Microwave component and module manufacture has been well understood and automated for many years. Myriad applications benefit from the capabilities driven by high volumes of microwave point-to-point links. But at frequencies around 60 GHz, the manufacture of millimeter-wave systems turns from a highvolume automated operation into a hand-tuned black art.
Success in building millimeter-wave modules and subsystems by integrating very high-frequency monolithic devices often requires years of hard-won experience and plenty of manual intervention. Industry veterans describe difficulties in designing module packages and how they tune components and add absorbing material in just the right places to make the modules work and optimize performance.
With few exceptions, millimeter-wave companies are small specialists, hand-building products for military, aerospace, and research applications. These packaging and manufacturing difficulties limit the deployment of millimeter-wave systems and the use of these bands because they prevent substantial reductions in cost, size, and weight.
Yet with increasing demand for video streaming to mobiles, backhauling high-definition (HD) video in broadcast and CCTV applications, there’s lots of interest in the vast bandwidth available at millimeter-wave frequencies. Uncompressed HD video needs 1.4 Gbits/s. In addition, 3G Long Term Evolution (LTE) service is just starting to be deployed in mobile networks, and each basestation needs 100-Mbit/s backhaul capacity. Up to 7 GHz of bandwidth is available for these applications in various millimeter-wave bands (U.S. band 57 to 64 GHz) (see the table).
The wide bandwidths available at these ultra-high frequencies offer enhanced resolution for radar and imaging systems, too. For example, the range resolution of a frequency-modulated continuous-wave (FMCW) radar is directly proportional to the swept frequency bandwidth. Accurate millimeter-wave radars are now being deployed at 77 and 79 GHz for non-military, highvolume applications such as vehicle adaptive cruise control and lane-change assistance.
The latest screening systems for detecting weapons and explosives under clothing also demand wide bandwidths in the millimeter- wave bands (Fig. 1). Active transmitting systems often rely on a “radar-like” swept frequency approach, and passive, receiveonly systems require good sensitivity and high gain across more than 20 GHz in the W-band around 100 GHz.
Most microwave and millimeter-wave components and subsystems comprise metal, or at least metal-coated, enclosures into which cavities for mounting miniature microwave integratedcircuit (MMIC) chips and other components are milled or formed. Often gold-plated, these enclosures physically protect the MMIC chips, wire-bonds, and other components from damage in manufacture and use and from the external environment. They also protect the components from interference caused by electromagnetic radiation from the electronics in the rest of the system and the operating environment.
Of course, such metal modules are far from ideal in terms of size, weight, and cost for many applications. For example, airborne systems that provide real-time imagery to assist pilots landing helicopters in the dark, in bad weather, and in “brownout” conditions (dust and sand clouds) need small and low-mass millimeter- wave components, as do the imaging systems now being tested on unmanned aerial vehicles (UAVs).
All microwave circuits radiate energy, from interconnect tracks, from bond wires, and from the chips themselves. When the wavelength approaches the dimensions of the MMIC chips, many electromagnetic effects, perhaps negligible at lower frequencies, become much more significant and can even dominate the functionality and destroy performance.
Radiated energy couples into other parts of the circuit and often causes unwanted, sometimes catastrophic, behavior. Examples include resonance in the “cavity” that houses the MMIC chips and non-bulk conduction in filter structures in planar circuit boards. Resonances often render a millimeter-wave module completely non-functional. The ease with which unwanted millimeter-wave radiation “leaks” into and affects all parts of a system makes realworld equipment building a substantial challenge.
One approach that can help mitigate these effects involves the use of flip-chip MMICs, where the chip is mounted face down onto a substrate with interconnections. Example substrates include thin-film or thick-film circuits formed on quartz or ceramic (usually alumina) or on a variety of organic highfrequency circuit board materials.
The chip connections, usually bumps of various types of solder, can provide transitions with low loss at high frequencies. But the proximity of even a nonconductive substrate surface can affect the chip’s high-frequency performance.
