BDTI’s DSP Insider Archives
|
||
BDTI’s DSP Insider Archives |
|
|
| HOME << |
|
|
|
||
This month:
Special Feature: New Chips for Wireless ApplicationsEditor’s Note: In the last month there has been a deluge of chips announced for wireless products. Many of these announcements came from industry leaders, and many of the announcements introduced significant new technologies. While we can’t cover all of the significant announcements from the last two months in this issue of the DSP Insider, we’ll highlight some of the most notable developments in the following feature article. This February, Samsung, STMicroelectronics, and Texas Instruments announced new “application processors.” Application processors are intended for portable wireless products—particularly “smart” phones and wireless PDAs—running a powerful operating system like Pocket PC, Symbian, or Linux. These products have two types of computational loads. The first type of load is the “application” software, which includes the operating system and software like Web browsers and MP3 decoders. Wireless communications software constitutes the second type of load. As the name implies, an application processor is intended to run the operating system and application software, but not the communications software. Like most existing application processors, the new processors—the Samsung S3C2410 SiP, the ST Nomadik, and the TI OMAP161x—are all based on the ARM architecture. These new processors are also alike in that they all use die stacking to place large memories inside the same package as the processor. Despite these basic similarities, these new processors are remarkably different. For example, the processors take widely differing architectural approaches to handling multimedia tasks. In addition to its ARM core, the ST Nomadik contains two “smart accelerators,” one for audio and one for video. Each accelerator consists of a ST MMDSP+ DSP processor core and audio- or video-specific hardware. In contrast to ST’s dual-accelerator approach, the TI OMAP161x supplements its ARM core with a single TI ’C55x DSP core. The Samsung S3C2410 SiP takes an even simpler approach: it contains no multimedia-specific features. Each of these approaches has tradeoffs. On one hand, specialized hardware tends to be more efficient than general-purpose hardware. On the other hand, specialized hardware tends to be harder to program and harder to adapt for unforeseen applications than general-purpose hardware. As mentioned earlier, application processors target products with two types of computational loads: application software and communications software. One way to handle these loads is to use an application processor for the application software and a separate “baseband” processor for the communications software. This is the approach taken by ST, Samsung, and TI application processors discussed above. This approach is attractive for products with demanding application and communications loads such as high-end phones. The main drawback of this approach is that it is relatively expensive. For mid-range products with less demanding loads, it may be possible to use a single “integrated” processor that handles both the applications and the communications loads. Compared to the two-processor approach, the single-processor approach may offer less computational power, but it may also reduce chip counts—and hence system costs and size. Last month Intel and Texas Instruments both announced integrated application/baseband processors. The TI OMAP73x contains three cores: an ARM core and a TI ’C54x DSP core for baseband processing, and a separate ARM core for application processing. Like the TI OMAP161x, the TI OMAP73x is available with “stacked” memory. In contrast to TI’s three-core approach, the Intel PXA800F contains only two cores: an XScale core (an Intel-specific derivative of the ARM architecture) and a Micro Signal Architecture core (a DSP core co-developed by Intel and Analog Devices). Baseband processing is split across the XScale core and the Micro Signal Architecture core; the XScale core also handles application processing. Like some versions the TI OMAP73x, the Intel PXA800F features large on-chip memories. However, the Intel PXA800F does not use stacked memory; the PXA800F processor cores and memories are combined on a single die. While processors are key components of cell phones, PDAs, and other wireless products, these products are complex systems with numerous important hardware and software components. This complexity has two important implications. First, cell phone and PDA technologies are maturing, leaving fewer opportunities for innovation at the “nuts and bolts” level of design. As a result, many product designers prefer to focus development efforts on high-level features like eye-catching packages. Second, even where there are opportunities for technological innovations, the increasing complexities of cell phones and PDAs mean fewer companies can take advantage of these opportunities. As a result, “end-to-end” solutions that integrate the critical hardware and software components are much more attractive than stand-alone processors. A number of companies announced end-to-end cellular handset solutions in the last month, including Analog Devices, Intel, and Texas Instruments. Of these announcements, the most significant may have been TI’s announcement of a 3G chipset, the TCS4105, which can be used with or without a TI OMAP161x application processor. No single feature of this chipset is particularly unusual. However, few other vendors can match TI’s portfolio of wireless-oriented components, which includes application processors, baseband processors, and analog components. As a result, there are few single-vendor offerings that are as complete and as integrated as the TI TCS4105.
