Page 40

EETE DEC 2013

DRIVES & CONTROLS The experimental setup used as background for this paper is based on an ARM CortexTM-M4 mixed-signal control processor from Analog Devices used with combined tools from IAR and Math- Works, to achieve complete implementation of the MBD platform. Each of the steps discussed above has a direct link to the available tools and to the overall implementation. As can be seen in figure 4, each tool-chain offers a range of value. In MBD, the designer must choose how to balance the use of these tool-chains with the overall value creation of the independent MBD platform. For the target platform a real-time development environment is now in place to model, simulate, evaluate, deploy and optimize overall system performance and capability. This is all based on MBD and the balanced selection of system parameters, resulting in best-in-class flexibility where specific optimization is needed. In this way a scalable model of a system has been realized which facilitates use and reuse of code, either based on existing legacy code or functions, or complete new building blocks based on standard C or graphical functionality (Simulink/MATLAB models aligned with the complete simulation and deployment phase as shown in figure 5). Not only from the software perspective is it possible to change the overall setup, but once the right device drivers for a system are developed, a designer can change system resources, hardware elements and overall applications software Fig. 6: Sketch of a “Drive System”. Fig. 7: Implementation and compilation. Fig. 8: Simulation and run-time data from a MBD build system (processor data vs. simulation). for the end-application or system. Also, by being able to control the real-time aspects of the overall system timing, optimization of system scheduling is possible directly through this environment. Taking a deeper look at the classic sketch of a drive system as in figure 6, one can now visualize the capability of this architecture. It is possible to optimize each of the elements in the “Drive” and focus effort on what element is most critical to the end system. This means that, for example, if protection functionality and scale is of key importance, then the focus can be placed on the mechanical system in combination with the electrical control and power system. A mix of simulated results and real-time data can be used for monitoring system behaviour, and together forms a “Live” optimization. On the other hand, if noise disturbances decrease the overall efficiency level of the system, measurements of these can be used in scalable filters and observers which minimize the hardware’s noise issues to give an optimal state. Once all factors have been modelled and gathered, the final step in the deployment phase can begin and a complete implementation phase on the target system is a reality. Through the MBD design flow, and via MathWorks and IAR, the implementation of an overall model and the compilation of code are now possible. Each of the stages or elements in the model of the “Drive System” is now represented by a MATLAB and Simulink model which is scaled at the right level for optimal design criteria. Each of the elements in the model is based on standard toolboxes and block sets from Math- Works and can be reused across any element in a particular design – see figure 7. These elements also represent the different domains of the drive system and can be finetuned fully to minimize the Model vs. Deployment error. Through real-time deployment methods and compilation in this mixed environment, it is also possible to combine existing C code written by hand with ARM Cortex M4 optimized C code generated by Embedded Coder, the production code generation tool for MATLAB and Simulink. This entire process allows users to reuse existing knowledge of motor control design at the right level. At this point the IAR Embedded Work Bench can take the generated code and compile the complete project for ARM Cortex M4, which closes the MBD implementation phase for this system. Since its inception, MBD has been questioned on its capability and functionality relative to traditional system development and on its efficient use of overall system resources. This is where strong efforts from component suppliers, simulation- and deployment-vendors, and tools complier providers have “merged” and results today are comparable to traditional deployment methods. Of course, any code development written for a real-time system can be created in an inefficient way, depending on the implementation method used. With MBD, profiling, cross-optimization options and strong advantages in safety-critical system development can be combined, so that code development overheads are minimized and highest performance results are achieved. MathWorks offers tool qualification for use of Embedded Coder with IEC 61508, ISO 26262, and related functional safety standards. This mix of capability is much more difficult to achieve in a standard design flow. In the example above a standard FOC model is implemented on Analog Devices ADSP-CM40x series. In this model, position and current loop feedback are executing in 15us, supporting real-time profiling of both the current scheme and debug facility. It also allows tracking functionality of the overall FOC scheme. Both MBD simulation results and real-time data can now be evaluated and compared with ideal system functionality in relation to the target specification, as shown in figure 8. This ultimately enables a designer to constantly improve system efficiency, functionality and performance, and to evaluate how a given element or component in the signal chain is performing against specification. 32 Electronic Engineering Times Europe December 2013 www.electronics-eetimes.com


EETE DEC 2013
To see the actual publication please follow the link above