ConnectorSupplier.com - The Top 10 Industry Trends

The Top 10 Industry Trends

Systems-in-Package

Moore’s Law
Gordon Moore postulated years ago that silicon ICs would double in transistor count (i.e. raw performance) every 18 months. That comment was etched in stone, as year after year Moore’s prediction came true and his status was elevated to the “Oracle of Silicon Valley.”

However, all things are subject to change, soft landings, or brick walls. Semiconductor technology is no exception. Things cannot get smaller forever. At some point, feature sizes reach molecular levels. We are not there yet, but we are getting closer. Current trends indicate that transistor counts have already slowed down as semiconductor feature sizes reach nanometer levels, with existing transistor populations well into the millions.

Photo courtesy of Evaluation Engineering

Alternative strategies are being developed at 65 nanometer node and below. These include Deep UV and X-ray photolithography to resolve nanometer feature sizes, as well as multi-core processing, where more than one CPU is diffused onto a single chip. Some predict that this latter trend will rise from the current dual- and quad-core Intel and AMD CPUs, to 16 or 32 cores within a decade. This trend will provide multiple independent processing units on a single logic chip, thereby increasing chip power via parallel processing, not sheer transistor count or speed.

The focal point of this development activity is in computer and related industries where embedded computing power is found in many products. The leader in this area is Intel Corp., with significant developments occurring at AMD, IBM, and other chip companies, ensuring that the multi-core phenomenon will become more widespread in the future.

While at a different level than the stamp-and-form/injection mold technologies used in the connector industry, IC technology and packaging developments have a profound impact on connector applications and demand.

MCM, SiP and SoC
Multi-chip modules (MCM) were reported to be the up-and-coming thing back in the `90s. The concept—taking off from the slowing Moore’s Law proposition—was that multiple chips, packaged together in a small module, would overcome the limitations of single-chip packages, operate at lightning speed, and provide a true system-in-package approach.

MCMs, more or less, flopped for one major reason: None of the major semiconductor companies supported it—except in their own proprietary multichip packages. Independent MCM startups were trying to create a paradigm shift without the support of major IC players; most were fabless and did not have native applications to promote.

Finally, industry began to pick up on the idea—mainly from companies who actually had a product to build. Thus, the System-in-Package (SiP) and System-on-Chip (SoC) concepts were developed.

ST Micro Camera-on-Chip

SiP and SoC mean what they say, although the term “system” is used rather loosely in the current vernacular. It more often means “subsystem” in-package or on-chip, rather than the complete system. Otherwise, you might be holding a TSOP (thin small-outline package) or BGA (ball grid array) to your ear, not a cell phone.

SiP is two or more die in one package, allowing partitioning into RF/digital/memory blocks.

SoC integrates all functions on one chip.

SiP/SoC is used (here) generically to describe a wide array of highly integrated planar, 3D, mixed technology, or MEMS semiconductor systems-in-package, which may drive the future electronics marketplace.

Moore’s Law is moving to 45nm, and small feature sizes will enhance SiP/SoC possibilities because geometries will allow five to 10 times more functions to be packed into a single chip.

SiP/SoC is now happening at the subsystem level in cell phones, digital cameras, GPS devices, and the like. Many of these products did not previously exist. What holds SiP/SoC back from universality is the cost of integration—it often takes more than one company to produce the chips, and the marketplace is oriented for mix or match designs composed of many different chips.

Also, the market has changed with its current high levels of digital convergence:

The industry is 10 years further down the CMOS learning curve.
Packaging has advanced with Via organic substrates and integrated passives.
High-volume miniaturized applications have somewhat predictable roadmaps.
SiP development cycles and investment costs are shorter than a SoC.
SiP or SoC can make a lot of technical and economic sense, depending on volume.
Volume and product life cycles are key, due to the high cost of development.
SiP/SoC can be produced by one company as the semiconductor industry has expanded.
Fabless IC companies are emerging to showcase their creativity.
Big companies, such as Intel, Qualcomm, and Freescale, can leverage their own killer applications.
Wireless mixed-signal applications tend to promote SiP and SoC technologies.

SiP/SoC Examples

Jazz Semiconductor 0.18 micron silicon radio platforms for cell phones and networks, single chip TSOP or other packages.

Amkor Technology SiP, organic MCM process, integrating ICs, passives, connectors, and other devices.

Ziptronix 3D stacked Si for 3D option to linear SiP.

IC Forecast: A Robust Industry? Or, Slowing and Maturing?
As we enter 2008, concerns are being expressed about the global economy and its impact on semiconductor demand. Certainly the IC market has become much more dependent on consumer spending, which is now exceeding 40 percent of the total demand. Gartner’s November 2007 forecast is shown below:

Gartner Worldwide Semiconductor Forecast 2002-2011
(November 2007)

Other industry forecasts bracket the above numbers, ranging from 3.9 to 8.1 percent projected growth (in dollars), higher in units. There remains the possibility of negative growth if a recession occurs
2007 shipments show how important consumer-driven market segments are. These segments would be primary candidates for SiP and SoC because of their high unit volume.
The ever-increasing need to reduce cost plays toward higher very-large-scale integration (VLSI) and ultra-large-scale (ULSI) levels
SoC is a potentially powerful tool for large multinational OEMs because such developments can potentially checkmate smaller rivals in an increasingly global, competitive marketplace.
SoC is the ultimate in vertical integration, providing maximum financial leverage.

Semiconductor Market Segments 2007

Source: Semiconductor International Webcast December 2007

Forecast Summary

Semiconductor sales have grown from $141 billion in 2002 to $257 billion in 2007—a 12 percent CAGR.
2007 was a correction year, primarily because of declining ASPs (average selling price).
2008 and 2009 should produce respectable growth, with unit volume up from 13 percent to 15 percent.
A somewhat bearish (monetary) forecast for 2006 to 2011 results in a modest 5.93 percent dollar forecast.
There are concerns that economic events could change this, in the short term, to a negative forecast.
SiP, SoC, and multi-core developments will continue because innovation and competitive positioning are core motivators for the semiconductor industry.
The incredible costs of facilitation will continue to restrict the number of major IC players.

Impact of SiP/SoC on Connectors

SiP/SoC will impact connector applications and use, both negatively and positively.
SiP/SoC are potentially a game-changer in terms of their impact on subsystem and system design.
They are logical evolutionary technology developments based on increasing levels of integration.
By combining chip functions, they have the potential to reduce a “system” to a chip.
Over the years, connectors have experienced successive levels of integration, and prospered.
Increasing levels of system integration resulted in new, high-demand electronic products that utilized more connectors, from IC sockets to I/O connectors and cables.

Examples where SiP or SoC might impact connectors include:

High-volume consumer products: cell phones, other handhelds, notebook and desktop all-in-one PCs, wireless networks and automotive electronics.
In PCs, there has been a slight reduction in connector count, mainly through standardization and removal of legacy I/O, but overall volume is up double-digits.
IC-induced miniaturization of outboard components, but with connectors there is a limit.
IC socketing: BGA and LGA socketing of large expensive packages with high I/O counts.
PCB connectors: Fewer per system, less outboard system bus and memory requirements.
Flex Circuitry: More high-density flex connector requirements, chip-on-flex, display I/O, etc.
I/O connectors: Fewer per system, as wireless/mixed-signal applications go SiP/SoC.
So far, the “fewer” scenario hasn’t happened, but it will develop with wireless USB.
Future PC board designs could follow ICs into much higher levels of integration, placing design pressure on conventional components while eliminating others.