In just a handful of years, the packaging and interconnect technology known as low-temperature cofired ceramics (LTCCs) has entered the mainstream of electronics design. LTCCs, which up until about five years ago were mostly relegated to the realm of low-volume military applications, are now being applied in high-volume consumer designs, especially in the wireless arena. And while wireless designs represent a growing market for LTCCs, other areas such as automotive, medical, and instrumentation also are making use of the technology.
LTCCs provide a modular approach to building highly integrated subsystems that are capable of operation at microwave frequencies. An LTCC module consists of a multilayer substrate constructed from layers of a dielectric foil (Fig. 1). RF passive components may be embedded within the substrate, along with the interconnect. Passive and active components both may be mounted or formed on the top-most layer.
RF design is aided further by the use of low-loss dielectrics and high-conductivity metals, which make it possible to construct high-Q resonators and filters. It's also possible to build fully shielded structures within the LTCC stack. For interconnect between the module and the pc board, wirebonds or solder balls may be used. In addition to high-frequency performance and circuit integration, LTCC modules offer superior thermal performance, hermeticity, and high reliability.
Construction of an LTCC substrate starts with the dielectric material. This material is a glass ceramic cast tape known as Green Tape, which is DuPont's name for its series of LTCC material systems. First, pieces of the tape are cut to the desired-size foils. Next, these foils are punched or drilled to create vias. Then, using screenprinting or photo-imaging methods, metallization for via fills and conductor traces are added (Fig. 2).
Passive Components Possible
Resistors and capacitors (thick film) and inductors and cavities can be embedded into the foils. The metals used for the substrate's inner layers, via fills, and wire-bond pads are typically gold or silver. Alloys of platinum and gold or palladium and silver may be used for solder assembly. The number of foils or layers required will depend on the complexity of the design. Substrates with over 50 layers have been demonstrated. Stacks with up to 50 layers are considered production-worthy.
Once the individual layers of tape are formed, they are collated, stacked, and laminated to create a "green" multilayer structure. That structure is then fired at approximately 850°C to 900°C to harden the material. After firing, the stack is cut to size. The top layer is then completed with the mounting of passive and active components. Bare die may be assembled using flip-chip techniques. Passive components may be formed by using thick-film techniques or by mounting surface-mount discretes on the top layer.
The process may be likened to techniques used to construct thick-film hybrids. Yet the LTCC process is more cost-effective because it relies on parallel processing of layers rather than the sequential buildup of layers in a conventional thick-film device. Another distinction is that conventional thick-film hybrids require multiple firing steps, while LTCCs are fired in one stage—hence, the label "co-fired," which reflects the fact that in an LTCC, the dielectric layers and conductors are fired together.
Among substrate materials, LTCCs face varying degrees of competition from other ceramic materials and from organic pc-board materials. In the ceramic domain, there are high-temperature cofired ceramics (HTCCs), which are multilayer structures based on dielectrics such as 92% alumina. Unfortunately for HTCCs, their high firing temperature (approximately 1500°C) precludes the use of highly conductive metals, which are required in high-frequency work. Typically, HTCC metals are tungsten or molybdenum (refractory metals) that are plated with nickel for solderability and gold to promote wirebonding and prevent corrosion of the nickel. On the other hand, the 850°C to 900°C firing temperature of LTCCs allows the use of highly conductive silver and gold traces.
Among organic materials, the LTCC weighs in against traditional board materials like FR4 and high-performance materials such as PTFE (teflon). LTCC either matches or exceeds the high-frequency and thermal characteristics of these materials.
Samuel Horowitz, marketing manager at DuPont Microcircuit Materials, Research Triangle Park, N.C., notes that the first generation of LTCC dielectric materials was better than FR4 in terms of high-frequency performance. But they weren't as good as the more-expensive PTFE. The current second-generation LTCC materials, however, now have the same dielectric properties as PTFE. Furthermore, the use of LTCC materials allows passive components to be embedded in the multilayer stack, while PTFE does not.
