Laminate structures and low-temperature co-fired ceramic technologies allow custom RF module designs for SiP architectures.By Michael Gaynor
The design and development of RF circuits is an ever evolving work of art. From elementary circuit design to transmitter and receiver functions, RF power amplifiers for base stations and full architecture RF circuit design for satellite phones, one size doesn't fit all.
LTCC and Laminate Modules
One area of designs includes integrating RF functionality on laminate substrates and LTCC. This article examines the advantages and disadvantages of these two substrates with respect to RF module design. A general design procedure will be presented with some module design examples.
The High-level Overview
A good approach is to analyze the RF module design as assess how to design the RF functionality into the module. The RF design process starts with defining the end-user's module requirements. An analysis is performed to develop module solutions to meet the desired size and RF performance.
A cost analysis of the partitioning of laminate and LTCC is reviewed. Usually each request reviews an all-laminate module, an all LTCC module, and a laminate module with some RF functions designed into LTCC. Modules that are entirely LTCC designs have currently been restricted to front end antenna switch modules (see Figure 1). For example, this module contains a diplexer, lowpass filters, two PIN diode antenna switches and three SAW filters in a 6.7 × 5.5 mm package.
Generally, past design experience provides a basis for accurate upfront estimates on the cost, size, and performance of the RF module under various analyzed options. This requires a few days to a couple of weeks depending on the level of detail, the number of options, and the commonality between the options.
Laminate is generally lower in cost than ceramic. Usually, the ceramic module must yield a size reduction to be cost effective. This is achieved through embedding more circuitry into the many layers of LTCC. The same size module will almost always be lower cost in laminate. However, laminate can be made costly when fine pitch flip chip die are attached. Fine pitch flip chip devices require higher cost high HDI technology. HDI can be lower or higher cost than LTCC depending on the construction. In some designs, the passives and die determine the module size. The module shown in Figure 2 contains two die and large value passives that cannot be embedded in LTCC. This design contained a balun and a filter with 40 dB rejection to DCS and PCS frequencies.
The higher permittivities and thin layers of LTCC allow for embedding small value capacitors within the LTCC layers. Some LTCC sources have layers as thin as 20 microns. Relative permittivities up to 80 are available in tape thicknesses of 40 microns after firing. This yields capacitor densities of 18 pF/mm2 for two dielectric layers. Laminate capacitor densities are limited to 1 pF/mm2.
This may yield a size advantage for ceramic over laminate. Ceramic also provides a wider range of dielectric permittivities. Relative permittivities range between 5 and 80 for LTCC versus 2 to 5 for laminate. Both laminate and ceramic offer various dielectric thicknesses, however, ceramic provides much thinner capabilities. This is advantageous for capacitors, but it can be a hindrance for some structures.
LTCC gains another size advantage with via technology. LTCC provides for a via in pad (see Figure 3). This allows components to be placed on the pad, since the via is solid metal.
Low cost laminate solutions utilize mechanical drilled vias that are 200 microns in diameter. The via hole is partially filled with metal. However, the hole is too large to completely fill with metal. The remainder is filled with solder mask material. This requires the via be moved away from the component pad since the solder will not adhere to the solder mask material. HDI or added via overplating processes can be utilized for via in pad technology with laminate. These issues usually add significant cost to the laminate product, however.
Another advantage of LTCC is its smaller via and via capture pads. This provides for more compact designs. Vias, however, must be placed farther from the module edge in ceramic than laminate. Therefore, ceramic's advantages are embedding small value capacitors, 30 pF or less, and smaller vias and capture pads. This provides for smaller designs than laminate when components, passives and die, do not determine the module size. This offsets the higher cost of the ceramic substrate and is especially advantageous for fine pitch flip chip die. It also can be more economical than HDI substrates.
