Sideline the Thermal Challenges of Embedded Systems
Key Takeaways
-
The thermal challenges in embedded systems range from accelerated ageing, premature failures, drop-in circuit speed, and increase in leakage power consumption.
-
Activity migration is one thermal management technique in which the tasks are migrated to different processors to reduce the heat generated.
-
The power and thermal management strategies commonly employed in embedded systems are dynamic voltage scaling ❲DVS❳, dynamic frequency scaling ❲DFS❳, and dynamic voltage and frequency scaling ❲DVFS❳.
Embedded systems can introduce some thermal challenges.
The ‘think big’ paradigm can be unhelpful when selecting embedded systems for your mechanical or electrical control applications. The large CPU size of embedded systems can cause heating problems, since the bigger the CPU power, the higher the processor-generated heat. The over-specification of CPU-size will turn out to be a bane to embedded system applications.
The complexity of an embedded system calls into question the reliability of its application. The amount of heat developed in embedded systems is directly related to its complexity. Design engineers have made several design changes to tackle thermal challenges in embedded systems. Modifications such as avoiding over-specification, incorporating highly integrated components, and components specially designed for embedded applications have offered some solutions.
Thermal Challenges in Embedded Systems
Computer chip size reduction has typically increased energy consumption and temperature in embedded systems.
We have seen the growth of embedded system applications in fields such as medicine, automobile manufacturing, telecommunications, and aerospace, to name a few. In all these applications, we are looking for reliable functioning of embedded systems. However, thermal challenges limit the reliability and real-time throughput of embedded systems. Thermal problems in embedded systems start with reduced carrier mobility, increased driving currents, interconnect delays, and increased power consumption, effectively putting the signal integrity and reliability in doubt.
Accelerated Ageing and Premature Failures
The downscaling in VLSI technology is proclaimed to be one of the greatest innovations in electronics research and development. It is true when we consider the high integration, power density, and cost reduction in the chip manufacturing industry that resulted from downscaling. However, chip size reduction has substantially increased the energy consumption and the temperature in embedded systems. The consequences of downscaling are seen in the reliability, performance, and power efficiency of embedded systems.
The significant negative impact of temperature increase is the accelerated ageing of components present in embedded system circuits, which ultimately leads to premature failure of the whole circuit. Temperature-driven phenomena such as electromigration, stress migration, hot carrier injection, and insulator failure are also responsible for shortening the lifetime of components.
Apart from the average and maximum values of temperature, magnitude and frequency oscillation also adversely affect the life of embedded systems, which we call thermal cycling. The embedded system design should consider a thermal aware design and process scheduling in the targeted application to prevent thermal cycle issues.
Reduced Speed of Embedded System Operation
Before you buy a phone or computer, you probably read reviews on its performance, speed, or memory. It is always concerning to see comments that a device you are considering gets heated quickly, runs out of batteries, or takes a long time to load certain features. These are all drawbacks of an embedded system’s circuit speed when the temperature increases. The embedded system speed of operation is dependent on the clock frequency—which varies with change in voltage and temperature.
When you are utilizing an embedded system for a critical control system in an oil-refining plant, the circuit speed matters a lot. The predictability and timing of the control loop employing the embedded system can go out of sync if there are any speed violations resulting from high temperatures. The on-chip temperature rise alters the clock frequency, which in effect lowers the embedded circuit speed. The increase in interconnect delay also adds to the circuit speed reduction, and infringes on the signal integrity and system reliability. In critical industries, the unexpected delays in circuit timing can cost billions of dollars.
Increase in Leakage Power Consumption
Another primary concern in the portable electronics industry is battery life. Battery life can be extended by keeping the device cool, which is the principle behind battery save modes. As temperature increases, power consumption—especially the leakage power ❲compared to dynamic power❳—increases in chip-based embedded circuits. The increase in leakage power and temperature are mutual, and this will cause thermal runaways and permanent damage to embedded systems.
Achieve Thermal-Aware Embedded Systems
From the discussion so far, we can conclude that the key to thermal-aware embedded systems is reducing power consumption and clock speed. The proper scheduling of activities in embedded systems can also help in achieving a good temperature profile.
Activity Migration
In Multi-Processor System-on-Chip ❲MPSoC❳ embedded systems, allocating and scheduling tasks alternatively to various processors can achieve a thermal-aware operation. By establishing this activity migration, we are distributing the computational tasks to multiple processors, thereby preventing the creation of hotspots and thermal runaways in chip-based embedded systems. This optimized execution of tasks fulfill all the performance constraints, as well as lower power consumption. A broad view of the application platform is essential, rather than board viewpoint, for ideally migrating tasks.
Dynamic Thermal Management System
DVFS can improve the performance of a CPU.
Dynamic voltage scaling ❲DVS❳ and dynamic frequency scaling ❲DFS❳ are the two techniques implemented to attain dynamic thermal management strategy in embedded systems. As the voltage and frequency are scaled, power consumption is drastically reduced and eventually brings a balanced thermal profile in the embedded system.
In DVS, the voltage required for logic transitions is scaled, which causes low power consumption. DFS or ‘CPU throttling’, is the technique in which the processor is run at less than maximum clock frequency to conserve power. The demerits of DVS and DFS, such as performance degradation and run time violations, can be overcome by the introduction of dynamic voltage and frequency scaling ❲DVFS❳. In this powerful thermal management technique, both the voltage and frequency are dynamically controlled for power conservation and temperature maintenance.
Thermal aware system optimization is essential when it comes to overcoming the thermal challenges of embedded systems. The SoC embedded system should be designed to give the highest performance without getting thermally unstable. It is the developer’s understanding of the application platform and requirements that determine the thermal management techniques to be employed in embedded systems.