Rosenheim Technical University of Applied Sciences Students Build Servo Drive Controllers Using Model-Based Design

In the mechatronic systems lab, Professor Rainer Hagl’s research team created a model to design and control various kinds of servo motors and their components. Created in Simulink, the model was primarily used to configure software code of motion controllers for different MCUs, but also enabled hardware implementation on FPGA or SoC platforms for high-dynamic systems.

Seeing an opportunity for students to learn ASIC development using this model, Professor Versen created the project, “Servo Drive Controller ASIC.” His goal was to show students that the same Simulink model could be used to build ASICs, enabling a higher level of integration in microelectronics.

From the inception of the project, however, it was clear that having students write VHDL or Verilog code was not an option. The faculty wanted to motivate students to work collaboratively using a familiar programming language.

TH Rosenheim has a campus-wide license for MATLAB and Simulink, which are used extensively across the school’s engineering programs. This meant that Professor Versen’s project team was free to decide what tools to use for ASIC design, verification, and code generation and could be sure that all students were familiar with this environment.

The team modeled the controller plant using MATLAB, Stateflow^®, and Simscape™. In Simulink, they developed a floating-point model of a three-cascaded motion controller. After verifying the controller performance with the plant model, they converted the controller model to fixed point to reduce resource usage and costs. Before generating code, they compared simulation results for verification using the built-in logging and recording functionality of MATLAB and Simulink, enabling live comparisons of current and recorded data. The simulation results of the optimized fixed-point model then served as the golden reference for subsequent stages.

The team then generated synthesizable HDL from the fixed-point version of the controller using HDL Coder™ and verified its behavior in cosimulations using HDL Verifier™ and Siemens^® EDA Questa^® ADMS. Getting identical results to the fixed-point model in Simulink, they were able to move to hardware, beginning with an FPGA implementation. This enabled identification of design defects that could only be revealed on an FPGA prototype running in the lab.

Based on this workflow, the team was confident that their HDL code would perform the same way on an ASIC platform. They implemented the HDL using Cadence^® Innovus™ and handed it off to the semiconductor manufacturer.

In Professor Versen’s course, “Digital Design,” the exam is based on this project. Students were given a Spartan^® Edge Accelerator board to use at home to gain hardware experience. They were able to collaborate on the project from home in a virtual lab environment, exchanging and versioning files using Git™ tools.

• Students motivated to collaborate on ASIC design. “Using MATLAB and Simulink—languages they already speak—it was easier for students to collaborate on large projects and learn all the complex steps and workflows in building servo drive controllers,” says Professor Versen.

• Verified design deployable to a choice of devices. “By using Model-Based Design, students only have to focus on the design and verification of the model one time,” says Professor Versen. “If they want to reuse it on a different type of device, they don’t have to redesign it.”
  
• Students and faculty were able to work on the project remotely. “Using Model-Based Design with MATLAB and Simulink, we created a platform where students could work together on designing and building servo drive controllers,” says Professor Versen.