How Design Software Changed the World, Part II
Before design software, microwave engineers spent their days with copper tape and soldering iron applying cut and try methods. Without software, the MMIC, the RFIC, the LTCC, the SoP and SiP would almost certainly not exist and we would not have today's smart phones or smart weapons.
"How Design Software Changed the World, Part I" (Microwave Journal, July 2009) looked at design software and microwave hardware from the early 1960s up through 1987. During this period, the US defense department needed to develop smarter systems to counter balance the superior size of the soviet military in Eastern Europe. The weapon systems envisioned required reliable, high performance RF/microwave electronics. This is when MIC and MMIC technologies came into favor and design began to rely on "home-grown" software programs.
The Department of Defense (DoD) helped fund software development that affected MMIC implementation, namely simulation, layout and work flows. The microwave and mm-wave monolithic integrated circuits (MIMIC) program promoted collaboration between defense contractors, software providers and branches of the armed services to further the adoption of software tools among engineers.
By 1988, the Reagan administration, which was responsible for the military build-up of the 1980s, was in its final year. Military spending, which peaked at 6.2 percent of GDP in 1986, was trimmed down to 5.8 percent and would continue to decline to the present day (roughly 3.0 percent). However, the MIMIC program was well-funded and entering its second year. Among the major commercial RF/microwave software vendors vying for contract money were EEsof, Compact and Hewlett Packard. Over the next decade, software would be influenced by the downsizing of defense spending, the emergence of the commercial mobile communications market, new technologies, start-ups, mergers, acquisitions and fierce competition among vendors.
Heading into the 1988 IEEE MTT-S Symposium in New York, EEsof unveiled a new system level simulator in the May issue of Microwave Journal. The product—OmniSys—allowed microwave system designers to predict the overall system performance (gain, noise, dynamic range, nonlinearities, spurious signals, distortion and more) from the performance data of the constituent components. The frequency-domain component data (scalar, S-parameter or pole-zero) could be synthesized from circuit simulations or test measurements. The initial release contained 50 internal component models and signal generators and was capable of simulating systems with up to 350 components, targeting radios, radars, EW systems, receivers/transmitters and feedback control loops.
The introduction was well-timed as many MMIC and MIC devices were becoming available for system integration. In the same May issue, an article by Bob Bierig of Raytheon and David McQuiddy of Texas Instruments entitled "Broadband MMICs for System Applications" described a series of MMIC-based building blocks for microwave systems targeting military applications. The authors' respective companies were partners in the MIMIC program. Noting that "one of the more notable features of MMIC technology is the relative ease with which components with broad operational bandwidths can be demonstrated," the article heralded a variety of successful MMIC designs including driver and power amps, mixers, phase shifters and switches; all front-end components for the defense departmentís next generation of military systems and undoubtedly made possible through simulation.
Early Design Environments
System simulation added to the number of discrete, dedicated software tools available, which now included linear and nonlinear simulators, measurement data acquisition, yield analysis, schematic capture and physical design layout. Around this time, software vendors began to market bundled solutions, followed shortly by the introduction of new design environments, which unified the user interface and integrated functionality. Pressure to develop powerful user interfaces was driven by the growing complexity of MMIC design, the introduction of schematic capture (over text-based netlist entry) by HP MDS and the need to support a variety of platforms from the IBM compatible PC to Sun, Apollo, HP and Digital Equipment Corp. UNIX workstations.
By early 1988, EEsof promoted the interoperability of its products: Libra, Touchstone, mw SPICE, MiCAD, ANACAT, E-Syn and Libraries across multiple computer platforms. While the company had initially gained market share through the PC, they now shifted focus to the same platforms as other CAD tools such as Mentor Graphics and Cadence, namely 32-bit workstations and mainframes. By April, EEsof introduced the MMIC Design Workstation and the EEsof Microwave/RF Design Workstation for board designs. These short-lived products were a stop-gap measure to counter the market gains made by the Design Capture System of HP's MDS and Compact's strong position among workstation users.
In November, six months after introducing the MMIC Design Workstation, EEsof introduced its new Microwave Design Environment ñ Academy, a front-end schematic and layout capture user interface for both Touchstone and Libra. Academy emphasized an integrated workflow, allowing engineers to create designs from either schematic or layout, moving seamlessly between editors.
