Simulation and Modeling Aid Prototype Development
By Rick Nelson, Executive Editor, March 2013
Software vendors are offering simulation, modeling, and real-time-testing tools that serve applications ranging from electric motor control to software-defined radio (SDR). As products move from design through prototyping, the tools support software-in-the-loop (SiL), processor-in-the-loop, and hardware-in-the-loop (HiL) test.
Ken Karnofsky, senior strategist for signal-processing applications at MathWorks, cited the latest release of Simulink as offering a completely revamped user interface and user experience to simplify model creation, editing, and debugging as well as simulation control. A key feature, he said, is the ability to rewind, replay, and step through a simulation to pinpoint where a particular event or problem occurred. The new release generates diagrams that are smart about routing lines in presentable formats, helping customers create very large models with many blocks and navigate through them.
Smart Signal Routing in Simulink
Courtesy of MathWorks
Karnofsky said MathWorks also has introduced built-in support in Simulink for inexpensive and widely used Arduino, PandaBoard, LEGO MINDSTORMS NXT, and BeagleBoard hardware. This capability primarily targets the academic world, Karnofsky said, making project-based learning accessible to students and educational labs. “We think [the Simulink and low-cost hardware combination] will dramatically improve the learning experience for students who are doing simulation and want to run a real-time experiment in the lab,” he added.
Another effort, Karnofsky said, is to integrate MATLAB into the model-based design environment to support the development of algorithms or to automate test. And new capabilities in the Stateflow tool also simplify modeling, with the Stateflow Editor now unified with the Simulink Editor, with support for MATLAB as the action language for states and transitions. That function, Karnofsky said, used to be served by a hybrid C-like language.
MathWorks also has enhanced its Simscape and SimMechanics physical modeling and simulation tools as well as its code-generation tools. That’s in support of industrial customers who use the physical modeling tools to model their plant and want to link the code-generation capability to MathWorks’ xPC rapid prototyping target. “With code generation, the whole model can go onto the target environment and execute in real time,” he said, adding “That’s another major enhancement to the combination of physical modeling and rapid prototyping.”
Karnofsky identified two key use cases for SiL and HiL test. The first involves algorithm engineers doing rapid prototyping for applications including control systems or communications systems. The need here, he said, is for better integration with turnkey hardware and making the experience seamless for the customer. He said he also sees customers designing control systems that are processor-based interested in moving to higher speed I/O capability enabled by FPGA plug-in boards.
He said MathWorks works with its partner Speedgoat, which delivers that type of hardware, to serve such applications. “The idea is to have somebody doing simulation get to real time, take real measurements, or perform high-speed I/O in as turnkey a fashion as possible.” He said he also sees a trend toward using FPGAs for signal processing in board-level systems that might have a discrete processor and an FPGA or in a single-device SoC such as the Xilinx Zynq.
The other use case, Karnofsky said, centers on an embedded scenario in which the trend is to perform comprehensive verification. “We see a lot of interest among customers to do very extensive verification at the model level using some of our formal tools in Simulink Design Verifier to generate test cases and to do coverage analysis and requirements linking,” he said. “The goal is to do as much simulation as possible at the model level.” The role of SiL and processor-in-the-loop test, he continued, is to apply tests to code executing in the target environment and verify that the results are the same as what has already proven to be correct in simulation.
Karnofsky said MathWorks also has enhanced its APIs to improve automation and support for custom targets. In the past, MathWorks has supported a few such targets directly, but now with so many different targets available, he said, “We are focusing on having reference examples for enabling our partners to use those APIs so they can get that last mile down to the target. That allows us to focus on the overall verification and code-generation workflows and then work with partners to support them in developing the target-specific piece.”
As designs move from simulation to prototype, it becomes important to make physical measurements. MathWorks’ approach to supporting instrumentation comes in the form of the Instrument Control Toolbox, which connects MATLAB to instruments such as oscilloscopes, function generators, signal analyzers, and power supplies. The toolbox attaches via instrument drivers such as IVI or via SCPI commands over commonly used communications protocols such as GPIB, VISA, TCP/IP, and UDP. With the toolbox, users can generate data in MATLAB to send out to an instrument or read data into MATLAB for analysis and visualization. In addition, users can automate tests, verify hardware designs, and build test systems based on LXI, PXI, and AXIe standards. “We are fairly vendor-independent when it comes to instruments,” Karnofsky said. “Our approach is to focus on the standards.”
