reduced silicon cost, and faster response to requirements will place them in the pole position.
So we can see how the new wireless processors will be able to deliver the higher compute speeds. The change to SDM will be beneficial but without the right tools will cause problems in the development of the new software.
This all sounds great. Can the network providers invest enough to support these higher speeds? It appears that the capital available for the providers will have to be mostly self generated through lowering costs and improving efficiency. It appears hard to raise wireless charges to create new capital at this time, given all the competition. The investment in ever increasing wireless speeds may slow due to capital restrictions. I think that the network improvement to offer higher speeds will also create an even bigger digital divide, leaving out low income, rural and under populated areas of the U.S.