Compact Software was also busy building up its product portfolio and bundling tools into solutions for MIC and MMIC design. Under Ulrich Rohde's leadership, Compact focused its attention on improving the capabilities (speed, accuracy and robustness) of its harmonic balance engine, Microwave Harmonica, as well towards the companyís noise analysis and model accuracy.
Along with its analysis tools—Super Compact (linear), Microwave Harmonica (nonlinear), Microwave SPICE and LINMIC (EM)—the company expanded its offering of synthesis tools that included Filter, PLL, and RF & Communications Design Kits along with the new Complex Match II, a unique impedance matching tool for designing feedback amplifiers, matching antenna impedances, transforming lumped designs into distributed ones and synthesizing commensurate distributed matching networks.
For design entry and layout, Compact partnered with third parties, releasing the clumsily named "Gas Station" (along with AutoArt) in March of 1988. This "mouse-driven, menu-oriented graphics system" generated layouts and geometrically related schematic displays from "hierarchically flat" Super Compact files. Unfortunately, this cobbled together user interface and layout tool was a lackluster competitor to the front-ends offered by HP and EEsof, and was a poor show of the solid engineering going into Compact's analysis capabilities. By December, Compact stopped actively promoting Gas Station and released the Microwave Design Workstation (MDW).
As described in a July 1989 MWJ product feature called "New Design Workstation Simplifies the Design Process," MDW combined Compact's nonlinear simulator, Microwave Harmonica with Design Framework—a schematic and layout tool from Cadence Design Systems—into a single CAD workstation. The combination of Compactís analysis and modeling capabilities with Cadence's powerful design automation made a formidable challenge to HP and EEsof's products.
The Compact/Cadence collaboration provided users with a number of powerful design features including forward and back annotation, open architecture (ability to create custom pull-down menus, define macros, etc.), design rule checking, hierarchical schematic, GDS II output for mask generation and a "zoom" feature. Compact focused on the analysis capabilities of Microwave Harmonica with new temperature-dependent models, improved noise performance predictions of active devices (MWJ cover story, December 1988) and a new harmonic balance algorithm resulting in a 300 time speed up in convergence.
In case any engineers were drawn into believing that these advances in simulation and design automation were going to free them from actual design work, future MWJ editor, Harlan Howe, brought optimization-happy designers back down to earth with his November 1988 special report, "Computers Don't Innovate…People Innovate!" Howe reminded readers that the human element was still very much needed in microwave circuit design and that ìa Calma tape full of cell designs is basically no more than the rack of plastic drawers with resistors, capacitors and inductors sitting against the back wall in every laboratory." Howe warned that ignoring MMIC process variations or the impact of temperature variations was a guaranteed recipe for failure.
By July of 1990, Compact was introducing its own system analysis tool, Microwave Success, supporting top-level block design capabilities, budget calculations, an unlimited number of cascade-able n-ports and unlimited number of sources including multiple carriers and modulations. A MWJ special report, also in the July issue, discussed the features of CAD/CAE tools on display at that year's IMS in Dallas, TX, noting that as Compact was introducing enhancements to new and upgraded elements, file syntax, time-domain steady-state analysis and graphics, EEsof was releasing Academy version 3.0 with a number of user interface and analysis enhancements.
While the development of an integrated design environment with schematic capture and layout took several twists and turns (MDW, Gas Station, EASI) from 1987 to 1990, Compact introduced a new design workstation called Serenade in August. Based on X-Windows, Motif and GKS, the product supported the use of remote terminals with networking capabilities that allowed the simulator engine, graphics and user-interface to run on different computers. The new "look and feel" of the user interface to Super Compact, Microwave Harmonica and the new Microwave Success was a big improvement to Compact Softwareís user interface. Though Compact now offered a more competitive PC-based CAD product, the complexity of MMIC design had shifted much of the engineering work back toward workstations.
In his March 1989 MWJ article, Carl Denig of M/A-COM presented a technique for designing microstrip interdigital filters, "applying an approximate method for modeling these structures in CAD programs such as Super Compact and Touchstone." Earlier approaches to modeling multiple-coupled sections using parallel sections of singly coupled lines ignored coupling between nonadjacent lines and resulted in errors that became more pronounced at lower substrate dielectric constants.
