Broadcom Chipset Could Boost US 3G Adoption by 2009

Just as in the microcomputer industry, the key to widespread dissemination of a design is mass production of its underlying technology. Standard chipsets are what make varieties of motherboards possible; and in the mobile phone realm, where form factors are much smaller, miniaturization demands that more basic features get crammed onto a single chip.

Today, Broadcom announced it appears to have overcome the myriad of timing issues associated with such amalgamation, and has begun sampling a system-on-a-chip (SoC) that combines all the underlying communication features phone manufacturers demand along with HSDPA/HSUPA high-speed data transfer.

The buildout list is somewhat staggering. If all Broadcom's plans pan out as planned, it means the company will have beaten rivals Qualcomm, STMicroelectronics, and Motorola to the mother lode: a true 3G baseband processor that adopts high-speed data standards much of the world has already embraced, and that's as easy for a manufacturer to implement as any other 65 nm CMOS.

Along with EDGE and WCDMA support, Broadcom's new BCM21551 unveiled today will include an HSPA modem capable of 7.2 Mbps of HSDPA download throughput, and 5.8 Mbps on the upload side. Not all networks worldwide actually display this speed just yet, but when they do, the bandwidth will be there. At the same time, the SoC maintains the engine for a 5 Mp camera and a 30 fps decoder for H.263 and H.264, MPEG-4, and WMV9 video at VGA resolution.

Getting these features, along with the now ordinary things like USB and BlueTooth support and RF/FM radio tuning, onto one chip is not just a feat of miniaturization. It's an achievement of static timing analysis, as indicated in this 2003 CommsDesign white paper that studied the problem in depth. All the different standards have already determined their own digital clock speeds, so to make them all work together on a single chip, something has to account for their synchronization.

Simply adding new clock tickers for this purpose can introduce a problem called phase error, a rough demonstration of which you can see most any day a TV reporter in New York conducts a live interview with a subject in Baghdad: When one person gets through speaking, there's a huge gap; and when one interrupts the other, that other person doesn't stop talking for a few seconds, after which the one who's interrupted has already given up trying to interrupt.

In the digital communications field, this self-multiplying pattern of overlaps followed by gaps followed by overlaps, is called jitter. Overcoming it is typically a nightmare unto itself. But in Broadcom's case, it was a nightmare double-feature: As its engineers were wrestling with jitters, its lawyers were feeling the jitters from Qualcomm, which kept up a brutal intellectual property fight against Broadcom and others up until last February. At that time, a federal judge ruled Samsung and Panasonic could continue to use Broadcom's existing WCDMA SoC chips in their handsets without fear of reprisal from Qualcomm.

With the clouds lifted from over the heads of two of Broadcom's most important customers, Broadcom announced last April at the 3GSM World Congress in Barcelona that it was going ahead with its plans to produce a 65 nm CMOS with 3G support. The industry took that proclamation very seriously, with Nokia responding by offloading its chipset business to STMicroelectronics.

How would Qualcomm respond? Back in September 2005, Qualcomm's own senior vice president for marketing Jeffrey Belk wrote a white paper called "Why MAX?" (PDF available here). There, he called the WiMAX standard "not a pretty picture," among other choice words. "In the system simulations we are doing on Mobile WiMAX," Belk wrote at that time, "it's not clear whether WiMAX has ANY advantages over the evolutionary path for CDMA2000 or WCDMA/HSDPA WWAN."

But in the wake of Broadcom's proclamation, Qualcomm announced last April it was purchasing the SoC production assets of TeleCIS. That company had been working to build its own 3G SoC with EV-DO and WiMAX, and Qualcomm said it would continue that project, despite its earlier quarrels with that technology. Perhaps Broadcom agreed with Belk after all.

As Broadcom stated this morning, putting all these backbone technologies into one chip enables manufacturers to concentrate on building handsets with more open features, such as supporting Symbian, Linux, or Windows Mobile. A Broadcom marketing manager told EE Times Europe that, as long as his company's roadmap continues undeterred, "We anticipate that HSUPA phones targeting the mid-tier price market using the part will be on sale during the first half of 2009."