Software and Instruments
One company offering extensive software tools and instrumentation is National Instruments. Nick Keel, NI VeriStand product manager, said one recent innovation is a new version of NI VeriStand, which was released in February. “I work with what we coined real-time testing—an umbrella term where we include embedded software validation, model-in-the-loop, and SiL test as well as test-cell control and monitoring applications,” Keel said.
Said Keel of the new release, “We added a lot of features for both embedded software validation and test-cell control.” NI, he said, works with a variety of simulation and modeling partners. “Based on customer feedback, we add different modeling environments.”
Keel sees growing interest in motor control. “This year, one of the big trends we’ve been focusing on is power-electronics test. We added to the VeriStand framework support for electric motors for electric and hybrid vehicles. We have seen a big part of automakers’ development budgets going into testing these vehicles.”
In particular, Keel said, NI has been working with JMAG, whose software tools have become a standard for electric motor design and test in the Japanese automotive market. “We worked with JMAG to integrate its simulation models into our HiL platform,” he said. NI also partners with Xilinx, he said, and has combined the latter’s FPGAs and JMAG’s modeling and simulation technologies to develop what he called a “high-fidelity electric-motor simulation platform,” which can apply to applications areas beyond the automotive space. In addition, he said, NI now can integrate Dymola models from Dassault Systèmes and has been working closely with LMS in France.
Furthermore, Keel said, the new VeriStand release provides broader support for conditioned I/O and high-speed data acquisition. Closed-loop-control systems, he said, often essentially have a single data point in and a single data point out, but with the high-speed acquisition capability, “You can acquire whole waveforms—you can capture transients and do additional data analysis.”
Keel said his goal is to serve the entire embedded software validation process and break down the walls between design and test. “In reality, you may have design teams and test teams, but everyone’s a test engineer to some degree,” he said. “If you are developing models, you’re building prototypes, and you are testing those prototypes. Wouldn’t it be great if you had one platform to do testing at all the different phases?”
With such a platform, users could create one set of test patterns and profiles and, said Keel, “…use those same profiles and view apples-to-apples reports of how something performs in simulation and how it performs in the prototype phase or HiL phase.” That, he said, is the concept of test component reuse—the ability to use a test component such as a test profile or even report template across all the different phases of embedded software validation.
Arduino, BeagleBoard, and LEGO MINDSTORMS NXT Targets
Courtesy of MathWorks
Ease of Use vs. Extensibility
As for supporting instruments, Keel said, “We try to strike a balance between out-of-the-box ease of use and an extensible and open framework. The software framework is very configurable. Users select the I/O devices they want to use and choose models from the wide list of modeling partners we work with. Ninety percent of application requirements are fulfilled very easily by what comes out of the box.”
Nevertheless, he said, “It’s virtually impossible to create a configurable software environment that’s easy to use and flexible enough to give everything to everybody.” Consequently, NI provides support for sockets or plug-ins, which can be written in NI LabVIEW, C, or any .NET environment to finish the final 10% or so of the application. Customers can write their own plug-ins or employ third-party developers; customers also can share plug-ins via an online plug-in community.
Keel said the HiL test systems he sees are likely to be built on the PXI standard with instruments from NI or third parties opening up a range of applications areas: semiconductor test, RF and wireless test, and data acquisition, for example. “We get to ride the coattails of NI’s investment in new products because NI plays in so many different spaces,” he said. “We get access to a lot of instruments and lots of I/O devices that a traditional company wouldn’t have any interest in building.”
For example, he said, NI’s machine-vision products could serve in infotainment test. Similarly, customers could integrate NI’s RF instruments into an HiL test system. As part of a $1 billion plus company that invests 15% into R&D every year, Keel said, his group gets the benefits of NI’s innovations in RF, vision, and other applications spaces essentially for free. Similarly, he said, his group can access Xilinx’s technology because of the leverage NI has with the FPGA maker.