Flip-chip assemblies lack a large-area contact to a thermally efficient substrate, though. As a result, the relatively high thermal dissipation of many high-frequency MMIC chips must be managed through the front face of the chip and the mounting bumps. Mismatch between the thermal coefficients of the MMIC chips and the substrate materials can also create reliability problems in operation.
Solutions to this issue, such as inserting non-conductive underfill between the MMIC chip and the substrate, often impact the microwave performance. In addition, the planarity of many MMIC chips may not be sufficient to provide reliable low-loss bumps for all connections. The inability to investigate, adjust, and tune or even rework flip-chipped MMICs also means that flipchip assembly isn’t an approach that’s generally applicable for all millimeter-wave subsystems.
It may be possible to make the cavity enclosing the active devices small enough to prevent a resonance mode from being set up at the fundamental operating frequency of the circuit. However, resonance modes at higher frequencies still couple into the devices and structures, so they contiue to seriously impact the circuit performance.
Continue to page 2
Absorbing material is unavoidable even with this approach. Also, at millimeter-wave frequencies such as 60 GHz and higher, it becomes impractical to make the cavity small enough to avoid these resonances, especially when access is needed for the lowcost automated assembly and bonding of the MMIC chips.
On top of that, there’s a strong trend toward higher levels of on-chip integration at 60 GHz and higher frequencies to substantially reduce system cost. This ultimately creates ever larger millimeter-wave chips. For example, a handful of integrated receiver and transmitter chips operate at 60 and 70 GHz, some based on silicon-germanium (SiGe) biCMOS, and wide galliumarsenide (GaAs) and gallium-nitride (GaN) amplifiers with onchip power combining to achieve high power output at 60, 70, and even 94 GHz.
Another approach to managing these problems is to package microwave chips with radiation-absorbent material (RAM). Reflection of microwave energy from, and transmission through, RAM is low. Coating the inside of the chip cavity and mounting blocks of RAM inside the cavity are common approaches.
However, some of these materials must be pretty thick to be effective so they can match the impedance of energy radiated at millimeter-wave frequencies. Consequently, it’s not easy to form RAM blocks to fit at the appropriate location and mount inside cavities with dimensions that are only a few millimeters. Often times, it requires manual intervention by trained engineers. This in turn brings higher costs and forces lower quantities.
Other thinner materials often rely on wavelength effects to absorb the radiation. It makes them inherently narrowband and possibly unable to suppress harmonics of a particular frequency and higher-order resonance modes of cavities. In addition, many materials that absorb microwave frequencies are far less effective on millimeter waves at 60, 70, and 100 GHz. When using materials that absorb millimeterwave energy in close proximity to MMIC chips to inhibit unwanted effects, great care is required to avoid degrading the wanted performance of the circuit and thus rendering it unusable.
For one or two millimeter-wave applications, such as very short-range wireless personal-area networking at 60 GHz or automotive radar at 77/79 GHz, the operating bandwidth is relatively narrow or the output power is low. Conventional semiconductor packaging, where a plastic material encapsulates the MMIC chips, can be used at low cost in these cases—but at the expense of performance. However, large losses at high frequencies mean this approach simply isn’t an option for most microwave, let alone millimeter-wave, components and systems.
Other solutions include ceramic packages that are utilized for semiconductors, such as low-temperature co-fired ceramic (LTCC) or alumina. Familiar in many military or aerospace components, such solutions usually provide good environmental protection and reliability, but remain very expensive for commercial use. Some of these approaches involve machining or etching and then coating silicon substrates, which is another high-cost approach. All of these techniques still require conductive coatings or surfaces, or even RAM, to absorb any unwanted microwave energy.
POTENTIAL SOLUTIONS: AUTOMATED OPTIONS
MMIC Solutions uses a simple low-cost technique to form a cavity for the MMIC chips in a multi-layer circuit board, with no metal milling required and a low-cost lid to encapsulate and protect the MMIC devices (Fig. 2). The easily fabricated lid can be automatically assembled. Its coated, conductive surface is precisely placed to absorb unwanted millimeter-wave energy and eliminate the resonance issues even at operating frequencies higher than 100 GHz. The technique has been used to build modules for radio communications in the V-band at 60 GHz (Fig. 3) and for passive imaging in the W-band at 100 GHz.
