Embedded Processor Watch

Senior Editor: Tom Halfhill

Contributors to this issue: Peter Glaskowsky, Senior Analyst; Jennifer Eyre and Jeff Bier, Berkeley Design Technology, Inc.; Michael Slater, Founder, MicroDesign Resources

In This Issue:

• Infineon Targets 3G with Carmel2000
• TriCore Turns 1.3
• JEDI Gets $18 Million Funding
• Slater Perspective: A New Era for Photography...and Me

Infineon Targets 3G with Carmel2000

By Jennifer Eyre and Jeff Bier Berkeley Design Technology, Inc.

Processor vendors everywhere want a piece of the third-generation ("3G") wireless market, which many expect to be the next DSP "killer app." Today's 2G digital-cellular phones are consuming hundreds of millions of 16-bit DSPs and 32-bit MCUs. Third-generation devices, however, will have performance requirements far beyond the capability of even today's fastest DSPs, including application-layer tasks like video decompression. And, of course, this blazing performance will have to be matched with extremely strong energy efficiency and reasonable cost. At Embedded Processor Forum 2000, Infineon explained how it will upgrade its Carmel DSP core to address the demands of 3G wireless applications through the addition of tightly coupled application-specific hardware accelerators.

With Carmel2000, Infineon takes its concept of "parallelism where you need it" a step further, allowing SoC designers to create hardware accelerators that can be tightly integrated with the basic Carmel architecture to speed up execution of inner loops even more. These accelerators, called PowerPlugs, can be thought of as an extension of the configurable-long-instruction-word concept. In addition to configurable instructions, you can customize the core itself to allow more parallelism where needed.

This approach blurs the line between custom hardware and programmable processors in an attempt to combine the best of both worlds, and Infineon is not alone in exploring this avenue. Indeed, both ARC and Tensilica allow designers even more latitude in customizing their processor cores.

But while PowerPlugs can provide a significant performance boost, 3G requires more than just computational power -- it also requires extremely high data bandwidth. Carmel2000's data bandwidth is constrained by the memory architecture of the original core. Although Carmel's data bandwidth is high compared with that of most other DSP processors (Carmel can read four independent 16-bit data words/cycle), it will likely prove insufficient for the bandwidth demands of a full 3G application.

A Carmel2000 with PowerPlugs that contain their own memory will be able to make more efficient use of that bandwidth, but it most likely won't be able to meet the roughly order-of-magnitude increase required by 3G. So, although Carmel2000 is a step closer to a single-core 3G solution, it appears we will have to wait a while longer before we see processors capable of implementing a complete 3G terminal without the aid of powerful off-core coprocessors. (The full version of this article is available online to Microprocessor Report subscribers at http://www.MDRonline.com/mpr/h/2000/0717/142902.html).

TriCore Turns 1.3

By Peter N. Glaskowsky

At last month's Embedded Processor Forum, Infineon announced version 1.3 of its TriCore architecture and described a "V1.3 MicroProcessor System" macro meant to be combined with additional application-specific logic. The design includes a 3mm^2 CPU based on the new core architecture that runs at speeds up to 200MHz in a 0.18-micron process.

At the Forum, Infineon also described how to integrate four of these macros on a single die, connected by their peripheral-bus interfaces. Though the small size of the TriCore CPU core makes this plan affordable, software development for such a chip would be more complex than for single-core processors.

Infineon expects to begin sampling application-specific standard products (ASSPs) based on the V1.3 core in late 2000. The new core is also available for licensing. More information is available online at http://www.infineon.com.

JEDI Gets $18 Million Funding

JEDI Technologies has raised $18 million of second-round funding from Alliance Ventures, Ignite Group, New Enterprise Associates (which previously gave JEDI $5 million in first-round funding), and Redwood Ventures. JEDI, founded in 1998 and based in Santa Clara, Calif., licenses JSTAR, a synthesizable coprocessor that translates Java bytecode instructions into the native instructions of a host microprocessor (see Embedded Processor Watch #92, http://www.MDRonline.com/epw/issues/epw_92.html). The additional funding gives a boost to JSTAR at a time when Java is attracting more attention from embedded-system developers, thanks partly to a real-time Java specification created by a consortium of companies. For more information, go to http://www.jeditech.com. --T.R.H.

A New Era for Photography...and Me

By Michael Slater

As followers of my columns have no doubt noticed, my interests in the past few years have shifted from microprocessors themselves to their applications, especially in consumer products. I have been particularly excited about the prospects for information appliances, and for digital photography.

Drawn once again by the siren call of entrepreneurship, I have launched a new business -- after 13 years at MicroDesign Resources -- to pursue both information appliances and digital photography: PhotoTablet, Inc. We are developing hardware, software, and services to make it possible to effectively use a digital camera without a PC. I will continue to contribute occasionally to Microprocessor Report, primarily in the area of information appliances, and I will work with the MDR team to ensure the successful continuation of the Microprocessor Forum and Embedded Processor Forum conferences.

I've written in the past (see "Digital Photography Coming of Age," http://www.MDRonline.com/slater/photoage/photoage.html) about the advantages inherent in digital cameras, most notably the instant viewing of pictures and the elimination of film and processing expense. The cameras have advanced another generation since I wrote that column, with three-megapixel cameras now available from all the major vendors. Resolution has ceased to be an issue for most consumer uses; prints from these cameras have great detail, even at 8 x 10 inches, and vanishingly few consumers ever make larger prints.

The cameras still have some drawbacks. The biggest quality issue is dynamic range; compared with film, it is more difficult to capture detail, in both shadows and highlights, with a digital camera. Usability is improving, but the user interfaces are still too cumbersome. The delay between pushing the button and capturing the picture is getting shorter, but it is still annoyingly long on most cameras. And the price premium remains substantial: a $300 film camera is superior in most respects to a $999 digital camera. If you want interchangeable lenses, the premium skyrockets.

I believe there is little doubt, however, that digital cameras will continue to improve, and that it won't be long before they become attractive to mainstream consumers. The tougher challenges lie in everything that happens after the pictures are taken, and that is where today's PC-centric solutions fall woefully short. Storing, printing, and sharing digital pictures offer great potential benefit but are just too cumbersome today for most consumers to accept. To make effective use of a digital camera, a consumer must master an assortment of PC applications, which vary in design quality from pathetic to barely acceptable, and must also learn to navigate several Web sites and endure tedious uploads to access printing and sharing services.

In the new world of digital photography, pictures are collections of bits, not arrangements of silver halide molecules. This is the first fundamental shift in photography since its invention -- and certainly the biggest change since George Eastman introduced roll film 120 years ago. Just as digital representation is now shaking up the music industry, it will also lead to a reshaping of the photography industry. The issues are quite different than for music, however, since most photos are taken by the user, whereas most music is purchased from a recording company. Sharing photos will become much easier, especially over a distance. Shoeboxes will be replaced by disk drives. And preserving photos will become a matter of safeguarding bits, not physical media.

As digital photography becomes mainstream, an entirely new infrastructure will be required. Already we see Web-based printing services, countless photo-sharing sites, and a few digital picture frames. In time, there will be many more devices on which people will want to view their pictures. As add-ons to a complex PC-based solution, however, additional devices and services only add to the already daunting complexity.

Consumers need solutions that make digital photos as easy to work with as film pictures while retaining all the advantages of the digital world. Five or ten years from now, the vast majority of the photographic infrastructure will have been replaced with new approaches, new technologies, and new companies. Massive transitions such as this create tremendous opportunities to create new businesses -- and this is an opportunity I feel compelled to chase.