Keel noted that NI VeriStand doesn’t come out of the box with RF support. Customers who use it in RF applications have employed the plug-in approach. NI does address RF simulation through its AWR subsidiary.
Said Haydn Nelson, product marketing manager for RF and wireless at NI, “Traditionally, we’ve observed a substantial organizational divide between design and test groups developing wireless products. However, one of the ways we’re bridging the gap is by enabling design and test tools—and therefore data—to be shared between both groups.”
For example, he said, a new feature in AWR’s Visual System Simulator software is cosimulation capability with NI LabVIEW system design software. “This feature enables engineers to use measurement algorithms developed in LabVIEW inside their EDA tools. As a result, design and test engineers are better able to correlate measurement results between simulations and measured performance of a DUT.”
An RF/Microwave Focus
One company that offers extensive RF and microwave design software and test instrumentation is Agilent Technologies.
Martha Zemede, an applications expert in the Agilent Microwave and Communications Division, sounded a note similar to NI’s Keel with regard to design and test cooperation. “Collaboration between design and test is very important because it helps accelerate development and minimizes risk throughout the product development cycle.” And Andy Botka, vice president and general manager of the Agilent Technologies Microwave and Communications Division, said R&D customers’ emphasis on shortening design cycles and removing uncertainty is a big driver for simulation software tools from Agilent’s EEsof Division (see “Mobile Data Is Key Driver for Test Business”).
Zemede commented specifically with respect to multistandard radios (see “Hardware and Software Support Multiple RF Standards”). “The complexities of devices with multiple radio access technologies along with time-to-market pressures make it critical to minimize system integration risk to avoid costly design turns and product delivery delays,” she said.
She explained, “Simulation tools such as the Agilent EDA tools help minimize integration risk throughout the product development life cycle, beginning with the initial design and extending through R&D hardware testing.” RF/baseband integration presents a potential source of risk, she said. Simulating and verifying RF and baseband designs together can help mitigate the integration risk and avoid unexpected behavior late in the system testing phase.
“As the design cycle transitions to the hardware testing stage,” she said, “Agilent’s design software can be combined with test equipment to verify system-level performance. With this approach, some portions of the design are modeled in simulation, and other portions are represented as the hardware DUT. The simulated signals can be downloaded to an arbitrary waveform signal generator to turn simulated signals into ‘real-world’ physical test signals. The test signals then are used as stimuli for hardware DUTs during R&D testing.”
Subsequently, Agilent’s 89600 vector signal analysis (VSA) software can capture digital and RF DUT outputs using a variety of instruments (an Agilent X-Series signal analyzer for RF DUT output, for example, or logic analyzer for digital DUT outputs) and transfer the data into Agilent’s design software for simulation post-processing.
“The ability to integrate EDA tools with instruments to stimulate, measure, and emulate hardware subsystems extends the reach of simulation tools further into the design life cycle. The ability to use the 89600 VSA software for both design simulation and testing real hardware provides unique measurement-algorithm and user-interface consistencies across the entire product development cycle,” she concluded.
Motors, Vision, and SDR
Looking ahead, Karnofsky of MathWorks said several areas are generating a lot of interest: “Motor control is certainly a hot area—we are seeing a lot of interest there. Computer vision is another. SDR is a third.”
Karnofsky elaborated on the SDR. The military and public safety sectors could benefit from the flexibility of SDR in meeting multiple standards and achieving cost reduction. However, he said he is seeing SDR used in the lab as a way to put some form of instrumentation and prototyping capability in the hands of design engineers. The end result may not be an SDR but rather an aid to the designer and a useful extension of simulation capability.
The motor control, computer vision, and SDR areas, Karnofsky said, “…are in addition to our traditional embedded markets where we have a lot of depth—the safety-critical aerospace and automotive markets.” In general, he said, there is an upward trend toward increasing software flexibility to put more intelligence in a radio, a camera, or electronically controlled motor.
For More Information