Automated assembly of modules and components brings repeatability—for instance, in chip placement and bond wire lengths—and more predictable performance. Indeed, automation will often lower losses and help improve performance, and it has been used for many years in the microwave component arena. This is evident even in a number of low-volume military and aerospace applications.
Automation also plays a significant role in driving down costs for microwave modules used in point-to-point radios for backhauling mobile operators’ network traffic. Furthermore, it’s used in the newer low-cost vehicle radar systems.
However, automated manufacturing of millimeter-wave systems at frequencies of 60 GHz and above is considerably more difficult. The tolerances required in the placement and bonding of MMIC chips are tighter, and fewer manufacturers have the more expensive equipment that’s required. The packages must also be large enough to allow access for the machine handling and pickup collets used in automated assembly, which creates problems with cavity resonances, absorber materials, and other factors, as outlined earlier.
The high system complexity such as high-order modulation and rapid frequency agility inherent in many commercial millimeterwave applications is not only pushing higher levels of chip integration and the problems it poses, it’s also driving the need to integrate more components into the module package.
These components can be “chip and wire” MMIC devices or surface-mount components from connectors and capacitors to regulators and oscillators. Conventionally, the metal enclosure is made larger to suit surface-mount circuit boards interconnected to high-frequency components by glass-to-metal seals. This approach isn’t a route to small size and low cost for widespread deployment, though.
A module with surface-mount components and high-frequency “chip and wire” devices assembled onto the same single circuit board is possible if the board material has good microwave or millimeter-wave properties. Multilayer substrate materials from Rogers (e.g., “Duroid”), Taconic, and others are well known for microwave apps.
However, most of these high-frequency circuit boards are made using glass fibers for rigidity and support. This can present problems when cut to form pockets or cavities for MMIC chips. Ragged edges due to the fibers can confuse the machine-vision system used for high-precision automated assembly.
Continue to page 3
Also, the layers can move relative to each other during high-temperature assembly processes, because the fibers in different layers are oriented in different directions. This may substantially reduce performance, or even be catastrophic, for the high-precision structures and placement necessary for automated millimeter-wave assembly.
In the multilayer circuit board mentioned earlier, MMIC Solutions uses new liquidcrystal polymer (LCP) materials from Rogers. They eliminate such effects and enable high-yield automated assembly of MMIC chips and surface-mount components into modules operating at 60 GHz and above.
Passive imaging at 100 GHz requires very good receiver sensitivity, usually specified as the noise equivalent temperature difference (NETD), which is the smallest detectable difference in wideband noise, of around 0.7K. This requires a low receiver noise figure (around 5 dB at 100 GHz) and very low high-frequency loss in the module circuit board.
MMIC Solutions measures less than 0.2 dB/mm loss for a buried stripline in the multi-layer LCP substrate of its MSi102 product for W-band imaging. The MSi102 mates to a standard WR-10 mechanical interface, also formed in the module substrate. The waveguide termination (backshort) is formed by the lid used to encapsulate the MMIC devices (Fig. 4).
In imaging, the packing density of an array of receivers is additionally important in determining the resolution of the image and, therefore, in the system’s ability to detect small objects that could be a threat. For this reason, imaging arrays are perhaps the most demanding of all millimeter-wave applications on the small size of the module.
Using the latest multilayer substrates to reduce size, MMIC Solutions developed its latest MSi200 series receivers for imaging systems. The modules are 18 mm long by 8.5 mm wide by 5. 5mm high, and they measure less than 850 mm3 in volume.
Not all systems working in the millimeter- wave band need small size, low weight, and reduced cost to support increasing volumes. Some military, aerospace, and research radio and radar apps are inherently small quantity-wise, for which “hand-tuning by experts” is an appropriate solution.
But increasing demand to use the large available bandwidth at millimeter-wave frequencies for commercial systems and services drives the need for new solutions. Smaller, lighter components than the milled metal blocks are welcomed by almost all and demanded by some applications.