The answer isn't always a single-chip SoC for lowest cost

An Infineon Technologies entry into the increasingly hot ultra-low-cost handset market this week illustrates an interesting point. Even in very cost-constrained systems, it’s not always the single-chip solution that makes the most sense. In this case, the object lesson comes from a reference design targeted to produce a GSM/Voice-over-WiFi dual-mode handset selling for under $100.

The elegant approach to this design would be to fire up a large design team and integrate RF front ends for both GSM and 802.11, two direct-conversion transceivers, a dual-mode baseband processor and an application processor on one 45 nm die. And that may happen some day, but that’s not the way Infineon chose to do it.

Instead, the company turned to massive production volumes, rather than massive integration, to slash costs. The company already has its X-Gold 101 single chip for ultra-low-cost handsets, of which they have shipped about 50 million units. And Infineon has a relationship with Atheros Communications that gave them access to the low-end single-chip AR6101 802.11 a/g hardware and software. The designers calculated that given the huge volumes in which these chips ship, they could not beat the bill-of-materials cost of the two, even with their full complement of supporting chips and passives (about 50 components each) with a new single-chip integration in today’s technology. "Even in the intermediate term I don’t see this becoming a single chip," says Infineon program manager Alain Fogel.

But the reference design still required more than just putting the two chips with their respective other components and antennae, down on a board. As it happened, both chips have low enough standby power that power management isn’t a huge issue, even with both chips in standby mode. Infineon estimates the battery life with a standard battery of 250 hours, and talk time of up to 6 hours.

RF interference was a problem. Of course GSM and WiFi operate near each other all the time, but not in the same package. Infineon had to add a pair of bandpass filters to the WiFi design and a low-pass filter to the GSM design to reach adequate noise margins, particularly to get the GSM radio to 3GPP levels.

Software is another issue. If this phone is going into developing countries, users will need to have an easy way of gaining access to WiFi hotspots, or they won’t spend the extra money for a dual-mode phone. That means the phone has to come with clients for GoogleTalk, Yahoo, and the rapidly-spreading network of FON communities. And, Fogel says, the handset will probably want a simple browser so that users can become clients on other sorts of WiFi VoIP services.

The browser raises a problem, however. Fogel says the reference design has enough memory and computing power for one, but browsing on the tiny screen of the ultra-low-cost handset is a non-starter. Adding a larger display would drive the selling price from about $100 to about $130. "And our studies indicate that this much of a price increase could mean a factor of six loss in volume," Fogel says.

If every dollar counts, sometimes existing, high-volume ASSPs can make more sense than an integrated design, even if they pull in substantial numbers of supporting components. It’s not obvious, but it is interesting.