Denig provided an eight-step procedure to analyzing a multiple-coupled line filter that included determining an equation relating even- and odd-mode impedances using linear regression, calculating even- and odd-mode phase velocities and estimating parasitics to include in the CAD netlist. While the author achieved reasonable results, the approach was not for the weak of heart and clearly illustrated the need for a more accessible and generalized solution.
Ray Pengally (formerly with Compact Software) further represented this need in his 1990 MWJ article "CAD for MMIC Implementation." describing how GaAs foundries would fabricate and characterize (from measurement) a large array of passive structures such as a multi-turn spiral inductor, each with different physical dimensions. Working with CAD vendors, equivalent circuit models were fitted to measured S-parameters over a specified frequency range. It was often possible to fit simple polynomial expressions as a function of geometry to these equivalent circuits, relying on less-than-reliable interpolation for unmeasured components.
A more rigorous solution using electromagnetic simulation for passive structure characterization was of great interest to engineers and university researchers alike. Pengally described several programs including Linmic+, EMSym, Sonnet Software and Stingray (a quasi-static analysis based on finite element for calculating local capacitance matrices for multi-coupled lines, Apogee Software, Oakland, NJ), which could simulate structures in two dimensions "since considerable reductions in computing time can be achieved if planar metalized patterns are assumed where two-dimensional current distributions exist."
The evolution of EM simulation technology occurred over multiple decades, producing a variety of techniques. The basic FDTD space grid and time-stepping algorithm can be traced back to a seminal 1966 paper by Kane Yee in IEEE Transactions on Antennas and Propagation, while the "Finite-difference time-domain" (FDTD) descriptor supposedly originated with Allen Taflove in a 1980 paper in IEEE Transactions on Electromagnetic Compatibility. Today, approximately 30 commercial and university-developed FDTD software suites are available. Word of early commercial offerings began appearing more regularly in publications during the late 80s and early 90s.
In 1977, Weiland introduced Finite Integration Theory (FIT), also known as FI. In this technique, the integral form of Maxwell's equations is approximated by summations around small loops. This technique is different than FDTD in which the differential form of Maxwell's equations is represented as finite difference approximations. In 1983, at the Deutsches Elektronen Synchrotron (DESY) in Hamburg, Weiland set up an international collaboration in order to develop the software package MAFIA (Maxwell's equations Finite Integration Algorithm) for 3D EM and charged particle simulation. By 1992, Weiland founded Computer Simulation Technology (CST) to commercialize MAFIA and focus on the telecoms industry along with three managing directors: Dipl.-Ing. Michael Bartsch, Dr. Peter Thoma and Dr. Bernhard Wagner.
In 1996, CST decided to implement the perfect boundary approximation (PBA) into its simulation software. Unfortunately, MAFIA could not be easily adapted into a conformal code so Weiland reformulated FIT in terms of global quantities assigned to space objects, allowing matrix formulations that were valid for irregular and non-orthogonal grid systems. In addition, CST decided to build a new environment from scratch, taking advantage of the improved usability offered by Microsoft's new Windows OS. The result was CST's current flagship product CST MICROWAVE STUDIO (CST MWS), which was first released in 1998. By that time telecoms weredominating CST's business and CST MWS was aimed squarely at people designing antennas and connectors for mobile phones and base stations.
In 1988, Sullivan, et al. published the first 3-D FDTD model of sinusoidal steady-state electromagnetic wave absorption by a complete human body, while Zhang, et al.—the founder of Zeland Software (and IE3D)—introduced FDTD modeling of microstrips. In February of 1990, Microwave Journal featured HP's high frequency structure simulator (HFSS) on its cover, "Software Computes Maxwell's Equations." Promising to replace empirical design procedures, the software, which was jointly developed by Ansoft Corp., was able to accurately simulate the RF performance of a coaxial-to-waveguide adaptor on an HP 9000 model 835 workstation, taking less than one hour for a single frequency point.
Using a mathematical technique known as finite elements, the simulator provided key information, including multi-port S-parameters and color displays of electromagnetic field plots. Based largely on the work of Ansoft founder Zoltan Cendes and some of his graduate students from Carnegie Mellon University, the technique had been used in civil and mechanical engineering before Cendes and his team figured out a way to address the spurious signal modes that were problematic at high frequencies.