Vern Stygar, product manager for LTCC materials at Ferro Corp., Vista, Calif., cites two market factors that are driving the adoption of LTCC solutions—rising operating frequencies and rising power-dissipation levels. In the past, when the big wireless applications were running at about a gigahertz, demands on the dielectric materials weren't so great and traditional organic pc-board materials were sufficient.
Bluetooth Changes Things
The arrival of the Bluetooth wireless protocol, though, with operation at 2.45 GHz, changes the situation. According to Stygar, "insertion loss and the dielectric constant become extremely critical" at this frequency and above. Consequently, Bluetooth provides a major opportunity for LTCCs. As operating frequencies go up into the tens-of-gigahertz for satellite, LMDS, fiber optic, and automotive collision-avoidance applications, LTCCs will find more uses.
The other driving force, increasing power-dissipation levels, also favors LTCCs. Stygar notes that the thermal conductivity of an LTCC is 2 W/m/k, versus 0.5 W/m/k for organic materials. That value could jump to 50 W/m/k for LTCCs with vias. Other thermal characteristics put LTCCs ahead, too.
The temperature coefficient of the resonant frequency (Tf) is a parameter that dictates frequency stability and depends on two factors—TK, the temperature coefficient of capacitance, and TCE, the temperature coefficient of expansion for a given dielectric material. Tf ≈ −1/2TK − TCE.
Tf for an LTCC material is said to be less than 10 ppm/°C versus 80 ppm/°C for FR4. The variation in TCE from LTCCs to FR4 is itself significant. LTCCs have a TCE that closely matches that of silicon and gallium arsenide. As a result, semiconductor die can be mounted to the top side of the multilayer stack through direct chip attachment.
When LTCCs are placed in extreme temperature environments, they also fare well. They can withstand temperatures greater than 500°C and are better able to withstand thermal shock than organic materials. Consequently, another big application of LTCCs is in automotive designs. While wireless markets represent the fastest growing application of LTCCs, Horowitz reports, the automotive sector captures the largest market share (35%) of LTCC devices. At the moment, he says, LTCCs are mostly being used in engine controls. But in the future, intelligent transportation systems (ITSs) will present a great demand for LTCCs. The ITS area encompasses a variety of functions, ranging from in-vehicle navigation and information systems to collision-avoidance, obstacle-detection, and adaptive cruise-control applications.
Beyond wireless and automotive applications, the next frontier for LTCCs may be fiber-optic networking. Horowitz says, "This looks like wireless looked to me four years ago." He notes that LTCC technology is equipped to dissipate the high power levels generated by lasers. It's also a better choice for the electro-optical interconnect. With signals running at 2.5 Gbits/s, organic circuit boards are already at their performance limits. The migration to 10 Gbits/s will only make matters worse for pc boards. A material such as PTFE may offer the needed high-frequency performance, but not the integration of LTCCs, which also satisfy fiber's demand for hermeticity.
An example application would be an optical switch. One version, which already exists, is a thin-film silicon MEMS device packaged in an LTCC. But in the future, Horowitz speculates, it may be possible to build a thin-film structure on the surface of the LTCC and then embed control circuitry within the stack.
For fiber-optic and other applications, the decision to implement LTCC designs rests heavily on system cost, rather than the cost of LTCC materials alone. "LTCCs," Horowitz claims, "can provide the lowest system cost for fiber optics." Then again, fiber optics and other new applications will benefit from progress made in building LTCC modules for wireless. "As circuit manufacturers have ramped up to meet the needs of the wireless industry, they've gone down the learning curve and brought down pricing," he continues. "This knowledge can be also applied to fiber optics."
The predominant method for laying down metallization in an LTCC is screenprinting. According to Bill Minehan, general manager at LTCC foundry Coorstek of Chattanooga, Tenn., the modules his company builds typically contain screenprinted lines and spacings with an 8-mil thickness. He observes that "screenprinting is effective for lines and spacings down to 3 to 3.5 mils." Beyond that range, photo-imaging techniques become necessary.
