There is no question that wireless connectivity dominates current thinking on the utility of new-generation portables. Third-generation (3G) cellular handsets will offer Internet connectivity in addition to voice messaging and telephone capabilities-even the ability to transmit data and pictures. Bluetooth promises wireless connectivity for cell phones and PDAs and may give portable computers wireless Internet access in hotel lobbies, convention centers and at airport gates. Portables using IEEE802.11 protocols in an office LAN may also be found frequency-hopping in the home using HomeRF protocols.
However, the question remains whether the RF transceivers that offer this wireless connectivity will be multichip modules packaged on dimensionally stable multilayer ceramics-as they are with current-generation cell phones-or single-chip devices mounted on low-cost printed-circuit-board materials. With Bluetooth, in particular, a number of semiconductor manufacturers are aggressively pursuing one- and two-chip solutions in CMOS or BiCMOS. CMOS integration is perceived as the key to the $5 solution everyone is hoping for, but some say the focus on silicon obscures the technology issues associated with putting RF transceivers on phenolic substrates. Putting aside the reliability issues-PalmPilots do not have to function in the same physical environments as aircraft navigation systems-microwave industry veterans will ask whether FR-4 materials have enough dimension stability to hold the interconnect tight trace lengths associated with 2.4-GHz transmission lines.
While they never go head-to-head, contributors to this week's Signals section address the packaging issues attached to new-generation portables. For example, Matthew Phillips of Cambridge Silicon Radio (CSR; Cambridge, England) and Steve Brown of Silicon Wave Inc. (San Diego) argue that carefully crafted Bluetooth transceivers will not only eliminate the need for the dozens of passive components ordinarily associated with RF transceivers, but also will make the transceivers relatively insensitive to the kinds of substrate materials used.
Indeed, current writings on new-generation RF transceivers indicate that extremely precise tuning could help implement some of those goals. In the architecture of a superheterodyne receiver, the modulating signal (voice or data) is extracted from a captured RF signal by mixing that signal with an out-of-phase replica of the RF carrier.
Ideally, the replica-a synthesized frequency-should be exactly 180 degrees out of phase with the RF carrier in order to extract the carrier. But sloppiness in the tuning of the RF signal and imprecision in the voltage-controlled oscillator that produces the synthesized replica will produce a variety of spurs or unwanted images of the modulation signal that will need to be suppressed or filtered. The additional filtering requirement often results in dozens of little RC "tank circuits," which must be crafted with precise geometries and mounted on nonabsorbent substrate materials. That packaging requirement prevents the captured signal from being totally absorbed in its own tuning circuitry.
|As head of the International Wireless Packaging Consortium, Don Brown is driving the development of EDA tools and databases that will support the rising time-to-market concerns of RF designers.
The new-generation frequency synthesizers, however, are much more adept at replicating the RF carrier in the first-stage mixing process. Integer-N and fractional-N synthesizers-implemented in digital CMOS-can adaptively account for imprecision in the tuning process. That is, it can adaptively track the tuned signal and ensure a precise phase relationship with the carrier it is trying to null out. That, precision synthesis, single-chip radio advocates will argue, not only eliminates the passive component count but also makes the transceivers relatively insensitive to their substrates.
Cambridge Silicon Radio is pushing a low-cost, one-chip solution that includes both a radio transceiver and a control processor, essentially a data formatter, implemented entirely in CMOS. But Silicon Wave argues that the integration of a radio and control processor may result in design compromises in the architecture of each and delays in time-to-market as the more complex part is tuned and tweaked. Silicon Wave advocates a two-chip solution, with the radio transceiver implemented in BiCMOS.
Both manufacturers insist that readily available FR-4 materials can be used to implement Blue, though the "fine print" in CSR's article reminds readers that a matching network-or ballun-is required to ensure an impedance match between transceiver terminals on their chip and the detachable antenna. Impedance matching, which is extremely important in RF circuits, ensures that signals will not be attenuated along the path from chip to antenna. The company recommends a printed-circuit board implementation, in which the impedance-matching network-representing fractions of the wavelength produced by the 2.4-GHz carrier-is etched into the board's metal traces.
National Semiconductor Corp. (Santa Clara, Calif. ), like CSR and Silicon Wave, is working on a single-chip Bluetooth radio and is looking to integrate the stack and control processors in successive generations. Bob Swann's contribution, though, comes at the packaging issue from a slightly different angle. He examines the use of chip-scale packaging and laminates for RF circuits, an option that offers further IC size reductions for devices packaged in SOT surface mount.
The contribution from DuPont's (Research Triangle Park, N.C.) Sam Horowitz explains the design benefits of using low-temperature co-fired ceramics (LTCCs). "Sandwiching passives and RF tank circuits with semiconductor devices between layers of ceramic or glass-and replacing chip resistors and capacitors with thick films in the sandwich-can result in considerable savings in size, space and component count," he writes. What's more, low temperature coefficients of expansion mean the LTCCs are dimensionally stable under a wide variety of environmental conditions, ensuring the integrity of the RF circuits they carry. It should come as no surprise that this is the preferred packaging for RF modules in cellular handsets. Ericsson's Bluetooth module offering, in fact, is an LTCC. It is somewhat more costly than an FR-4 PCB implementation, but Ericsson's module represents a stable reference design.
Bridging islands of automation
A variation of LTCC called Multi-Mix, where stable phenolics as well as ceramic and glass are used to package high-density RF circuits, is described by Richard Dec of Merrimac Industries Inc. (West Caldwell, N.J.). The technology is so well-suited for handsets that Ericsson invested in the company, buying 17.5 percent of Merrimac's stock last April.Meanwhile, of prime concern to consultant Don Brown, who leads the International Wireless Packaging Consortium (IWPC), is a shortage of published knowledge on the characteristics of packaging materials, substrates and components-especially as those are applied to RF transmitters and receivers. Even where EDA tools are able to encapsulate knowledge of materials and models of components, the tools exist as "islands of automation," with few ways of moving design data from one tool to another in the design process without manual reentry.
The IWPC is developing a tool set and database standard that would enable RF engineers to access reliable design information quickly-ideally from the Internet. Brown identifies specifications of special interest to RF designers-including operating frequency, bandpass design and tolerance requirements, transmitted and received RF power., thermal stability of material properties, and-above all else-cost. Brown's 1998 survey of 250 engineers and manufacturers revealed not just a shortage of details in such areas but also a lack of standards among RF material and component suppliers.
A standardized database and easy access to design information, Brown believes, would speed time-to-market for millions of new cell phone designs, pagers, GPS devices, pico cell links, wireless PDAs and automotive RF products. By holding close to a dozen workshops each year and developing an Electronic Product Design System tool set, the IWPC is showing progress toward that goal.