How do you optimize embedded system hardware?

Optimizing hardware in embedded systems involves choosing microcontrollers and components thoughtfully, considering environmental aspects. Prioritizing low power usage, reducing interference, and using low-power sleep modes are key for energy efficiency. Clearly defining component requirements ensures smooth compatibility and functionality in hardware design optimization

I would agree with some of the contributions, first we need to know what to optimise for. Some of the possibilities are: - Power Consumption (ie. remote sensing) - Component count and Price (ie. mass production) - Size of the PCB (ie. wearables) - Testability (ie. prototyping) These are only some of the optimising goals and we can have one of them or some combination. When we decide their priorities we can start with the component selection and hardware design.

Choosing the right components also includes putting environmental constraints into consideration. I worked with a team to prepare for a Tech Exposition some years back, and one of the major things we did to build our applications was to define the requirements with respect to the selected available components, in this case it helps to ensure that effective compatibility and functionality are intact.

A common mistake that can be made during the acquisition process is choosing a more expensive component than what the potential use case might require simply because someone might be interested in learning the tech instead of basing components on the overall goal of the project.

The top priority fact each embedded hardware designer must know before the process starts is the purpose of the device, the key features, and tasks the device will handle. Most embedded designs are already optimised and are far from being universal. The performance demands, as well as the final application tasks determine the device components. For example a need for a battery backup requests a completely different power supply structure than a power supply fed from an AC mains power source.

Power consumption is still a major challenge open for experts resolution. I come from the western part of Africa 🌍 where beyond the impact of high power 🔋 consumption on the hardware system, the cost of maintenance and even availability is a major challenge. To incorporate effective power consumption model in systems, there has to be a sync in the software and hardware technology. I believe having a system with low power consumption model must have this two (software and hardware) modelled to manage each stage of the process flows.

Yes, I agree. The code-related techniques are always crucial and are a must. The other focus should still be kept on the hardware part. All pull-up resistors, voltage dividers and voltage biasing components are small energy eaters. The trap is that those eaters are hidden. Once an engineer starts calculating, the hidden trap is usually revealed. For the power supplies, we must always make crucial calculations of the efficiency. Sometimes linear regulators perform better than switched-mode power supplies when talking about efficiency. It is all about the voltage drop, current consumption, quiescent current, etc. To calculate efficiency of the switched-mode power supply is not always simple but charts in datasheets always help.

It takes skills and experience to look at power saving at each stage of product design, component selection and implementing the functionality both in HW and in SW. Count trice, measure twice, implement once...

Async programming and hardware feature usage is the key to minimize latency. Ideally you like to offload logic to peripherals/FPGAs/etc., memory transfer to DMAs and keep the communication between different system components like codec/MCUs/etc. minimal This requires a deep understanding of the processes that are realized. By offloading and using DMAs you can put your MCU/DSP to sleep. That will save you some power or gives you more MIPS for other stuff. As well as decoupling data streams from events can speed up system reaction time/data throughput.

I will speak to continously system testing and setting a benchmark. Having effective monitoring tools can help address and monitor latency bottlenecks, beyond monitoring they give real-time data and are smart enough to keep engineers informed in case of any spikes. The evolution of AI should be our focus in minimising latency in systems.

This is all about code efficiency, using correct programming techniques and strategies. A programmer must consider the pros and cons of the used framework and especially its overheads. Sometimes, the framework has higher complexity than the running tasks. Everything can be for good but we have to engineer the most suitable solution for the job.

Many embedded designs are on low power budget. Thus, minimising data transfer, where it costs energy is critical. Latency is part of design.

The cost of neglecting security is more than the cost of security. Beyond handling company's sensitive data, security should be the other of the day attachments for any technical team. The rise in crime in the cyber space is alarming, and this should be a major reason why all organisations irrespective of size should ensure being cyber secured.

Security is beyond to just talking about confidentiality, in fact this is only one of the three basic information security criteria, here the other two: integrity and availability of the information (according to the ISO27001 standard). For availability, we can make the idea that it is necessary to consider some HW and SW aspects like robustness and liability, in what environment it would be going to deploy, etc., so that the system always guarantee availability when it is required or needed by the users. About the integrity of the information, you can have all secret keys you want, and a robust HW or SW design, but if you are storing or sharing corrupted data, you would be actually doing nothing. Lots of tests should be considered for it.

Utilizing established best practices while being mindful of the potential risks associated with your actions is a great first step to ensuring cybersecurity for yourself, your company, and those around you.

Security is an area that is often neglected. Especially the embedded designs that grew up from small DIY projects lack even the basic elements of secure functions.

It all depends on balancing the costs, risks, usage scenario. One thing is low power sensor in synoptic monitoring and ease of decoding data is priority. Another is where data is in critical hardware control systems, where cryptography and access control must be applied to allow no gap for intrusion.

Consider EMC compliance from the start. Don't wait until the design is finished and a test house fails the electronics because of EMC issue. Slow down signal rise times whenever you can.

The development of embedded systems over the past 10 years has brought a need for the tech community to look into the following areas for effective productivity. ■ Policy and regulations. ■ Risk Mitigation (in terms of environmental and climate change). ■ Availability and Durability. ■ Lifecycle Management.

Other points to consider: - price - market availability - country of origin - component life-time - availability of similar components that can be used as a substitution - important standard compatibility, for example AEC-Q100 standard for automotive semiconductors

Take the jobs you have the skill for and will (want to see done), and there are customers who need it and will pay for. The landscape is changing quick. Embedded designer should try to learn new trick every week.