The use of photo-imaging to pattern metallization and create via fills has let LTCC designers incorporate finer features, including landing pads for high-density applications such as flip chip. Photo-imaging also enables higher circuit densities and better high-frequency performance. Ray Brown, manager of RF module design at LTCC manufacturer National Semiconductor, Santa Clara, Calif., says that his company's current design rules permit 2-mil lines and spacings with photo-imageable conductors. According to Brown, 1-mil lines and spacings may be ready for production in about a year.
One of the benefits of narrower lines and spaces has been the ability to design coplanar waveguides (CPWs). Unlike microstrips and striplines, which are 3D structures with ground planes above and below the transmission line, CPWs include a ground plane and a signal line on the same level. This feature makes them easier to design and isolate. These structures couldn't be built with screenprinting of conductors because the gap between the transmission line and the ground plane couldn't be made fine enough to achieve sufficient capacitance.
Ferro's Stygar notes that it's now possible to pattern conductors with line widths of 20 µm or less. Tolerances of ±5% on these lines isn't unreasonable, and even ±2% precision may be achievable. High-resolution photopatterning can produce not only high-density interconnects, but also spiral embedded inductors and interdigitated capacitors (Fig. 3).
Meanwhile, advances in capacitor materials are improving the performance of embedded components. The development of high-K materials makes it possible to build higher values of capacitance in smaller spaces. While past designers may have been limited to materials with a K below 30, LTCCs now offer Ks beyond 100. For example, Ferro has a dielectric material with a K of 130. This material is compatible with the company's A6 tape system. There also are efforts to develop materials with very low K, which are needed for high-speed digital signaling.
Low-loss dielectrics are another element critical to high-frequency operation. A recent advance in this category is DuPont's 943-A5 Green Tape. This material system has performed well at frequencies up to 40 GHz, when tested against 99% alumina and other low-loss ceramic materials (Fig. 4).
Other improvements in the works include the development of tape with a higher TCE. A higher TCE, closer to that of pc-board materials, will provide a better match between the LTCC module and the board, improving reliability. Additionally, innovations in materials and process are expected to improve the accuracy of embedded resistors.
The difficulty with buried resistors is that they can't be laser-trimmed, so they generally have tolerances no better than 5% to 10%. As a result, they're typically limited to digital applications. For analog functions, the use of laser-trimmable surface resistors is an option.
When it comes to designing an LTTC module, there are no standard package types. Designs are custom, and designers must model the package to simulate its electrical performance. One way to do this is with a tool such as Momentum, which is a part of the ADS design platform from Agilent EEsof EDA. This software could be described as a two-and-a-half dimensional planar tool because it allows modeling of a rectangular shape with depth. It can model the individual LTCC layers with their dielectric, vias, and metallization. To model the stack, Momentum takes a composite of all the layers and performs the analysis.
Chris Mueth, product marketing manager for RF and microwave products at Agilent EEsof EDA, says Momentum is particularly adept at performing simulations up to 40 GHz and beyond. For more complicated geometries or arbitrary shapes, though, it may become difficult to model LTCCs in a two-dimensional planar fashion. In these cases, 3D analysis such as that performed by Agilent's High Frequency Structure Simulator (HFSS) is required.
Examples of arbitrary shapes would include a circle, an S-shape, or a complicated bend in the stack, as viewed from above. Even bond wires, which may be treated as rectangular structures at lower frequencies, have arches. At high frequencies, it may become necessary to model them in three dimensions.
Of course, the need for accurate data lies at the heart of all simulation. As the frequencies for LTCC applications rise and line widths shrink, it becomes extremely critical that the industry has accurate data on material properties. Designers particularly need accurate values for the dielectric constant and the loss, as well as the conductivity of metals used for interconnect.
The problem is multifaceted. Dielectric and metal properties change as the LTCC material goes from being an unfired Green Tape to being a sintered (fired) substrate with embedded components and interconnect. Properties change once more after the ceramic-metal composite is sintered. And at different stages of processing, different test methods are required to measure the same properties.
