New Interconnect Technology Addresses System Integration
Mr. Sherrer was founder and president of ACT MicroDevices, Inc., later named Haleos, Inc., a manufacturing and product development company for micro-fabrication based components and assemblies. After Rohm and Haas acquired Haleos (a publicly traded $8B company with over 18,000 employees serving the semiconductor industry) in 2002, Mr. Sherrer took over responsibilities for new product development and commercialization of micro-fabricated products. In this role, Mr. Sherrer has managed the development of Rohm and Haas’ Si-Pak™ hermetic wafer-level packaging technology, and PolyStrata microwave device and packaging technology. Mr. Sherrer has a MS in EE and BS degrees in physics and philosophy from Virginia Tech. He has over 100 patents issued or pending.
MWJ: The PolyStrata micro-fabrication technology creates a unique transmission line structure by depositing multiple metal/plastic-patterned layers on a flat sheet to form an air-filled rectangular micro-coaxial line with ultra-low dispersion for broadband RF components. The resulting structure allows thousands of devices to be integrated onto a single substrate. Before we discuss some process details or characteristics of the resulting components, let’s talk about the history of this technology. I understand the research was done in partnership with University of Colorado. How did this research come about and what can you tell us about the team that worked on it?
DS: The research came about as a result of winning a substantial Department of Defense Advanced Research Projects Agency (DARPA) program, which we developed with BAE Systems. The program leveraged our PolyStrata™ microwave process technology with BAE System’s vision for low-cost phased array communications systems. So far, we have demonstrated genuine success with verification conducted by independent test and measurement laboratories. Our future plans are to offer Polystrata™ for commercial and military components. Also on our team are Dr. Zoya Popovic and Dr. Dejan Filipovic of the University of Colorado, Boulder. They serve as the team’s microwave physics, design and modeling partners. Their expertise (as evidenced in two dozen publications and two PhD dissertations on the topic) largely transformed our concepts and fabrication know-how into new low-cost microwave components and sub-systems.
MWJ: The process creates “rectacoax” transmission lines. Could you describe this type of transmission line and its benefits to others such as microstrip or coplanar stripline?
DS: In short, the rectacoax loss is 5 to 10x lower compare to coplanar waveguide (CPW) on conventional alumina substrates, the isolation is 1000x higher, it is dispersion free past 100GHz, and enables real 3D and multi-layer circuits with 90° Z turns. Polystrata propagates a distortion free, single TEM mode at all frequencies of interest. In fact, the measured results are so good that we typically achieve virtually identical responses compared to our simulation models. The third dimension is powerful – imagine really having a cube of microwave circuitry with the design freedom to fold signal lines in any dimension, embed active devices, all while providing superior thermal conductivity because the entire circuit is connected through 400 W/m K copper ground plane structures. Using conventional transmission lines it is encumbering to crossover two signal lines without excessive coupling and loss problems. However, PolyStrata enables complex 3D wiring networks with hundreds of crossovers into small chips that perform non-blocking NxN routing switch matrices at frequencies up to 100 GHz with nearly immeasurable isolation.
Our PolyStrata rectacoax structure is a fully shielded copper outer conductor with a copper center conductor surrounded by an air dielectric and behaves mostly like circular coax transmission lines. The square or rectangular shapes fit with our PolyStrata fabrication method. The unique properties and characteristics of it have been pretty well described in papers such as the references in your article.
MWJ: I understand the rectacoax lines can be quite small and offer superior performance at upper microwave/millimeter-wave frequencies. What’s the relationship between the line dimensions and electrical performance?
DS: The PolyStrata rectacoax structure is on the order of a couple of hundred microns in size for our Ku and Ka-band applications. Doubling the height and width of the center conductor reduces the loss by half in dB. Loss will increase with the square root of the frequency due to the skin effect. There isn’t a significant loss or dispersion component from a solid dielectric so this simplifies things. You can find those rules of thumb right on the Microwaves101.com website for a simple analysis.