Laminate is a lower cost material and wirebond die can be inexpensively protected with injection molding. Ceramic requires a more costly dam and fill operation (see Figure 4), and a cover for pick and place. Laminate currently provides similar or smaller line widths and spaces. Laminate provides 65 micron widths and spaces in high volume with 50 micron capability, whereas many LTCC sources use 80 to 100 microns with some down to 60 microns on inner layers. In addition, laminate uses thicker metal with higher conductivity. This provides lower resistance and inductance. Larger line widths are required in ceramic for the same resistance and inductance. Laminate solutions also offer better second level reliability due to the closer TCE with the mating PCB. Ceramic exhibits a TCE of 7 × 10 4 whereas laminate and the mating PCB have a TCE between 12 × 10 4 and 14 × 10 4. This yields less stress of the solder interconnect during a second reflow for the module attach. In addition, a third reflow may be required if the PCB is two sided. The matched TCE also yields lower stress on the mechanical solder interconnect to the PCB over the product environmental thermal conditions.
Ceramic modules typically employ solder bumps or balls, in a BGA package, to help relieve the solder interconnect stress due to the mismatched TCE of the ceramic and the PCB. In addition, critical interconnects are moved in one row and removed from the higher stress module corners. They may be duplicated for added reliability. The size of the package is also critical to reliability1. The laminate reliability, however, is not easily achieved.
Advancements in the technologies
Recent technological advancements have begun to blur the once sharp division between ceramic and laminate.
One of the advantages of ceramic is the capability to embed capacitors. New technologies allow embedding capacitors inside of the laminate. The current technologies are suitable only for large value capacitors. They use X7R dielectric2 or very thin submicron films3. However, lower value embedded capacitors are on the horizon.
Samples of this technology have been proven but it is not currently qualified for high volume manufacturing. Expectations are that laminate modules with embedded capacitors will be in volume production in 2004.
Furthermore, both of these processes allow for embedded resistors. The Shipley process is limited to the same material Ohms/square throughout. The Dupont process can mix and match Ohms/square paste at a small, added expense than a single paste. Both of these technologies are cost effective today if the embedded passive count approaches six components per cm2. It is typically more costly with fewer components, however, expectations are for lower costs with higher volume.
On the ceramic side, injection molding processes for ceramic has been developed. This reduces the overall package cost of the ceramic module by optimizing the many factors that affect reliability.
Although laminate exhibits higher dielectric loss, it has better metal systems than LTCC, LTCC provides better dielectric loss while sacrificing the metal. The fired metal is thinner and lossier.
Filters, Baluns and other devices
Laminate filters are available for BT and WLAN application at 2.4 GHz and 5 GHz (see Figure 5). Baluns and other structures have also been developed at these frequencies. These devices lower the overall package cost while providing RF selectivity to the receiver4. The filter protects the receiver from PCS/DCS and cellular. It also provides some attenuation to the harmonics of the transmitter and systems such as 802.11a operating in the 5 GHz range.
The amount of filtering depends on the desired protection level, distance and dynamic range of the receiver, and the compression point of the LNA. However, the compression point is tightly correlated with the current drain of the LNA.
Filtering cannot provide protection to in-band interferers like 2.4 GHz wireless phones and leaky microwave ovens. LNA compression is the only protection to in-band interferers. Filtering can provide protection to out-of band interferers. There is still a balance with the LNA compression and filter selectivity.
Adequate filtering may not be achievable for a low P1dB LNA without higher insertion loss. However, the insertion loss will impact the overall receiver noise figure since it is in front of the LNA. This higher filter insertion loss may require an unachievable LNA noise figure to met the overall receiver sensitivity. The opportunity to embed the filter in the substrate arises from using a HPF instead of the traditional BPF. The benefits include eliminating the component, less required space, BOM cost reduction, and cost reduction through the use of a less expensive standard moldcap.
Ceramic filters have a higher profile and demanded a higher cost dam and fill process with a cover. Advanced designs allowed enough selectivity to be designed into the laminate filter to remove the ceramic and provide a smaller height and lower cost alternative.
Other componentsAnother technology that lowers the overall system cost is integrated antennas. A full BT module is depicted in Figure 6, an external reference signal is required. It contains one die with digital and RF functionality. The design includes embedded filters and a balun. The antenna is integrated into the package. It is a 93 pad BGA that measures 15 × 15 × 6.5 mm, however, this could be reduced to 4 mm in height.
Embedded shielding is also a factor in cost reduction. Shielding may be required to reduce emissions to meet regulatory requirements, provide immunity from interferers in proximity, as well as provide proper operation of the transceiver.
