The goal of this series is to help you better understand what are the main aspects to take into account to perform a good thermal design. We will cover the basics you need to know about heat transfer, applied to electronics and embedded systems, while avoiding as much as possible the need for complex math or expensive simulation tools. The objective is not to be rigorous, but to help get you on track and provide you with the basic knowledge and tools so you can start applying these concepts in your own designs.
Relevance of thermal design
Thermal design is becoming more and more important in electronics and embedded systems because it is increasingly affecting the performance and reliability of these devices.
-
Regarding performance, modern processors can scale their operating voltage and frequency depending on the temperature. These processors are designed in such a way that even the same processor model will perform better or worse in different devices or scenarios depending on the heat that can be dissipated at any given time. In other words, the amount of heat that our system is able to dissipate will be one of the key factors limiting the performance of our device.
-
Regarding reliability, it is necessary to ensure that these devices will operate reliably and safely throughout the entire life of the product, avoiding malfunctions or even, in extreme cases, smoke or fire. This will require ensuring that all components in the system will always be within the absolute maximum temperature limits for the specified operating temperature range of the device.
Therefore, it is essential to perform a proper thermal design from the initial design phase. If the thermal design is not performed and no countermasures against overheating are taken from the beginning, issues might be detected in later phases of the testing or manufacturing processes when they would be way more time-consuming and expensive to solve.
A bit of history
In the past, thermal design for embedded systems was not as critical as it is today. Printed circuit boards and components were much larger. Each component was usually capable of dissipating its own heat to the air by convection, thanks to the huge surface of the package. PCBs were also not intended to conduct heat. Heat conduction from the components to the PCB was not great given the through hole packages with their long and spaced leads. The low copper layer count of the PCBs didn’t help to conduct heat either.
Due to these reasons, heat generated in one component barely affected the temperature of the others, so it was common to consider that every component was capable of dissipating its own heat, without further considerations. Only when a specific component was particularly critical (for example, a powerful microprocessor or a linear regulator), some calculations were made to check if that particular component needed some kind of additional dissipation, but without considering the rest of components at all.
The motherboard of the NES is an example of such a design. The entire system can draw up to around 9W of power, which is quite low given the relatively large dimensions of the system. Only around 5W are actually used to power all the circuitry in the motherboard and in the cartridge. This power is evenly distributed among all the chips, there are no chips that draw much more power than others. All of these chips are also large enough to dissipate its own heat by convection, without any special consideration.
The remaining 4W of power are wasted in the input rectifier and especially in the inefficient 7805 regulator that powers the entire system. Given the high power to dissipate in this single component, it needs an additional heatsink that can be seen in the right picture. This entire metallic subassembly is used as the heatsink for the regulator and also as an electromagnetic shielding for the analog video circuitry.
The current approach
Nowadays, electronic components are getting smaller and more powerful. PCBs are also getting denser, accomodating more components and functionality in a smaller area. Some of these components can generate a lot of heat, and their packages are not large enough to dissipate this heat by themselves. Modern surface mount packages are designed to transfer this excess heat to the PCB, which now acts as a heatsink. The dedicated thermal pad included in some packages helps in this regard. Current PCBs also have more copper layers so they are better heat conductors.
The heat generated in a component will then be conducted through the PCB and eventually dissipated by convection making use of the entire PCB area. The temperature of each component cannot be independently considered anymore: heat generated in one component will greatly affect the temperature of others, even if they are located at opposite ends of the PCB. So, a better understanding of how heat is transferred within the PCB and to the ambient is now required to achieve proper results.
An example of this more modern approach is the Switch motherboard. The Switch can draw around 11W when gaming, in docked mode and fully charged. Even though this power consumption is quite low by today’s standards, the PCB dimensions are rather small so it is consuming many times the power per PCB area of its old predecessor. Moreover, this power consumption is not evenly distributed, but most of the heat is generated in the SoC. This means that there is a lot of heat to dissipate coming from a single, small source. In this case, given also the compactness of the device, the PCB alone is not sufficient to dissipate all the heat, so a more complex dissipation system by forced convection is necessary.
When thermal design is relevant
A good thermal study of your design is always recommended, but depending on the characteristics of your project it may be more or less important to do so. The key aspect is to consider the power/area ratio. For example, 1W of power is not a problem if it is dissipated evenly throughout the PCB, but it can be a problem if it is dissipated in a single small component. For the most common cases, some guidelines are:
- If your whole PCB draws less than 0.5W in total, there is usually no need to worry about it. This is very common in a lot of cases, for example in designs based on low power microcontrollers and sensors.
- If you have more than 0.5W in a single component, or around 2-3W in the whole PCB, a more thorough analysis starts becoming necessary. It is the case when you are dealing with powerful microprocessors, high currents through regulators or switches, or components such as motor drivers or powerful LEDs. In these cases, it is usually enough to use the PCB itself as a heatsink and to dissipate the heat through natural convection and radiation, if a good design is made.
- If your PCB draws around 5-10W or more, then a proper thermal analysis is essential to avoid overheating issues. Depending on the PCB dimensions, in these power ranges the PCB area may not be enough to dissipate the heat by natural convection, and the use of heatsinks and forced convection becomes necessary.
Sometimes, thermal aspects must be considered even if the power dissipation is very low. For example, if your product contains an ambient temperature sensor, you may want to isolate it as much as possible from other heat sources, to avoid disturbing the readings.
The intended application of the product also plays an important role. For example, for an indoor consumer electronics application, the maximum ambient temperature could be around 35°C, so you have plenty of margin to heat up the components without damaging them. However, in other applications such as industrial or automotive, the ambient temperature could be 85ºC or even 125ºC, so the margin is very low.
References
- ROHM Semiconductor. Thermal Design (Basic) Application Note: What Is Thermal Design?
- José F. Toledo Alarcón. Diseño térmico en equipos electrónicos: una introducción.