Juggling competing demands is a challenge as old as technology. When it comes to embedded systems, the story is often about how to maximize performance while minimizing power consumption. It’s a story that never gets old, and in a recent embedded.com article, Emily Newton has a few suggestions for best practices, design steps that will help “achieve good power management in embedded systems while ensuring they meet or exceed customers’ needs and expectations.”
First up: pay attention to device architectures, separating hardware from software when you’re plotting your architecture out, a separation that enables “independent testing and development, and forces the issue on focusing on hardware constraints. At this point in the design process, designers articulate what their power management targets are – and how they’re going to meet those targets. Awareness of what constraints you’re dealing with, and what resources are available to handle your constraints, “guides future choices.”
Having a detailed device architecture also provides the opportunity to define data inputs (information types) and outputs (data sinks); to specify the role subsystems will play; and to think through the protocols and communication approaches that will meet interface and component needs.
Once people create comprehensive device architectures, they can understand how certain choices could affect specific system parts or operations. Seeing that interdependence helps designers make the best decisions to optimize performance while saving power.
Newton has plenty of suggestions for other areas as well. One is to include in your design a block diagram illustrating “average power consumption and minimum energy used.” Understanding how programming relates to energy consumption “can show which parts of the design need further improvements and whether the selected battery for the system is sufficient to maintain reliable performance.”
Some programming choices that will reduce power consumption while maintaining performance are keeping the circuit board’s operating voltage as low as possible and selecting integrated circuits that align with the power-saving goals.
Another thing that should be looked at is turning the system, and its components, off when they’re not in use, which “will curb unnecessary power usage.” She also suggests utilizing operating system features that can keep power consumption down.
Newton also addresses some basics: making good choices when it comes to materials, and being very strategic when it comes to deciding the system features that are absolutely necessary. (As we all know, there’s always a features continuum that ranges from essential to “nice to have” to absolutely unnecessary.) Decisions on features will definitely impact power consumption and performance.
Newton stresses the importance of testing to make sure that power management and performance live up to expectations.
All tests should mimic real-life conditions as closely as possible. That’s the best way to check that the power management in embedded systems will provide customers with consistently excellent performance, providing the reliability they demand and expect…Plan your testing schedule to include embedded and software testing. Whereas the second type only involves software, the former encompasses all the system’s aspects.
The story wraps up with a reminder that there are always emerging technologies that engineers should be open to using. (An example she cites is machine learning.) She adds:
Some designers also find digital twins useful in their process, particularly when improving power management in embedded systems that are more complicated than most. The digital twin environment allows studying how different power management decisions affect performance before implementing these options in real life.
There’s nothing all that revelatory or novel in Newton’s article, but it’s still a story that, when it comes to designing embedded systems, never gets old.
Source of Image: Dreamstime