A case of the latter is depicted in Figure 7. A signal can couple in the circuit board after the BT or WLAN front end filter. This may cause a larger noise level into the BT or WLAN LNA than an interferer that is external to the product. Circuit coupling may also affect the PCS receiver. This may be a direct coupling as depicted in Figure 7, or it may couple later in the receiver chain. It may enable AGC if it is within the AGC bandwidth. The AGC bandwidth is typically larger than the IF bandwidth. This could enact 30 dB of AGC in the receiver degrading the receiver sensitivity. Care must be taken during the circuit layout to avoid coupling.
In addition to these direct coupling mechanisms, the BT and PCS transceivers must cohabitate with each other's clock and spurious products.
Effects of these emissions are not easily predicted. Shielding at the package level can contend with these requirements while meeting the regulatory stipulations imposed on the system. Shielding is typically accomplished at the product level, however, package level shielding can provide a cost reduction for many product developers by eliminating expensive and logistic manufacturing provisions. One alternative solution is shown in Figure 8 where the shield is encapsulated with the die. The module can be incorporated with multiple shields to protect against baseband and radio interferers or transmitter and receiver circuitry.
In addition to these technologies, other processes, such as chip and wire, flip chip, stacked die, embedded passives, and double sided surface mount can be part of the solution.
Where the rubber meets the road
The greatest impact for product cost reduction is obtained early in the design cycle. If package cost, size, and performance for various options can be determine early on, many of the restarts, redirects and failures can be eliminated.
Some typical assembly and substrate cost factors are given in Figure 9. Following these factors, the architecture can be optimized to incorporate low cost RF designs and yield the optimum module solution. An example tri-band amplifier is shown in Figure 10.
The initial design includes tuning and values and part locations are adjusted in the first prototype stage. The best results have been achieved from loadpull device data or application board measurements, however, design with device models is also possible.
Most RF power amplifier design projects include thermal management simulation and design. This can be designed in any substrate. It can include thin film, glass, or silicon for integrated passive networks.
Tomorrow's Sip Trends
RF SiP is increasing in popularity. It provides simplification of the mating system board, increased functionality per area or volume, reduced final assembly cost and parts count, improved electrical performance, greater final assembly yield, improved time to market, and reduced RF expertise of the end user or assembler.
As a result, RF SiP is gaining widespread acceptance within the industry for high volume packaging systems.
 Richard E. Sigliano Kyocera America Inc., "Ceramic Vs. Plastics: Part I Is Perception Reality?" MEPTEC Report Quarter One 2003 pp.26-28
 William Borland, "Designing for Embedded Passives", Printed Circuit Design, August 2001
 Percy Chinoy and Marc Langlois, "Insite Embedded Resistors and Capacitor Technologies", IPC Designers Summit, March 23-27, 2003
 Michael P. Gaynor and Douglas J. Mathews, "System-in-Package for WLAN/PAN Aids Coexistence with Digital Cellular", High Frequency Electronics, January 2003, pp. 30-41.
Figure 1. LTCC Cellular Front End Switch Module Design. Figure 2. Bluetooth Laminate Module. Figure 3. Two Layer with and without Plated Via. Figure 4. Dam and Fill Die Protection with a Module Cover. Figure 5: Laminate 2.4GHz WLAN and Bluetooth Embedded Filters. Figure 6. BT Module with Integrated Antenna and Shield. Figure 7. Coupling Resulting in Improper Transceiver Operation. Figure 8. Shield Integrated into Package Overmold. Figure 9. Some SiP Cost Factors. Figure 10. Tri-band Power Amplifier Design.
ACG - automatic gain control BGA - ball grid array BOM - bill of materials BPF - band-pass filter BT - Bluetooth DCS - digital communications system HDI - high-density interconnect HPF - high-pass filter LNA - low-noise amplifier LTCC - low temperature co-fired ceramic PCB - printed circuit board PCS - personal communications system PIN - positive-intrinsic-negative SAW - surface acoustic wave SiP - System in package TCE - temperature coefficient of expansion WLAN - wireless local area network