The bevy of processors and chipsets announced in the last months is a
mixed blessing for designers of wireless terminals. The good news is
that there is a wealth of new choices available to system designers.
The bad news is that choosing a solution will require careful
evaluation of a large number of complex offerings.
BDTI Case Study
This Month: Porting Multimedia PC Apps to Multiple Embedded PlatformsAs consumer electronics gain powerful multimedia capabilities, many PC multimedia software applications are being adapted for use in these products. Unlike the PC marketplace, which is dominated by the x86 architecture and the Windows operating systems, consumer products use a wide variety of processors and operating systems. Therefore, to succeed in the consumer arena, PC multimedia applications typically must be adapted to run on many embedded platforms. The process of adapting a multimedia software application to run on a wide range of embedded platforms can be streamlined by first creating a processor-independent, “embedded-friendly” implementation of the application, sometimes called a “porting kit.” This porting kit can then be adapted for specific platforms as needed. Each such adaptation requires much less effort and carries much less risk than adapting software directly from the PC version. This reduction in effort and risk has numerous benefits, including an increased likelihood that the application will succeed in the consumer space. Creating a high-quality porting kit for a PC multimedia application presents several challenges. While PCs have vast processing power and memory resources, embedded platforms typically have tight processing and memory constraints. Therefore, it is important to optimize the porting kit in ways that will reduce the processing requirements and memory footprint of the final implementations. This can be challenging because optimizations that reduce performance and/or memory requirements on one processor may have the opposite effect on another processor. The goal in creating a porting kit is to use optimizations that will be effective on a wide range of embedded processors. Another challenge flows from the difference between the data types used by PCs and by embedded processors. Most embedded processors are fixed-point machines, but PC multimedia applications—particularly audio applications—often use floating-point arithmetic. As a result, developing a porting kit often involves converting a floating-point algorithm to a fixed-point algorithm. The challenge in this conversion lies in maintaining sufficient precision and dynamic range throughout the application to meet the relevant signal quality requirements. Creating an effective multimedia application porting kit requires in-depth knowledge in three areas: signal processing theory, processor architectures, and embedded software development. For example, identifying optimizations that apply to a wide range of processors requires a broad knowledge of optimization techniques and of the effectiveness of these techniques on a variety of processors. Similarly, converting a floating-point algorithm to a fixed-point algorithm requires a thorough understanding of algorithms as well as experience in testing signal processing software to ensure quality. BDTI has extensive experience in all three of these areas, and has used its skills to create porting kits for a variety of multimedia applications. For example, in several porting-kit projects, BDTI verified that the porting kit code was indeed portable by testing it on a variety of embedded platforms. As part of these porting kits, BDTI also created embedded-oriented test infrastructure (test vectors, test benches, and documentation) to help porting kit users quickly verify their processor-specific implementations.
For more information on how BDTI can help you build a porting kit for
your PC multimedia application, contact Jeremy Giddings
(giddings@BDTI.com), or visit
http://www.bdti.com/products/services_software.html.
Impulse Response, by Jeff Bier
Follow the LeadersIn January, IC Insights released a study showing that the four largest semiconductor vendors in 2002 were Intel, Samsung, Texas Instruments, and STMicroelectronics. Last month, these same four companies all announced processors targeting wireless products like cellular phones and wireless PDAs. Is it a mere coincidence that the world’s largest semiconductor vendors seem to be in lock-step pursuit of wireless products, or is this a sign of a fundamental shift in semiconductor markets? To answer this question, consider three key trends in the semiconductor industry. First, there is a longstanding trend of migration away from expensive, centralized computing towards inexpensive, decentralized computing. History has already recorded the progression from mainframes to minicomputers to PCs; PDAs, cell phones, and smart displays are the next logical step in this sequence. Secondly, electronic systems are increasingly characterized by their communications capabilities. Consider the impact of file-sharing networks on the PC. For music lovers, a PC that can access libraries of digital audio from the Internet is far more compelling than a PC that can access only a few locally stored songs. Consequently, the speed of the Internet connection has become more important to many users than the speed of the processor. Wireless communications promises to cause similar realignments of priorities across a wide range of electronic systems. While these trends suggest wireless communications holds huge opportunities for chip vendors, a third trend, increased complexity, suggests these opportunities are available only to a few vendors. Wireless communications standards are becoming more complex, and wireless communications is increasingly combined with other complex features. For example, early cellular phones offered simple features like a basic address book. In contrast, today’s high-end phones include powerful features like Web browsers and built-in cameras. Crafting processors that meet the needs of these complex products requires enormous engineering resources—resources that may only be available to the largest semiconductor vendors.
