HiL Testing Accelerates the Move from Digital Concept to Hardware Reality
Yves Gerster from Speedgoat talks to PSD about how industry trends are leading to the growing importance of hardware-in-the-loop (HiL) testing.
PSD – What do you see as the main industry trends at the moment and how does HiL testing help meet those challenges?
YG - As designs become more complex, one industry trend is more virtual engineering, so more of the design is generated on the desktop. More teams are involved in each design, and they're spread across multiple geographical locations, where they develop different components or subsystems, which also adds to the complexity. Additionally, code is often AI-generated, and finally, safety standards are changing all the time to become more stringent. Together, those factors lead to what can be perceived as a black box design, built using advanced tools. That type of design can be hard to trust, especially in terms of proving that it fully meets the relevant standards.
HiL testing is vital for the transition from digital concept to hardware reality. As the trends mentioned above progress, and systems become get larger and include more components, there will be an even greater need for HiL testing, as these subsystems will all require end-to-end testing along with the final integrated system. This is where we see a growing need to gain trust in the final system, as you may have less understanding of the overall design itself. The strength of HiL testing is the ability to see all of the responses and how the design behaves with different phenomenon, such as transient behaviour. The ability to test it fully in the loop is very important, especially if you want to ensure multiple components work together and integrate correctly.
PSD - HiL is Speedgoat’s speciality. Can you tell me a bit about the company?
YG - Speedgoat build hardware test systems. The company was founded by former Mathworks employees, and Mathworks is a shareholder, so our hardware is very closely integrated with Mathworks software. In fact, in the hardware testing space, Speedgoat is the only company that can offer full integration with MathWorks through Simulink Real-Time. So, Mathworks has the simulation environment, while Speedgoat has the hardware. To test something effectively, you need to have both. The models which you use to to develop and simulate the design are also used directly in the test phase. Simulink Real-Time runs the test within the Mathworks environment. The models are automatically loaded and deployed on the Speedgoat hardware, and the close integration allows you to run all tests, while working in a familiar software ecosystem.
PSD - Can you tell me more about the hardware test systems?
YG - We call our products real-time testing machines. We offer around 200 different I/O protocols. Individual I/O modules can range from pretty simple to very complex. The more complex ones include FPGAs, which can be used to represent the missing part of the design if all components are not currently available. The components can be simulated first in the Simulink environment, and if the results are acceptable, the complete, tuned and tested model can be integrated into the remainder of the design. This provides complete flexibility by using only one model and one environment. If protocols are not available directly from an I/O module, programmable FPGA modules and signal conditioning modules that have a variety of capabilities, including shifting voltage levels and re-routing signals are available. So it is possible to construct the protocol from the ground up.
The I/O modules are normally integrated into a target computer to provide the different features that the test requires. There are several target computer designs, each with a variety of capabilities, for example the Performance real-time target machine has a motherboard that runs an RTOS, and also includes FPGAs to help with tasks such as customizing protocols or level shifting voltages, the Pulse system is intended for desktop testing of components, with the ability to easily swap out modules, and the Mobile target machine is compact and rugged for portability. For more complex testing, several target computers can be used together on a rack. Each target machine can replicate one or several components of the system, or provide the protocols and stimuli necessary for testing the system hardware. I/O modules and target machines are connected together using PCIe or dedicated FPGA links depending on the required speed. Other common protocols and low-level standards, for example I2C, are also supported. The system is built for flexibility, and to test the widest possible variety of applications.
You’ve spoken mainly about protocols and stimulus signals, but many applications require power delivered for testing?
Yes, one one side of the equation is that communication layer, and the other one would be the actual power layer. If power HiL testing is required, some of the I/O modules can deliver power for some lower power applications. We also have modules that can directly control motors. For higher power applications, the target machine and modules can easily interface with power electronics systems.
What do you see are the main advantages of HiL testing?
If you want to test something like a flight controller, it should be in a situation where it feels like it is actually in an aeroplane doing its mission, getting failures and having to deal with everything thrown at it. The correct stimuli are needed; the inputs, outputs, the protocols, the voltage levels. Speedgoat provides those in real time – sensor inputs and then control outputs, running in a loop. That is the strength of HiL. It's seeing the correct responses, and how the system behaves with phenomenon such as transients.
Another example would be a controller in a vehicle. You have all the CAN buses and different inputs, so the controller, or the monitor, would see the actual CAN messages coming in, with the correct delays. It's not just having the correct ones and zeros, but it's also about having delays, fault insertions, and all the timing issues, which can be controlled.
So the Speedgoat HiL hardware is mainly for testing the subsystems and complete systems, rather than boards and single components?
Our ultimate goal is to provide a way to allow the testing of subsystems, and then to integrate them and use the same environment for systems testing. If you only want to test some boards, you definitely can do, but it's not the main use case. We do have some academic customers which use products to test very small PCBs or single components.