Ansoft also developed an adaptive mesh refinement algorithm, capable of generating and refining the mesh automatically. With adaptive meshing, engineers just had to accurately define the structure, its materials, boundary conditions and port excitations in order to obtain correct results, at least in theory. HP contributed by defining the scope of functionality, managing the product development, QA, testing and marketing.
While HFSS allowed microwave engineers to address 3D problems such as waveguide filters and antennas, planar geometries could be solved more rapidly with a 2D or 2.5D approach. The July 1990 MWJ report on CAD/CAE at IMS reported on planar EM technology from Sonnet software as did a May 1991 product feature, "EMSim 3.0" by EEsof. The Sonnet product "em" operated on a UNIX platform with a mouse-driven user interface. The designer provided the multi-point polygons of the layout; input the dielectric constants, port definitions, simulation frequencies and the grid size.
EEsof's EMSim applied a similar method-of-moments technique using a closed-form solution to Green's function to characterize planar microwave passive circuits. In the article, EM simulation of an edge-coupled filter was compared to results based on a Touchstone microstrip model and measurements of the fabricated filter. The agreement between the EM results and the measured filter were significantly better than the Touchstone response, which neglected inter-element coupling.
Within the year, Compact began offering its planar EM solution, Explorer v1.0. In 1992, HP introduced Momentum, originally developed by Belgian company, Alphabit, a spinoff from IMEC, Europeís largest independent research center in nano-electronics and nano-technology. Today, planar and 3D EM simulation provides such an improvement to circuit simulation accuracy that all major RF/microwave circuit simulation environments have some form of integrated EM technology.
Active Device Modeling
Active device modeling plays a critical role in accurate RF circuit simulation and thus HP introduced a new approach to transistor modeling based on its expertise in test and measurement. In September 1991, David Root of HP wrote the MWJ cover feature, "A Measurement-based FET Model Improves CAE Accuracy," describing a technique for improving large-signal simulations by explicitly constructing device nonlinearities from measured data. An automated procedure of collecting DC and S-parameter data was proposed as a replacement for conventional parameter extraction and physical or empirical model generation.
Nonlinear model functions were stored in tabular form as functions of two independent controlling terminal voltages. The lookup-table model "avoided the ad hoc procedure of optimizing equation coefficients of process specific models to fit measured data." They were also process technology independent, meant to alleviate the need for specialized active device modeling personnel. The HP IC-CAP model extraction software program provided the graphical interface and instrument drivers to generate the compiled table-based "Root" model for use with HP MDS. Root continued working with HP (and later Agilent) researchers on measurement-based models and nonlinear vector network analyzer measurements, resulting in today's X-parameters and NVNA technology. For more details, see the MWJ March 2010 cover feature, "Fundamentally Changing Nonlinear Microwave Design."
Compact Software's Scout™ was a PC and workstation-based active device extraction and large-signal modeling program with interactive modeling features. Users could modify parameter values and observe the effects on both DC and S-parameters. MMIC-focused modeling software from other companies included Optotek's Small and Large Signal Analysis (SALSA™), dedicated to MESFET and HEMT modeling, compatible with the company's simulation software MMICAD™. Optimization Systems Associates' (OSA) offered a nonlinear device-modeling program called HarPE™, which included popular FET models, a Gummel-Poon model for BJTs and HBTs, models for HEMTs, as well as user-defined models. In 1997, OSA was acquired by HP for its EM optimization technology to complement its Momentum and HFSS products.
Mergers and acquisitions are common in the software world where markets can only support one or two major vendors and smaller companies are eagerly acquired for their game-changing innovations. In 1992, the RF EDA market was fiercely competitive with three major players. That fall, EEsof launched a stealth marketing campaign for a new high-frequency analog design suite that the company claimed would "redefine the industry standard." Two months later, the company revealed EEsof Series IV, a new design environment that offered unprecedented ease-of-use, eight simulation engines, time, frequency electromagnetic simulation, statistical design, yield optimization, models, libraries and synthesis tools. Series IV was well-received by engineers and went on to have great success in the market despite a short-lived ad campaign that ended in December, when the independently-owned EEsof disappeared from the pages of Microwave Journal forever.
Something momentous was about to occur.
