LG claims first LTE terminal modem chip design

Will brings us up to date on Long Term Evolution (LTE) development issues, including its role as a catalyst for new modem (baseband) designs.

Announced in early December, LG Electronics claims to have developed the world's first Long Term Evolution (LTE) terminal modem chip. Developed at the company's Mobile Communication Technology Research Lab in Anyang, Korea, the device is said to have achieved wireless download speeds of 60 Mbps and upload speeds of 20 Mbps. The fastest HSDPA cellphones currently on the market have a maximum download speed of 7.6 Mbps.

Considering that the initial 3GPP (Release 8) LTE specification was frozen only about three weeks ago, and final approval is not expected before March 2009, we doubt that the chip is even close to being productized. Besides,  it requires substantial backward compatibility with GSM/GPRS/UMTS and/or CDMA 1xEV-DO to be a real cell phone product. No information has been released on process geometry or power consumption, so it's unclear how far LG is from having a viable product. And when the chip is ready for production will LG cease buying advanced basebands from its current main 3G supplier, Qualcomm?

 A slap at Qualcomm?

LG claims that through its own LTE development it has secured 300 (applicable?) patents. According to a company spokesman, LG “will see a dramatic decrease in its license burden.” LG's spokesman also says that, “since it now has the technology, it has the upper hand to greatly lower LTE license costs through negotiation. And if it supplies chips to other firms in the course of commercialization, it can secure additional license cost reduction.”

Multicore DSPs for LTE

With the additional processing demands that LTE brings over that of UMTS, we are hearing of new modem (baseband) designs that will employ between six and eight very small DSP cores. Tensilica's announced licensees in Japan for multicore implementation include Fujitsu, NEC, and Panasonic. We suspect that the Japanese semiconductor houses, long noted for strong inter-company relationships, have an overall grand architecture in mind. We have also heard rumors of a six-core baseband architecture underway in the West, due to be announced this summer.

With several tiny (45-nm) programmable cores, each with a dedicated set of tasks and algorithms, it becomes possible to employ different clock speeds and individual power control for each core and so to generate additional computing horsepower at lower overall power consumption. Such a feat is impossible with a single-DSP approach...that would require a higher clock speed, resulting in greater power consumption. In a multicore architecture, a dedicated FFT engine (for OFDM) could clock at a higher speed than a core dedicated to, say, channel estimation. We'll keep you informed as we get more information.

Freescale enters netbook chip market

Following on the heels of Qualcomm's announced netbook market entry, Freescale has announced its entry to that market with its i.MX515 application processor chip. The i.MX515 is based on the ARM Cortex A8 architecture, but with some differences from other ARM V7 implementations (like Qualcomm's Snapdragon and TI's OMAP 3). The Freescale product supports both mobile DDR2 and far-cheaper DDR2 memory (at one-fifth the cost of mobile DDR2). The company claims that a complete bill of materials, including the i.MX515, associated chips (PMU, audio codec, accelerometer, and touch sensor), and Linux can be as little as $20.00, compared to about $60.00 for an Intel Atom-based XP solution...leading to the possibility of a $200 netbook with much longer battery life.

Massively parallel processor chip companies: Fewer in number

With the recent demise of Ambric, Inc., we felt that it was time to review the status of companies working on massively parallel processor chips. Although Ambric's technology appeared to be solid and was gaining market momentum, in November Ambric abruptly announced that it was shutting down its operations and would be trying to sell the company and/or its IP. The company's VC backers declined to invest at the next level, presumably feeling that under current market conditions their exit strategy (usually an IPO) would be a long way off.

Table 1 presents our understanding of the current status of the most recent crop of massively parallel processor chip companies. If any of our readers have more recent information, I'd love to hear from you.


Table 1: Massively Parallel Processor Chip Companies


Wireless infrastructure holding up well

Shipments of traditional DSP chips were down again sharply in November (results released 12/31/08). Revenue shipments were down a whopping 33 percent from October and down almost 49 percent from November 2007. Shipments to all regions were down over 30 percent, including Asia Pacific (Taiwan, China, Korea, and Singapore). All applications were down substantially, with the exception of wireless infrastructure, which was down only a modest 3.4 percent.

Cell phone processors, consisting of both DSP basebands and application processors, are reported in two major categories by WSTS (cell phone DSPs and cell phone SoCs). Combined, their November revenue shipments were down only 12 percent from the previous month.

Overall DSP shipments appear to be headed for a 2008 revenue drop of 14 percent from 2007. To put things in perspective, the DSP drop in 2001 after the “dot-com” bubble burst was a much worse 24 percent. So don't you feel better now?

As always, I invite your comments.

Will Strauss

President and Principal Analyst

Forward Concepts    wis@fwdconcepts.com