CEVA in First Commercially Available LTE Handset

Last week, Samsung Mobile Launched the Samsung Craft, the world’s first commercially available 4G LTE handset.  Specifically designed for the MetroPCS network, and initially available in the Las Vegas, Nevada area, the Craft is a multimode handset, with LTE data and CDMA voice and texting capability.  The LTE modem is Samsung’s own, based on CEVA DSP core technology, while the CDMA modem is from Qualcomm.  It is a bit curious that Samsung chose not to employ the readily-available Qualcomm LTE modem.  Perhaps Samsung likes the licensing terms better with its own LTE solution…or maybe they believe it is a better performer (after all, Samsung is getting up to 80 Mbps with that modem in Sweden).

CEVA also Provides SoC Route for TI DSP Chip Users

Earlier this month, CEVA, Inc., the leading licensor of DSP cores, introduced a new VLSI+SIMD DSP that provides users of Texas Instruments C64X family of discrete DSP chips a capability of migrating to an SoC solution, while preserving much of their existing C64X C-code.  The new CEVA-X1643 is said to be highly energy-efficient, and capable of operating at 1GHz using standard 40nm CMOS process technology.  Although applicable for a broad range of applications, the full support for TI intrinsic functions and data cache the X1643 will be appealing to existing TI customers who want to incorporate additional features into a single-chip solution specific to their product differentiation.  CEVA claims a Tier-1 wireless vendor has already licensed the new product.  My guess is that it is for a wireless base station application.

The X1643 also features an AXI bus, making it easy to join up with a licensed ARM RISC processor.  But, TI is not rolling over and is also offering a number of new combo C64X-ARM processor chips…but they are still discrete solutions, available off the shelf to anyone.

As we have pointed out before, SoC solutions (and FPGAs) make up almost 90% of the $30 billion DSP silicon revenue TAM (in everything from MP3 players and disk drive controllers to cellphone modems), with discrete DSPs accounting for the other 10%.  Not only that, just a half dozen or so chip suppliers offer discrete DSP chips (with TI and Analog Devices leading the pack), while about 100 chip suppliers offer SoC products that incorporate DSP as a necessary technology (with Qualcomm leading that pack).

More Leaks on Verizon’s iPhone

We speculated in our last newsletter that Intel was able to acquire Infineon’s cellular chip business on the cheap because of the looming CDMA version of the iPhone for Verizon.  More indications of the CDMA iPhone have been leaking out of China, with Taiwan-based DigiTimes reporting that a China-based analyst firm has stated that Qualcomm is replacing Infineon in future iPhones and another article states that “Pegatron Technology is expected to begin volume production of a CDMA-version iPhone 4 in November with total shipments likely to reach 3-4 million units in the first three months, and 10 million by mid-2011, according to industry sources.”  Of course, we can’t verify the accuracy of the “industry sources.”

Advice to Intel

Now that the company is acquiring Infineon’s GSM/EDGE/WCDMA/HSPA (and LTE) cellular chip business, Intel needs to expand on that base to incorporate CDMA capability; otherwise Qualcomm will be the only company that can supply chips for high-end handsets that serve the largest American market. Actually, Intel was once in the CDMA cellular chip market…after it acquired DSP Communications Inc. (in October, 1999), but their biggest customer was Kyocera…and Qualcomm subsequently bought the Kyocera handset operation in San Diego (in December, 1999), effectively killing that market segment for Intel.  Note the dates.  Wow, talk about bad timing!

So, Intel can now either invest a few hundred million dollars and at least two years developing its own CDMA capability or it can acquire the only other vendor of such chips: Via Telecom.  Although not well known, Via Telecom is shipping CDMA 1x chips in volume, primarily to handsets in China and India (and Sprint in the U.S.).  Its customers include Nokia, Pantech, Samsung and white-box handsets in China.  The company claims to also have CDMA 1xEV-DO Rev. A chips, but we haven’t seen them shipping in handsets, yet. But, Intel’s financial and marketing muscle could ensure that happening.

We know that CDMA will ultimately give way to LTE, but for at least the rest of this decade CDMA will be required as a legacy fall back on LTE networks from CDMA operators like Verizon, Sprint, and MetroPCS, as mentioned above.

Via Telecom is the San Diego subsidiary of Taiwan-based Via Technologies, Inc., the number-three X86 processor supplier. I suspect that Intel could offer enough money to get Via Technologies’ attention, but Via might also be open to negotiating better licensing terms for its X86 processors, too.