The industry is applying a number of different test methods to gather data about LTCCs. There's also no consensus among the various material suppliers and LTCC foundries on standards for testing LTCCs. This situation may be changing, though. At the National Institute of Standards and Technology (NIST), Boulder, Colo., the Radio Frequency Technology Division is working to help the LTCC industry obtain accurate measurements of dielectric properties at various stages of processing. NIST is evaluating test methods being used in LTCC production and developing new methods to ensure accurate and repeatable test results for measurements up to 100 GHz and above.
According to Jim Baker-Jarvis, project leader of NIST's Electromagnetic Properties of Materials Group within the RF Technology Division, "It turns out the test problem is extremely difficult." Measuring the performance of the individual dielectric layers may be relatively straightforward. But the big problem comes after the stack is fired. Baker-Jarvis says that LTCC foundries and designers need to know how they determine the properties after they've put the layers together.
Among its various test activities, NIST is evaluating the "T" and "ring" resonator techniques, which LTCC foundries commonly use to test fired materials. In these methods, a T or ring structure is placed on top of an LTCC stack and driven to resonance. By measuring resonant frequency and Q, it's possible to determine the dielectric constant and the loss tangent of the material. But despite their popularity, the accuracy of these methods is unknown. NIST is evaluating their precision with an eye toward developing better test techniques than the T and ring resonators. Baker-Jarvis notes, "We don't believe that these are the most accurate methods for loss measurement."
One of the advanced methods being pursued is the Fabry-Perot technique for measuring the dielectric constant at the sintered stage for frequencies ranging from 60 to 110 GHz. The Fabry-Perot method uses an open-cavity resonator consisting of two separated mirrors. One mirror is placed directly underneath the sample. The other mirror, a concave device with a coupling aperture in it, is placed at some distance (perhaps 20 cm) above the sample.
With a waveguide feeding them, the mirrors are driven to resonance, enabling the calculation of the dielectric constant and the loss tangent. The concave shape of the top mirror focuses the beam sufficiently and thereby limits radiation leakage along the sides. This reduces the influence of beam diffraction at the edge of the LTCC sample.
Another approach to testing sintered materials, the "whispering-gallery" mode, provides highly accurate measurements of dielectric loss. In this method, a disk-shaped LTCC sample rests on a dielectric rod. The sample is driven into resonance with coupling loops. Because of its high accuracy, this method helps reduce conductor losses.
Although measuring sintered materials with metals attached is the primary challenge, there is still a need to gather information on unsintered materials. The split-post resonator method now in development offers both simplicity and accuracy. It lets designers measure the dielectric constant within ±0.5% and read the loss tangent with 1 × 10−5 accuracy. In this approach, a thin sample of material is placed between two fixed dielectric resonators. Selected for their low dielectric constant and low loss, these resonators are driven using loop coupling. Changes in resonance and Q of the fixed resonators with and without the sample in place can be used to calculate the sample's dielectric properties. This method works for frequencies ranging from 1.5 to 12 GHz.
As the industry refines and standardizes its test methodologies, LTCC designers and manufacturers will benefit. Better test data will help LTCC material developers and foundries improve their processes, reduce costs, and extend the limits of high-frequency performance. At the same time, designers will find it easier to compare different material systems in their applications.
For further reading:
- "Current and Future Applications for Thick-Film and Low-Temperature Cofired Ceramic Technology in the Automotive Industry," Samuel J. Horowitz, DuPont Microcircuit Materials, Summer 2000 MMRC meeting, Dearborn, MI. Go to www.dupont.com/mcm/automotive/mmrc.html.
- "Frequently Asked Questions for LTCC Vendors" and "Design Rules for Physical Layout of Low Temperature Cofired Ceramic Modules," Revision 8.1, Jan. 5, 2000, National Semiconductor. Go to www.national.com/appinfo/wireless/0,1822,357,00.htm.
- "Ferro LTCC System for Wireless Packaging Applications," Liang Chai and Simon Turvey, Ferro Corp. Go to www.ferro.com/content/product_render.asp?use_case=EM2&Phase=products&FamilyId=18&DivisionId=3.
|Companies That Provided Information For This Report|
Agilent EEsof EDA
National Semiconductor Corp.
National Institute of Standards
RF Technology Division
Electromagnetic Properties of