Rectacoax behavior pretty much follows the circular coax trends, even though there isn’t a clean formula for loss, size and frequency, however people have derived formulas for both analytical and numerical cases as early as T.-S. Chen in 1960, if not earlier. To give you an idea, at Ka band, the losses of 50-Ω coax lines with a 650um outer conductor cross-section measure 0.08 dB/cm, while a 350um is 0.18 dB/cm, and 250um cross-section measures about 0.22dB/cm.
MWJ: Have you developed a set of design rules related to the process such as those that might dictate line dimensions or special line configurations such as a bend, tapered line or tee?
DS: Yes, we have put in place design rules for the PolyStrata process, very similar to the one you will find for IC or MEMS fabrication. The specificity of this 3D fabrication allows us to fabricate structures as small as 5μm in width while keeping micron tolerances across a 6” wafer and having structures as tall as 1mm in the Z direction. We generally use between 8-15 independent layers, so we can monolithically build rectacoax with 20μm line width on a 130μm pitch while connecting it to a 1 mm tall cavity resonator.
We have done tapering of the center conductor for bends, as well as for tee junctions in previous designs. This has been done for a few cases with good results.
MWJ: What types of microwave passive components such as a Lange coupler or waveguide resonator do you know of having been implemented in this process?
DS: We have fabricated and measured high Q cavity resonators, WG resonators, coupled-line directional couplers, branch line couplers, rat-race hybrids, filters, diplexers, baluns, and micro-connectors. Many of these have been built for frequencies ranging from 2 to 94 GHz. Some of the most exciting designs that are currently under development are the tunable components using integrated MEMS that are being developed in conjunction with Air Force Research Labs.
MWJ: Are there any design guides for engineers developing new passive components? Have any electrical models been developed?
DS: A significant portion of all our rectacoax structures is readily available in order to provide design guidelines, including design rules, electrical models, test data, and even yield analysis. Some basic circuit models have been developed, while others are currently under development. Currently, we are working on building a library of components including several standardized transmission lines that will make it easy to build circuits in programs like ADS or Ansoft designer. The lines themselves of course can be optimized for given applications. Currently we are expending a concerted effort to make the design process more accessible to practicing engineers.
MWJ: I read that the process can be fabricated on any flat substrate including a silicon wafer. What are some of the more common underlying substrates?
DS: Silicon wafers are mostly used because our process is currently implemented on an existing 6” wafer tool set. Furthermore, 6” wafers are readily available, flat, and cost effective. Depending on the application, we could fabricate this technology directly on Si, SiGe, GaAs wafers or perhaps ceramic carriers like AlN or SiC. Most of the applications we are currently investigating leverage our innovative release process which enables us to make free-standing components which can be flip-chip assembled into the customer’s existing circuits. Releasing the parts from the substrate has also enabled us to develop techniques for stacking and tiling the parts like LEGO® bricks into larger devices. They are surprisingly robust and we believe will usher in a new generation of modular plug-n-play microwave and millimeter wave capabilities.
MWJ: What should someone consider when choosing a substrate?
DS: The substrates need to come in 6” flat disks around 1 mm thick. If the application requires a different substrate, then the requirements should consider process handling, etch and release fabrication cycles, and final functionality such as thermal conductivity. We can fully support the customer with a full specification of process requirements.
MWJ: The PolyStrata process allows components to be integrated into a single module via the fabricated transmission line network. Could you describe some of the more challenging “systems” that have been created?
DS: All of the most challenging parts we have built and tested are for military applications with our defense contractors. The first big challenge arose from the DARPA 3D-MERFS Phase II program, which required a monolithic 16-element phased array using 12 strata with all the active circuits integrated directly into the PolyStrata substrate. Indeed this was extremely challenging, since we were still refining the fabrication process and material technology at the time. An even more ambitious phased array demonstrator is underway for the Phase II that involves stacking and tiling many functional levels into large panels. In yet another DARPA program we are working to create a 160 W ultra wide band GaN hybrid microwave amplifier where everything that would normally be on a MMIC is monolithically created in PolyStrata. PolyStrata is ideal for this application, being that the GaN real estate is too expensive for passives, and there are some superior wide-band power combining techniques you can do in PolyStrata. We are also working with AFRL on tunable components with MEMS actuators monolithically fabricated to create mm-wave tunable filters, phase-shifters, and VCOs. Most of these components are expected to yield record-setting performance; our vision is to unleash the creativity of the system architects and device designers with a PolyStrata tool-kit that will enable new capabilities, which cannot be achieved with any other technology.