Evolved EDGE: A Technology Who’s Time has Come

The industry buzz is all about LTE as the next big thing.  However, much of the world would be happy simply to get better results over the installed (and still growing) EDGE network.  EDGE represents improved data rates over that of earlier GPRS, with average bit rates in the neighborhood of 200kbps…which is almost 5 times faster than dial-up Internet access.  But, it can be better. The 3GPPS Release 7 specifies EDGE evolution, which some refer to as Evolved EDGE or simply eEDGE.  eEDGE promises dramatic improvement in data rates, which can typically be about 500kbps, with a peak rate of 1Mbps (with a dual-carrier implementation). Thus, eEDGE can reach low-end DSL data rates (and end-to-end round trip latency of about 70ms compared to 160ms for EDGE), with only a software upgrade in base stations.  However, cellphones do require minor hardware changes (adding about $4.00 to the chip BOM).  The first commercial eEDGE network is said to be operational in Istanbul, Turkey (Turkcell).  Trial networks are also underway by Ericsson in China and India, and Research In Motion (RIM) has developed its own eEDGE ancillary chips.

With its eEDGE handset product on the way, RIM probably leads in the number of air interfaces supported by a single cellphone company.  It has handsets for GPRS/EDGE/WCDMA/HSPA/CDMA-1XEV-DO/iDEN and, soon, eEDGE.

Freescale Re-discovers AltiVec

Back when Motorola Semiconductor was supplying Apple with PowerPC-based G4 and G5 processors, the company supplied a significant SIMD processing capability (similar to Intel’s SSE…streaming SIMD extensions) called AltiVec.  Meritorious in concept, the vector and DSP capability that AltiVec instruction set provided found its way into IBM’s Cell BE processor and Wind River (now owned by Intel) also embraced AltiVec.  But, as the company separated from Motorola and became Freescale, AltiVec continued shipping into high-performance military and bio security and print imaging applications, but was relegated to catalog status and Apple changed over to Intel Pentiums.  Meanwhile, Freescale beefed up its PowerPC heritage to create the QorIQ (pronounced “core IQ”) family of communication processors which dominate the cellular base station RISC market for layer-2 processing.  Although some members of the QorIQ family incorporate Freescale’s e500 DSP core, Freescale re-discovered the merits of adding AltiVec to that processor family.

Today, Freescale made an announcement that it was adding an enhanced reincarnation of AltiVec technology into the QorIQ family, but with new logic and a dozen new instructions that assist compiler optimization, etc.  The company expects AltiVec to improve layer-2 base station processing by 30-50%, while providing IPV6 support and improved security capabilities.  As this is an architectural announcement, we can expect products added to the company’s QorIQ 5500 family portfolio early next year.

AT&T Femtocells Will Ultimately be “Free”

AT&T has been criticized for charging $150 for their residential femtocells (MicroCells); devices which can enhance AT&T’s otherwise patchy cellular coverage. Compared to Japan’s SoftBank, which provides femtocells to its cellular subscribers for free, this seems a bit high.  Industry sources indicate to me that the reason for the relatively high price for a device that is equally beneficial to the customer and to AT&T is that the devices are currently supply-constrained by component shortages.  It appears that when the supply shortage is relieved, AT&T may be providing their femtocell devices for free (OK, subsidized)…since they can enhance network coverage and improve customer satisfaction, leading to reduced churn and higher overall revenues.

Component Shortages are Real

Several handset vendors have recently cited component shortages that are constraining their revenue growth, including Motorola in the U.S.  The average handset consists of hundreds of items, from tiny resistors and capacitors to chips and displays.  The shortage of a single 5-cent component can prevent a $200 handset from shipping (times hundreds of thousands of units).

History proved this when Ericsson was in the cellphone business (before Sony-Ericsson) in the early part of the last decade (c.2002). Ericsson could not get adequate quantities of tantalum capacitors because of high global demand and ethical restrictions on purchasing from sources that could not guarantee that the raw material for tantalum (coltan) was not from the Congo.  You’ve probably heard of “conflict diamonds” from the Congo, but “conflict coltan” has also been very profitable for warring rebel groups there.  The impact on Ericsson was such that its very survival was in question.  Some feel that this single event led to the merger of the handset operation with Sony.  Tantalum is not the constraint today, but hopefully the current component shortages will soon diminish.