In the previous article, we explained the systematic approach to EMC design. In this article, which is FC29, we will cover thermal design. We have calculated the losses that cause heat generation in previous power supply designs (FC21–FC23), but in this article, we will systematically explain how to dissipate that heat through thermal design at the PCB and package levels.
Why Thermal Design is Necessary
Heat generation on FC boards is primarily caused by MCUs, DC-DC converters, LDOs, and voltage regulators. As calculated in FC22, losses such as LDO voltage difference × current, and DC-DC switching losses manifest as heat.
Problems caused by heat occur in two stages. In the short term, if the operating temperature of the IC itself exceeds the maximum rated junction temperature, performance degradation (clock throttling, reduced accuracy) or immediate failure occurs. In the long term, repeated thermal cycling or continuous high-temperature operation causes fatigue failure of solder joints and shortens the lifespan of capacitors (there is a rule of thumb that the lifespan of electrolytic capacitors is halved for every 10°C increase in temperature).
The impact on sensor accuracy is also important. As explained in the previous series (IMU, barometer), temperature change is a major factor in sensor bias error. When heat from heat-generating components is transferred to the IMU or barometer, bias drift occurs due to temperature gradients. Thermal design, like EMI countermeasures, is a foundational technology directly linked to sensor accuracy.
Three Mechanisms of Heat Transfer
As a basis for thermal design, we will organize the three mechanisms by which heat moves.
Conduction is heat transfer within a solid. Fourier's Law:
Q = k × A × ΔT / d
k is thermal conductivity, A is cross-sectional area, ΔT is temperature difference, and d is thickness. Heat transfer through PCB copper foil and vias falls under this category.
Convection is heat transfer through a fluid (air). There is natural convection (flow due to air buoyancy) and forced convection (forced airflow by a fan). Most FCs rely only on natural convection, but industrial high-heat units may employ forced air cooling.
Radiation is heat transfer as electromagnetic waves. It follows the Stefan-Boltzmann Law and is proportional to the fourth power of temperature. In the normal FC operating temperature range (up to 100°C), the contribution of radiation is small compared to conduction and convection, and it is not the main focus of design.
Thermal design in FC design mainly targets two mechanisms: conduction (IC → PCB copper foil → external) and convection (PCB surface → ambient air).
Concept of Thermal Resistance
The most important concept in thermal design is thermal resistance. It can be handled in a framework similar to Ohm's Law in electrical circuits.
Thermal resistance θ has the relationship ΔT = Q × θ (where Q is heat flow, i.e., power consumption). It has the same structure as voltage = current × resistance in electrical circuits, with the correspondence of temperature difference = heat flow × thermal resistance.
Junction-to-case thermal resistance θjc is the thermal resistance from the IC die (junction) to the package surface (case). It is a major parameter listed in the datasheet.
Case-to-board thermal resistance θcb is the thermal resistance from the package to the PCB, which depends on the solder joint and the thermal path of the package.
Junction-to-ambient thermal resistance θja is the total thermal resistance from the die to the surrounding air, equivalent to the sum of θjc + θcb + θba (board-to-ambient). It is often listed directly as θja in datasheets, but since it strongly depends on PCB design (especially copper foil area), it should be treated as a reference value.
Calculation of series thermal circuits is performed in the same way as series resistance in electrical circuits:
ΔT_total = Q × (θjc + θcb + θba)
If there are multiple heat dissipation paths, calculate them as parallel thermal resistances.
Calculation of Heat Generation—Comprehensive Evaluation of Power Systems
Integrate the heat generation values calculated individually in FC21 to FC23 for the entire system.
The heat generation of a DC-DC converter is calculated from its efficiency (typically 85–95%):
P_DC-DC = Pout × (1/η - 1)
For example, if Pout=5V×2A=10W and η=90%:
P_DC-DC = 10 × (1/0.9-1) = 1.11W
This is distributed between the DC-DC IC and the inductor and appears as heat. The heat generation of the IC itself must be estimated individually from the efficiency curve in the datasheet.