MWJ: I get the sense that the process supports integrating different package types and components or being integrated itself into some hybrid module. Could you tell us more about what engineering teams have been able to construct to date?
DS: There are reasons for using both models: inserting PolyStrata devices into other systems and inserting other devices into PolyStrata systems. Some of our customers are interested in individual or simple components fabricated using the PolyStrata where these devices are then inserted into existing circuits. While other customers are motivated to integrate devices directly into a PolyStrata backplane to create a system, such as the DARPA amplifier program where GaN transistor amplifiers are integrated into PolyStrata. Both instances require similar fabrication process steps. As such, we have performed wire-bond attachment of devices to PolyStrata and, alternatively, PolyStrata structures can be wire-bond attached to external circuits. Notwithstanding we have developed an integrated thin-film solder deposition for the PolyStrata fabrication process, which allows flip-chip attachment to also work for either application (either attaching a PolyStrata structure into another system or mounting other devices into the PolyStrata environment).
MWJ: How are components such as an RFIC or surface mount components attached?
DS: We are creating sockets compatible with transistor and RFIC flip-chip attach, which support direct flip-chip attach and have integrated solders. For some conventional chips, we can develop a lead-frame like coaxial interconnect structure; the internal circuits can be connected with wedge-bonds.
MWJ: Any limitations worth mentioning?
DS: We are actively seeking both collaborators and customers for this new PolyStrata technology. For custom applications, our staff includes microwave design specialists who can understand a customers unique requirements for us to provide cost and delivery quotations. Our business model will support the provision of design rules for those customers wishing to design into our process. We will also consider rapid prototyping and accelerated developments as well. Finally, our product road maps will provide future plans for the PolyStrata component families.
MWJ: You have an extensive background in advanced package development. Is this new process unlike any other or is it an extension of technology that has been developing for some time?
DS: PolyStrata is the first technology that can create highly-complex millimeter-height structures from metal, air and dielectrics with micron level tolerances similar to semiconductor fabrication technologies. There has been work done with multilayer LIGA and metal MEMS for many years, but I am not aware of any that had the ability to integrate dielectrics as well, which continues to be their greatest limitation. Multilayer MEMS technologies are based on thin-film processes and still are mostly 2D and are very fragile. With our PolyStrata technology, we can release a small complex component from the substrate that you can pick up and not damage. It is more than a packaging technology; it can produce complete self-packaged complex RF, active, and electro-mechanical devices and subsystems. The genesis of this technology leverages the leading three competencies for Rohm and Haas electronic materials — photo resist, CMP, and plating. Rohm and Haas materials experts worked jointly with our micro-fabrication process engineers in each of these areas to develop our PolyStrata technology.
MWJ: Could you tell us a little bit about the company Rohm and Haas? You are a large company, but possibly not well known among microwave engineers. What is the make-up of the group dedicated to developing and supporting this process?
DS: Rohm and Haas is an $8 B, 18,000-person global company headquartered in Philadelphia, mostly focused in the materials business. The company supplies key materials at all levels of electronics manufacturing from the IC to the circuit board to the metal coatings on plastics on your cell phone. We are in paint, in flat panel display films, and water purification. You don’t see us that much because we provide enabling materials to the brands that you do know.
The group responsible for this technology represents an investment by Rohm and Haas in step-out growth. We are a group of engineers with a 6” wafer fab and back-end packaging capabilities, whose charter is to create disruptive platform technologies and grow through licensing and making devices for other companies. We also make other components as well, for example, we have a wafer-level hermetic micro-cavity package for optoelectronics, microwave devices, and MEMS called the Si-Pak packaging platform.
