Design Autopsy: Dissecting BoMs To Defuse Costs

These days, shrinking a product's price tag usually involves cutting the cost of service-related elements, the performance level, or the number of bells and whistles. But these may not be the only options, nor the most viable. Too often, we neglect to scrutinize the bill of materials (BoM). We might assume those costs are already minimal, so there's hardly room for improvement. That could turn into a very expensive mistake, though.

To demonstrate this fundamental precept, let's use a feature-laden cell phone as an example. If it's possible to reduce the cost of goods for this device, it's very likely that the same principles can reap similar rewards for a broad array of designs. This article examines the factors driving the rise of new features in mobile phones, analyzes different approaches to phone design and part selection, and suggests ways to keep costs under control.

Not long ago, a mobile phone was simply a mobile phone. It performed voice communications and a few related tasks, such as displaying caller ID information on a monochrome screen. Today, however, a mobile phone can be a digital still camera, a personal digital assistant (PDA), or a handheld video game system.

Additionally, while mobile phones were once simple devices, almost spartan in their features, they now seem out of date if they don't come with polyphonic ring tones, a built-in camera, and at least one color display—if not two. In the near future, phones that don't sport Bluetooth connectivity, active-matrix thin-film-transistor liquid-crystal displays (AM-TFT LCDs), and 3D graphics accelerators may be relegated to antique status.

Mobile phones are undergoing a feature frenzy, enabled by the evolution of standards and technology and driven by wireless carriers hoping to lure subscribers with flashy new attractions. This feature boom is bringing a veritable explosion in sales (see "New Features Fire Up Mobile-Phone Sales," p. 56).

But this proliferation of features comes at a price. Most new elements require additional transistors, components, and board space, factors that drove up the BoM for mobile phones in 2003. With buyers sensitive to price and with wireless service carriers often selling their handsets at a loss, preventing the BoM from ballooning is key to success in the mobile-phone world.

To remain cost-competitive while simultaneously satisfying demand for exciting new features, makers of mobile phones must reduce the BoM for some of the less glamorous, but no less important, segments of mobile-phone designs, such as the memory and radio-frequency/intermediate-frequency (RF/IF) subsystems. Based on the teardowns and comparative analyses of more than two dozen mobile phones, iSuppli's Teardown Analysis Service identified a number of opportunities for cutting costs in these areas.

THE COST OF INNOVATION Phone makers know that adding new features pumps up costs. "Features are the main drivers of material costs in mobile phones," says Andrew Rassweiler, who manages the Teardown Analysis Service. "In the phone teardowns we conducted, we've seen products with varying features ranging in manufacturing cost from as little as $40 to almost $200. The addition of new features not only boosts electronic content but increases the overall complexity of mobile phones, requiring more memory and even more programming, assembly, and testing."

The flood of new features in 2003 saw the worldwide factory average selling price (ASP) for mobile phones rise to $160, up from $157 in 2002. This is remarkable, given that factory ASPs declined during every year at least since 1995—even during the boom years of 1999 and 2000, when some component prices soared due to shortages. The main driving force behind the continual decline in phone factory ASPs is none other than Moore's Law, the basic semiconductor dynamic that describes how the cost of semiconductors falls, while functionality rises.

This puts the wireless carriers in a predicament. New features can cause factory ASPs to rise, but simply boosting end-product prices isn't an option in the highly competitive mobile market. In fact, most phones are sold at a discount, and some are even given away for free, to lure buyers.

"There is significant price elasticity in the mobile phone market," says Dale Ford, vice president of market intelligence services for iSuppli. "Time and time again, it has been shown that phones have to be at the right price points to be successful." Also, wireless carriers subsidize the cost of the phones. The more the carriers subsidize, the thinner their profit margins.

Thus, mobile-phone designs are continually affected by two opposing forces: the costly surge of new features and the ebb tide of Moore's Law, which brings greater functionality at lower expense. Yet with the flood of new features outpacing the inexorable progress of Moore's Law, mobile-phone makers will have to take advantage of technological and design innovations to reduce their BoMs and keep costs in check.

DROPPING THE BoM But where should designers cut to reduce their mobile-phone BoMs? First, let's pinpoint where the costs lay. Figure 1 shows a cutaway view of a typical mobile phone. It illustrates the various subsystems on its pc board, including two user interfaces, baseband circuitry, a trio of battery/power-management sections, and the memory and RF/IF portions. Obviously, the display isn't shown here.

Figure 2 shows the BoM broken down in greater detail by subsystem. Clearly, the display and baseband portions run neck-and-neck for the most expensive elements of the phones, followed by user interface, memory, RF/IF, mechanical/electromechanical, battery, accessories, and power amplifier.

While new, more integrated ICs can lessen baseband-subsystem cost, adding transistors to support new features tends to counteract the reduction in expense, according to Scott Smyser, iSuppli's senior analyst for frequency control, RF, and wireless. The arrival of 3D graphics processors and other devices intended to offload tasks from the system microprocessor probably will prevent much erosion in overall baseband costs, he notes.

This is also true for the display and user interface, which will likely see normal cost reductions largely reversed by the use of more advanced screens and electronics. This leaves only a few areas where costs can be cut significantly in mobile phones: memory, RF/IF, and the power amplifier, Smyser says.