The heat generation of an LDO is calculated using the formula (Vin-Vout) × Iout explained in FC22. For a GNSS LDO (5V→3.3V, 200mA):
P_LDO = (5-3.3) × 0.2 = 0.34W
The heat generation of an MCU depends on the operating clock, voltage, and activity. The power consumption of the STM32H7 is listed in the Power Consumption section of the datasheet, providing current values for different operating modes (Run, Sleep, Stop, etc.) and clock frequencies; this is calculated from the consumption of VDD (3.3V) and VCORE (1.2V) respectively. A few hundred mW to 1W is a typical value during full operation at 480MHz.
For sensor systems, the IMU, GNSS, compass, and barometer each have a power consumption of several tens of mW. While small individually, the accumulated heat can create local hotspots when placed in close proximity.
When the total heat generation is accumulated, a total heat output of about 2–4W is common for typical industrial FCs. This must be dissipated within the limited area of the FC board (usually several cm² to over ten cm²).
Thermal Design Using PCB Copper Foil
The primary heat dissipation path for an FC is the PCB copper foil.
The thermal conductivity of copper foil is very high at 385W/(m·K), and it functions as an effective heat dissipation material if it has an appropriate area. The basic design is to place a wide copper foil (GND plane or dedicated thermal diffusion pad) directly under the heat-generating IC.
A thermal diffusion via array is a configuration of multiple vias used to transfer heat from the exposed pad on the back of a surface-mount IC (many DC-DC and LDOs use this package type) to the inner and back layers of copper foil. By arranging multiple vias (about 9–16) with a diameter of approximately 0.3mm in a grid pattern, the thermal resistance of the vias themselves is reduced through parallelization.
Via filling is a process of filling the inside of a via with conductive or thermally conductive material, used for high-heat components that require higher heat dissipation performance. In standard FC designs, standard (hollow) vias are often sufficient, but it is worth considering for high-current DC-DCs of 3A or more.
Utilizing inner layer copper foil in multilayer boards is a design that actively uses the inner GND and power planes as thermal diffusion layers in boards with 4 or more layers. Not only the surface copper foil but also the inner planes play a role in diffusing heat laterally.
As a general guideline for heat dissipation pad area, approximately 1000mm² of copper foil (1oz thickness) per 1W of heat generation is considered a standard for natural air cooling, but in practice, it is determined by the balance between the layout constraints of the entire PCB and other design requirements (EMI, wiring).
Utilization of Thermal Simulation
Simulation tools are effective for verifying complex thermal designs.
Preliminary evaluation using simple calculations involves estimating the rough temperature rise from the datasheet values and the approximate thermal resistance of the PCB copper foil using the aforementioned series and parallel thermal resistance calculations.
Finite Element Method (FEM) simulation provides a more accurate evaluation, using tools such as Ansys Icepak, SolidWorks Flow Simulation, or free tools (such as those based on OpenFOAM) that integrate with KiCad, depending on the developer. It visualizes the temperature distribution by inputting the actual PCB copper pattern, component placement, and surrounding environment (convection conditions).
The practical advantage of simulation is the ability to identify problem areas (hot spots) before prototyping, allowing for copper pattern modifications and via placement optimization at the design stage. Countermeasures can be implemented at a much lower cost than modifications after prototyping.
While simulation is a standard process in industrial and large-scale FC development, in small-scale development, a combination of simple calculations and actual measurements is often sufficient, and the decision is made based on the balance of cost and risk.
Thermal Design Verification through Actual Measurement
We will organize methods for verifying the effectiveness of designed thermal countermeasures on actual hardware.
Measurement using a thermographic camera is a technique mentioned in the previous series regarding EMI verification, but it is the most direct evaluation method for thermal design verification. By capturing the temperature distribution of the FC board under maximum load conditions (all sensors operating, high-speed MCU operation, maximum current), you can confirm consistency with design predictions.