Taken together, these three trends suggest that it is no coincidence
that the world’s four largest semiconductor companies are increasingly
focused on wireless applications. Indeed, success in wireless
products may be the factor that separates leading companies from
also-rans in the coming years.
BDTIsimMark2000™ for the NEC SPXK5BDTI has released a BDTIsimMark2000™ score for the NEC SPXK5 DSP core. This score is based on recent benchmarking of the SPXK5 using the BDTI Benchmarks™. For this score, as well as BDTImark2000™ and BDTIsimMark2000™ scores for other processors, go to http://www.BDTI.com/bdtimark/BDTImark2000.htm.
The BDTIsimMark2000 and BDTImark2000 are summary measures of DSP speed
distilled from a suite of DSP benchmarks developed and independently
verified by BDTI.
“The Future of DSP Engineering” at ISPC/GSPx in DallasMake plans to join BDTI General Manager Jeff Bier and leading members of the DSP community in a panel discussion on “The Future of DSP Engineering” at ISPC/GSPx in Dallas. Come listen to prominent industry figures as they consider the challenges DSP engineers must face and the skill sets needed to face them. The event is scheduled for Wednesday, April 2 at 1:00 PM. Also plan to attend BDTI’s two conference presentations, DSP Benchmarks for the Latest Processors and Evaluating FPGAs for Communications Infrastructure Applications. DSP Benchmarks will include results for several new processors, including TI’s OMAP and the Intel XScale. Evaluating FPGAs will provide important insights into the strengths and weaknesses of DSP-enhanced FPGAs for signal processing applications—including the first-ever independent DSP benchmarks for FPGAs.
ISPC/GSPx will be held at the Hotel Intercontinental in Dallas, Texas,
from March 31 to April 3, 2003. More information is available at
http://www.BDTI.com/bdti_whatsnew.html#gspx.
BDTI Spotlights Consumer Media Products at ESC SF 2003The new Consumer Electronics track at this year’s Embedded Systems Conference features classes that provide important perspectives and key insights for engineers, marketers, and managers—including four of BDTI’s highly rated classes. The Consumer Electronics track also features a panel discussion moderated by BDTI’s Jeff Bier. This panel brings together leaders from content providers, home entertainment equipment manufacturers, and consumer advocacy groups to discuss “Key Challenges for Consumer Media Products.” Come hear how these diverse constituencies see the challenges impeding the widespread adoption of new consumer media technology.
For more information and a free exhibits pass, go to
http://www.bdti.com/bdti_whatsnew.html#esc.
BDTI Products and ServicesAre you aware of these BDTI products and services?
You will find more information on these and other BDTI products and
services at http://www.BDTI.com. Or contact Jeremy Giddings at
giddings@BDTI.com.
About BDTIBDTI is an independent source for DSP technology analysis and optimized DSP software. From rigorous technical analyses of processors for DSP, such as the Inside series of processor analyses, to highly regarded technology training classes, BDTI is the trusted independent source for reliable information on DSP technology.
For more information, visit our Web site at http://www.BDTI.com.
The next issue of BDTI’s DSP Insider is coming in April. Previous issues of BDTI’s DSP Insider are archived on BDTI’s Web site. Follow the link from http://www.BDTI.com/dspinsider.htm. If you have comments, suggestions, or other feedback about the DSP Insider, please send email to dspinsider@BDTI.com. BDTI’s DSP Insider is a free monthly electronic newsletter published by Berkeley Design Technology, Inc. If our newsletter was forwarded to you and you would like to receive it regularly, please register at http://www.BDTI.com/dspinsider.htm.
If you no longer wish to receive the DSP Insider send an email
message to dspinsider@BDTI.com with the words “Remove me” in the
subject line.
BDTI’s DSP Insider © 2003 Berkeley Design Technology, Inc. |