REDUCING MEMORY COSTS Increased functionality has triggered a dramatic rise in the amount of memory consumed in mobile phones, as well as a proliferation in memory types employed, according to Betsy Van Hees, iSuppli's principal analyst for memory.

For a basic phone with a monochrome display, 2 to 4 Mbytes of memory is adequate for code storage. With a color screen and a few additional features, the required memory density climbs to the 8-Mbyte range. For color-screen phones with cameras, the density jumps to 16 Mbytes. And with smart phones, density reaches the 64-Mbyte level.

Besides code storage, mobile phones also require separate memories for data storage and, in more advanced models, buffering. To implement these three functions, phone designers face a dizzying set of choices of densities, memory types, and packaging options. Memory types found in mobile phones include NOR flash, NAND flash, SRAM, pseudo-SRAM (PSRAM), and SDRAM.

The optimal memory combination for today's richly featured mobile phones is NOR flash for code storage, NAND flash for data storage, and PSRAM for buffer memory, says Van Hees. The teardown analysis reveals that using multilevel-cell (MLC) flash memory can significantly reduce flash-memory costs.

Recent teardown analysis of two phones revealed nearly identical memory configurations: 16 Mbytes of NOR flash, plus half a megabyte of SRAM. But one phone used an MLC-type NOR part, which costs $7.00 in quantity, compared to $10.00 for the non-MLC flash part in the other phone.

Because SRAM costs the same in both phones, the product using the MLC flash memory achieved a $3.00 savings from the total phone BoM. With an estimated 575,000 of these phones expected to ship this year, the product's maker achieved a $1.73 million cost savings on this model alone.

In general, MLC flash memory delivers a 20% to 30% cost savings compared to equivalent-density single-cell flash, says Van Hees. Intel Corp. (www.intel.com) offers MLC NOR flash under the brand name StrataFlash, Sharp Corp. (www.sharp.com) sells compatible parts, and Advanced Micro Devices Inc. (www.amd.com) has the MirrorBit line.

CUTTING TRANSCEIVER COSTS Further opportunities for cost reduction reside in the mobile phone's RF/IF subsystem. One obvious route to cutting costs is to eliminate the IF segment of the subsystem entirely and move to a Zero-IF (ZIF) architecture. This cuts an estimated 75 cents to $1 from a mobile-phone BoM by eliminating the need for an IF surface-acoustic-wave (SAW) filter and its related passive components. ZIF also reduces the overall complexity and size of mobile-phone electronics.

ZIF already has thoroughly penetrated the market for digital cellular systems that use the Global System for Mobile (GSM) communications, largely in Europe and Asia. GSM uses narrowband time-division multiple access (TDMA), which permits eight simultaneous calls on the same radio frequency.

However, phones using code-division multiple access (CDMA) have been slower to adopt ZIF due to the complexity of their transceiver systems that transmit and receive signals simultaneously, says Smyser. CDMA, which uses spread-spectrum techniques, doesn't assign a specific frequency to each user. Instead, each channel uses the full available spectrum.

In 2002, Qualcomm Inc. (www.qualcomm.com) released the first CDMA ZIF design, and penetration of the technology reached 35% in 2003, iSuppli estimates. This year, penetration will rise to 60%. Smyser advises that CDMA phone designers should move quickly to adopt the ZIF architecture to take advantage of its cost, space, and complexity savings.

The teardown analysis also revealed additional savings via an RF transceiver IC that integrates a voltage-controlled oscillator (VCO) or phase-locked loop (PLL). A good number of the torn-down phones used standalone VCOs for frequency synthesis with the RF transceiver. Integrating the VCO into the transceiver can cut the BoM by 75 cents.

See the table for potential cost savings from using an integrated transceiver by comparing two phones with similar features. Phone A uses a standalone VCO, while Phone B uses an integrated transceiver. Although Phone A achieves lower cost in its front-end switching and filtering and power amplifier, the BoM for its RF/IF subsystem is 53 cents more than that of Phone B due to the cost savings on the integrated transceiver.

Another BoM-cutting opportunity for mobile phones is to use a front-end module (FEM). A FEM combines an antenna switch module, a pin photodiode, two RF SAW filters, and a few passives, all presently standalone parts on most mobile phones. Though FEMs offer significant advantages in space savings and manufacturing cost, they're more expensive than their alternative standalone solutions. A FEM runs about $1.40 in quantity, compared to $1.30 for a standalone approach.

But iSuppli expects this cost differential to disappear by the end of the year. As FEM prices decline, their popularity with designers rises. In 2003, only one of the 20 mobile phones torn down by iSuppli used FEMs. In 2004, iSuppli has torn down five phones, three of which contained FEMs.

THE CHALLENGE: INNOVATE WHILE CUTTING COSTS The new features that electrified the mobile phone market in 2003 represent only the first wave of a new era of innovation in cell phones, as wireless carriers intensify their quest to find the next big thing (see "Mobile Phones Promise A Slew Of Innovations," p. 54). But with each new feature, the mobile-phone BoM blows up a little more. As a result, mobile-phone designers continually must find and take advantage of new technical innovations that can cut the complexity and cost of their products.