MWJ: Are there microwave design and modeling (simulation or measurement) engineers within the group?
DS: Yes, we have assembled a world-class team in microwave design, simulation, and test. Our CAD/CAE environment includes HFSS™and Ansoft Designer™, combined with Solidworks™ and Ansys™ modeling. One of our engineers labored his PhD thesis on PolyStrata, so indeed we do have a very strong core competency in this design space. We have in-house test and measurement capability to 40 GHz and can perform measurements up to 110 GHz at our university partner, Virginia Tech. Our collaboration extents to Zoya Popovic’s group at University of Colorado at Boulder for advanced simulation, modeling, and characterization.
MWJ: Who are some of the organizations that your team has worked with and how closely do you collaborate from a design and fabrication perspective?
DS: Most of the customer interactions occur under NDA. Those relationships that we are free to disclose include AFRL, BAE Systems, Virginia Tech, and University of Colorado at Boulder. BAE Systems has worked with us the longest and they do their own design work. For new customers and partners, we usually begin with a simple inquiry regarding application and specifications. From this information we can provide a full quotation for cost and delivery.
MWJ: How do you envision your future relationship with new customers? Will Rohm and Haas be like a foundry?
DS: Our business model will be like a MMIC or ASIC foundry with library support for customers to design into our process. In addition, we will offer design support services as well and are developing circuit level models to support our customers. We can perform full finite element analysis (FEA modeling) of physical structures to determine optimum design trade spaces. Therefore we are developing a comprehensive library of all the basic components, including RF parameter-based models and design rules for custom hardware.
MWJ: Regarding design support, your process can work directly from 3D CAD databases. Which ones do you support and do you have a library of component drawings already existing that customers can take advantage of?
DS: There are several file formats that we support in our CAD/CAE environment. We have worked with such 3-D file formats as SAT and STEP files, in addition to SolidWorks files, Autocad files or HFSS models. Other formats will be supported as we bring PolyStrata into full production mode. As we create masks for the fabrication, we will convert the 3-D models to 2-D file formats such as GDS and DXF. Our library will consist of a comprehensive listing of pre-designed components and transmission lines, along with physical design rules to allow customers to make their own structures.
MWJ: How long is a typical design cycle and how long is the fabrication cycle?
DS: Our design cycle will be very competitive with that of the microwave module and packaging industry. In other words, it should take no longer to design a PolyStrata module than a conventional microwave module like alumina or duroid. In fact, we believe that in the very near future, a design in PolyStrata will actually be quicker than the competition because the designer need not have to fix a problem associated with electrical isolation. Using PolyStrata, time to market will be accelerated compared to other technologies. Thus stated, obviously the length of the design cycle is quite dependent on the complexity of the component or circuit that is being designed. Some designs could be finished in less than a week, while more detailed system designs might take a month or more. Similarly, the amount of time necessary to complete a fabrication cycle currently can vary from four weeks to twelve weeks depending on the number of layers to be fabricated. As stated earlier, we can provide accelerated fab cycles if the customer requires it.
MWJ: Lots of MMIC foundries offer the benefits of fast turn around, so that designers can prototype quickly and hone their designs through a few iterations. Do you see a similar approach for people designing with the PolyStrata process?
DS: Yes, one of the key benefits of PolyStrata will be shorter development time due to projected “high probability” first-pass success with the resultant parts. This is mainly due to the simple fact that the electrical properties of bulk copper and coaxial lines that are 99 percent air-filled are well understood. Design for manufacturing and first pass success will be significantly enhanced as we can predict to a high degree of certainty the propagation effects of a practically pure TEM mode in precision hardware that is accurate to just a few microns. In contrast, conventional, microwave module technology is limited by modal effects caused by the waveguide, substrate, and dielectrics lack of precision tolerance. We believe that PolyStrata will improve the time to market for new products.