By 1993, EEsof CEO Chuck Abronson had been contemplating a bold strategic move. HP and EEsof were wrestling for the top position in the RF software market, each company realizing approximately $20 M in annual revenue. The new Series IV design environment took three years to develop and was designed to go up against Cadence, Mentor Graphics and Avanti in the emerging wireless chip set market. To compete, EEsof needed to dominate its current market so that its resources could re-focus on design tools for the RFICs being built for the next generation of handsets.
While the EEsof sales force was bullish on competing against HP, Bill Childs, EEsof's head of development, was less optimistic about the company's prospects of forcing HP out of the RF software business. Abronson had two options. He first contacted HP's executive leadership to gage its desire to sell the EDA group to EEsof. As a strategic part of its design through test solution, HP was not interested in leaving the software business. Therefore, Abronson proposed that HP acquire EEsof, a transaction that was publicly announced the following September. HP GM Jake Egbert led the new HP EEsof from 1993 until 1999; Bill Childs stayed another year through the transition while Chuck Abronson retired from EDA to focus on the California real-estate market.
In the 1990s, the 'second generation' (2G) of mobile phone systems emerged, primarily using the GSM standard. These 2G phone systems differed from the previous generation in their use of digital transmission instead of analog and also by the introduction of advanced and fast phone-to-network signaling. The rise in mobile phone usage as a result of 2G was explosive and would greatly impact the opportunities for design software as well as the required functionality.
In 1993, as HP-EEsof shifted its focus toward this new market, the new organization found itself serving roughly 80 percent of the traditional RF market with two similar products. Setting a course for new product development, HP executives identified two requirements they considered essential to making the next leap forward in simulation software. The first was the advancement of system simulation and support for the digital modulation that was driving future mobile phone technology. HP initially intended to address this first requirement with its acquisition of EEsof's Omnisys.
Coinciding with the introduction of 2G systems was the trend away from the larger "brick" phones toward tiny 100 to 200g hand-held devices. This drove the other requirement for RF circuit simulation, namely enhancing EM simulation to address the component miniaturization for smaller handheld devices. For this second requirement, HP had its planar EM tool, Momentum and an exclusive marketing agreement to sell HFSS under the HP name. Any new design environment would look to further integrate these simulators into the engineering workflow.
As HP was developing its new platform, the relationship with Ansoft was beginning to strain. In 1992, Ansoft introduced a new software product, Maxwell Spicelink, which targeted signal integrity applications. Spicelink essentially combined Ansoft's EM technology with SPICE for high-speed electronic circuit design. While HP had the exclusive rights to sell HFSS, a technical clause in the contract allowed Ansoft to sell Maxwell Spicelink and HFSS in a bundled a package called Eminence, which essentially allowed Ansoft to compete in the RF/microwave market as well. Meanwhile, HP was developing its own version of HFSS.
During the early 1990s, Ansoft grew steadily, as management prepared to take the company public. Revenues grew from $3.47 M in 1993 to $6.15 M in 1995. In 1996 Ansoft made an initial offering of stock to the public, netting some $12 M and allowing the company to acquire technologies that would further expand its product portfolio and market reach. It was also during this period that Ansoft and HP became competitors, as both companies vied for supremacy in the 3D electromagnetic modeling market.
Revenue from the HP agreement accounted for 12 and 13 percent of Ansoft revenue in fiscal 1997 and 1996, respectively. When the HP agreement expired in January 1997, it was not renewed, although HP maintained the non-exclusive rights to sell HFSS through January 1998. While Ansoft was directly marketing and selling HFSS version 5, HP introduced the HFSS product it had been developing internally, resulting in two products of the same name and plenty of confusion among engineers.
The New Kids on the Block
HP's acquisition of EEsof resulted in considerable overlap in simulation technologies and model sets. Rather than support two similar products or select one environment over the other, the HP management team directed the joint group of developers to start work on a totally new platform. The plan for this multi-year project was to combine the best technologies from both products while adding the new functions required for the emerging wireless technology. The product known internally as DE 1.0 (Design Environment 1.0) was re-named to Advanced Design System (ADS) prior to its introduction in June 1997.
ADS was promoted as the industry's first integrated, end-to-end signal path design solution for communications products, providing design technologies—from circuit and electromagnetic simulation to digital signal processing (DSP) synthesis and physical design—all in a single environment.