Point measurement using thermocouples is cheaper than thermography and is suitable for continuously monitoring the case temperature of specific ICs. It records temperature transitions during long-duration flights or in high-temperature environments (outdoors in summer) to confirm the temperature in a steady state.
Junction temperature estimation is calculated backward from the measured case temperature and the thermal resistance θjc:
Tj = Tc + Q × θjc
Using the measured case temperature Tc, the datasheet θjc, and the calculated heat generation Q, estimate the junction temperature and check the margin against the datasheet's maximum rating (usually 125°C).
Checking environmental temperature dependence involves testing (such as operation tests in a constant temperature chamber) assuming high-temperature environments in summer (ambient temperature of 40°C or higher) to evaluate the temperature margin under worst-case conditions. This verification is particularly important because industrial drones are expected to operate in high-temperature outdoor environments.
Placement Strategy for Heat-Generating Components
We will organize considerations for thermal design in the floor plan of the entire FC board.
Separating heat sources from sensors is a principle that has been repeatedly emphasized in the previous series (IMU, barometer). Physically separate temperature-sensitive IMUs and barometers from heat-generating components such as DC-DC converters, MCUs, and LDOs. This is a design decision in the same direction as ensuring distance from EMI noise sources, and in many cases, EMI countermeasures and thermal countermeasures can be achieved simultaneously with the same layout changes.
A floor plan that considers heat flow involves placing heat-generating components at the edges of the board or in areas with good ventilation, and planning the layout so that heat does not conduct toward the sensor area.
Distributed placement of multiple heat sources avoids the formation of localized high-temperature hot spots by spreading multiple heat sources such as DC-DC converters, LDOs, and MCUs across the board. Concentrating them in one place can cause mutual heat to overlap, leading to a temperature rise greater than expected.
Coordination with enclosure design is essential for comprehensive thermal design, which must include not only the FC board itself but also the ventilation of the final enclosure (case), the presence or absence of heat dissipation fins, and the enclosure material (whether it is a thermally conductive aluminum enclosure or a thermally insulating plastic enclosure).
Additional Measures for High-Temperature Environments
We will outline additional measures for cases where standard natural air cooling is insufficient.
Adding a heat sink is a measure to attach a small heat sink to DC-DC ICs or MCUs that generate significant heat. Due to space constraints on the FC board, large heat sinks are often difficult to use, so small stamped heat sinks (a few mm square) are used.
Thermal pads (thermal conductive sheets) are a method of adding a heat conduction path to the enclosure, in addition to the PCB copper foil, by sandwiching them between the IC surface and the inner wall of the enclosure. This is sometimes used in industrial sealed enclosure FCs.
Forced air cooling (fans) is the most effective, but it involves trade-offs in noise, power consumption, reliability (fan failure), and weight increase, making it a rare choice for drone FCs. While it may be adopted for high-heat companion computers (such as Jetson) in large industrial machines, its application to the FC itself is limited.
Implementing thermal throttling is a software-side measure that involves monitoring the MCU's internal temperature sensor and using protection logic to lower the clock speed or temporarily pause non-critical processing when a certain temperature is exceeded. While this type of protection mechanism is limited in ArduPilot by default, it is worth considering in custom firmware development.
Summary
Thermal design is a process that begins with calculating the heat generation of DC-DC converters, LDOs, and MCUs, estimating the junction temperature using series and parallel thermal resistance models, and designing heat dissipation paths using PCB copper foil and via arrays. The physical separation of heat-generating components and sensors is a design principle that aligns with EMI countermeasures, and in many cases, the same layout changes can address both issues. The validity of the design is verified through a combination of thermal simulation and actual measurements (thermography, thermocouples), and for industrial applications where operation in high-temperature environments is expected, coordination with heat sinks and enclosure design is also necessary. In the next installment, FC30, we will explain mistakes that cannot be prevented by DRC alone—the limits of automated design rule checks and areas that require human review.