"The HP system provides new DSP design and synthesis software in addition to significant new design capabilities for microwave and radio-frequency integrated circuit (RFIC) design. The integration and co-simulation of RF and DSP analysis is unique in the EDA industry, as is the software's availability on both UNIX and PC platforms."
initial press release, June 2, 1997
ADS included two new DSP tools—DSP Designer and DSP Synthesis. The DSP Synthesis software targeted communications products with features for optimizing and implementing high-level DSP designs into application-specific integrated circuits (ASIC) and field programmable gate arrays (FPGA), including both behavioral and register transfer level (RTL) VHDL/verilog code generation, simulation and synthesis capability, outputting the hardware description language (HDL) in industry-standard formats for logic-synthesis tools.
DSP Designer merged software from HP research and technology from the University of California at Berkeley Ptolemy project. This simulation engine facilitated co-simulation of time, frequency and data flow technologies for mixed RF/analog/DSP communications projects. The system simulation had also been enhanced with harmonic balance co-simulation to allow full-budget simulations on any RF topology using patented new technologies, which included new high-frequency SPICE, harmonic balance (speed enhancements up to 100 times faster than comparable solutions with a reduction in memory usage by up to 15 times) and new Circuit Envelope simulation technologies.
HP was ambitious in both the scope and schedule set for ADS. According to Larry Lerner, Agilent EEsof R&D Manager, "The decision to build ADS was a risky one. It was by far the most complex and most expensive project ever attempted by the HP Test & Measurement Group. It was also a heavily constrained project with many must-have requirements." Many engineers expected the new product to provide the same functionality and feature sets found in both Series IV and MDS. Designers accustomed to a particular user-interface found the new environment non-intuitive and complex, especially with the new system, circuit and physical design capabilities rolled into one. In addition, migrating legacy designs into the new environment was problematic. Needless to say, ADS got off to a rocky start, with many customers delaying their upgrade from MDS or Series IV. Ultimately, the functionality goals initially set for ADS would prove to be well-aligned with the needs of RF/mixed-signal IC and multi-chip module design. ADS (as well as Ansoft, CST and AWR products) would also find a customer base among the emerging high-speed (giga-bit) design market.
While the ADS product was stabilizing and designers became familiar with its operation, HP competitors sensed a market opportunity and pounced. Since the introduction of SpiceLink, Ansoft viewed EM-simulation as a critical component to high speed/frequency circuit design. In April 1997, the company made its move to complement HFSS with a dedicated RF/microwave circuit-design tool, acquiring HP EEsof's largest competitor in the RF space—Compact Software. In August, the company also acquired its own planar EM solution with the purchase of Boulder Microwave Technologies Inc. for $1.5 M.
After competing for several years, HP EEsof left the 3D EM market with the sale of its HFSS product to Ansoft in 2001 (Agilent returned in 2008 with EMPro), allowing Ansoft to focus on its next generation architecture/environment while continuing to develop the Serenade product suite—Serenade (circuit), Symphony (system) and Ensemble (planar EM). The dual development efforts were a strain on Ansoftís limited resources and delayed the product release until 2003. The resulting platform, Ansoft Designer, raised the bar on EM/circuit integration, yet the window of opportunity for replacing ADS had passed.
By 2003, HP was no longer an Ansoft competitor. In November 1999, the T&M division—including the EDA group—had been spun off into an entirely new corporation called Agilent Technologies, executing the largest IPO in Silicon Valley history. In addition to Agilent, Ansoft also faced new competitors utilizing Microsoft's core set of application programming interfaces (API) and Windows operating system. With advances in programming, software companies were able to create RF design tools with the feel of standard Microsoft Office programs.
Joe Pekarek joined Hughes Aircraft in 1988 as an engineer. By the early 1990s, he was leading the development of a low-cost radar receiver chipset in Hughes's solid-state microwave division. As a hardware designer, he was an experienced software user, yet he was dissatisfied with the functionality from a user's perspective. In particular, he wanted tools that would help him to design high-frequency electronics at the system and circuit levels while laying out the physical components in order to reduce errors.
While at Hughes, Pekarek entered the doctoral program at UCLA to focus on harmonic balance circuit simulators, but switched subjects to numerical electromagnetics after his dissertation advisor left the university a few years into his studies. Undaunted, his prior EDA experience inspired him to change his dissertation topic toward developing a new environment explicitly for simulating high-frequency chip design. In 1994, with a year left to go on his doctoral studies, Pekarek left Hughes to start Applied Wave Research (AWR) and develop the company's first product, Microwave Office (MWO).
Pekarek and two other engineers worked on a shoestring budget to prepare MWO for market in 1998; credit-card debt being among their few sources of financing. MWO debuted at MTT-S IMS 1998 in Baltimore, with a limited set of features compared to the products from Ansoft and HP EEsof. Nonetheless, the graphic oriented user interface, friendly design methodology and real-time tuning (using fast, nonlinear Volterra Series analysis rather than harmonic balance) caught the eye of many attendees and put Microwave Office on the map.
The MWO interface was conceived for the "typical" RF engineerís workflow, placing emphasis on productivity and utilizing an underlying architecture that leveraged Win 32 APIs and object-oriented programming. Modeled after established design frameworks from EDA companies such as Cadence, the environment was constructed as a socket for supporting technologies, allowing AWR to partner with third-party technologists in order to supplement product functionality and offer designers a choice in solvers. This approach provided AWR with the time needed to enhance product functionality toward support of complex end-to-end communication systems being targeted by Agilent and Ansoft.
While MWO capability was being developed for the wireless chip set market, the product's ease of use and lower price tag made it a hit with the traditional RF board market. AWR uncovered or at least demonstrated the existence of a market for tools that were less complex, used primarily by individuals or small groups working for cost-sensitive companies. This non-RFIC market now included a sizable number of MMIC and MIC design houses.
This market was also fertile ground for Eagleware (known as Circuit Busters until 1991), which had quietly been selling its linear simulator (=SuperStar=) and EEpal tools since the company was founded by Randy Rhea in 1985. A product feature for EEpal appeared in the May 1991 issue of MWJ. This program was a combination of scientific calculator and microwave product/vendor selector guide. Users could solve polynomial equations, the Fourier coefficients of a square wave or look up the fax and phone numbers of microwave filter vendors.
In 1995, Todd Cutler joined Eagleware as CEO. Cutler had spent 20 years at HP, where he worked as a development engineer, led applications development in the European Microwave and Communications Group, and was a founding member of the HP microwave design software business, serving as Marketing Manager of HP-EEsof after the merger. At Eagleware, Cutler steered the development and marketing of the Genesys RF and microwave circuit and system design software platform in a direction similar to Agilent, Ansoft and AWR. The company began promoting its various synthesis tools (PLL, Filter, oscillator, Tline and matching), schematic and layout editors and linear simulator under the Genesys name. Over time, Genesys added a harmonic balance nonlinear simulator (Harbec) and a planar EM tool (EMPower).
In 2004, the company acquired Elanix Inc., an electronic-system level tool provider, thereby adding SystemVue signal processing design software for behavioral modeling of communication and DSP systems to their portfolio of high-frequency simulation, analysis and synthesis software. By the middle of the decade, software vendors offering an RF design solution with circuit, system, planar EM simulation, integrated into a single user environment included Agilent, Ansoft, AWR and Eagleware-Elanix.
In 2005, Cutler's career went full-circle as he helped broker the deal for Agilent EEsof's acquisition of Eagleware-Elanix, complementing their portfolio of RFIC products (ADS, Golden Gate, RFDE) with products intended for smaller enterprises, individual professionals and personal (contractor) use. Meanwhile, AWR has made several significant acquisitions including ICUCOM (system-level simulator in 2000), APLAC (harmonic balance simulation in 2005) and Simulation and Applied Research or STAAR (3D FEM in 2009). In 2008, Ansoft was acquired by Ansys. Continuing to sell its HFSS, Ansoft Designer/Nexxim, SIWave and Spicelink tools, the company recently announced further integration of its HFSS product within Ansoft Designer and HSPICE co-simulation.
The introduction of new simulation technology and integration with powerful design entry tools reflect the major advances in software development over the past two decades. The wireless revolution gave birth to several software start-ups and mergers. Through it all, the needs of the microwave circuit designer have been met with innovative software products supporting the ever-increasing complexity required for the next generation of wireless devices and defense systems.
This article looked at the larger RF circuit/system-level design software vendors as well as two innovators in EM technology (i.e. CST and Sonnet). A complete guide to all the software companies, researchers and products that have been developed for the wide range of microwave frequency design challenges could fill a book.