If your product gets hot, think about how to protect printed circuit boards from excessive thermal stress!
Printed Circuit board materials expand with increasing temperature. If there is a mismatch between this expansion of the circuit board and the board’s mounting structure, the board will warp or otherwise deform. Mechanical fasteners will break or loosen, parts will absorb moisture, and the product will fail. Understand board temperature, choose good fasteners, and avoid excessive temperature swings in order to avoid this problem.
Typical electronic component failures due to thermal expansion are ￼cracked capacitors, broken component leads, broken PC board traces, cracked solder joints, PC board delamination, electrical shorts, and plated via barrel-to-pad disconnection. If this sounds a lot like the effects of excessive vibration, there is a good reason for that!
In a robotics application, typical sources of heat are power supplies, motor drive circuits, motors, and batteries. One challenge is to keep excessive heat away from the batteries.
Consider a simple board mounted to a metal plate with four standoffs. The goal is to see what happens when there is too much power dissipated on the PC board, and then find the maximum power level that the design can handle.
Here is the example design:
|FR4 PC board||64 mm by 65 mm, 1.55 mm thick|
|Screws||M2.5 x 12mm, Stainless Steel|
|Nuts||M2.5 Stainless Steel|
|Aluminum plate||64 mm by 65 mm, 1.55 mm thick|
|Standoffs||5 mm height, 5 mm diameter, Al|
|Power dissipation||12.9 W (2 W/in²)|
The image is a 3D model of the board entered into Fusion 360. I used the thermal and thermal stress simulation capabilities to analyze the board.
What happens when too much power is dissipated? At 2 Watts per square inch with convection cooling, the board gets hot! My rule of thumb for convection cooling is that the temperature rises by 140°C/W-in². Counting both sides of the circuit board, the area is 12.9 in², and the expected rise is 70°C. Since the simulation sets the plate temperature at 50°C, I expect a 120°C board, and this is about what the simulation shows. Notice that the aluminum plate keeps the standoffs, screws, and small amount of the board area at a lower temperature. The heat conductivity will be improved by the copper layers on the board, which are not modeled in this simulation.
As the temperature rises, the PC board expands more quickly than the aluminum plate. The screws are tight in the simulation, and this causes the board to warp. Since the board expands faster than steel, the screws get tighter. If there is too much warping, the board will fail.
The simulator exaggerates the warping to show the effect. The next pictures shows a visualization of the actual amount of board deflection. The top of the rear screw head is the reference plane. The deflection is the amount of change relative to this reference.
When mechanical stress is too high, there are several possible outcomes. One is that the heads pop off the screws. A more likely possibility is that the PC board material will flow, and change shape to relieve the stress. After cooling, the permanent change in shape is called creep, and temporary changes in shape are reverted through elastic recovery. Creep takes time to happen, and it happens faster at higher temperatures. Elastic recovery time also increases with temperature.
When there is creep, as the board cools, the now-thinner board contracts, and the fasteners will loosen. After a few extreme temperature cycles, the loose screws weaken the mechanical product structure, possibly unseating connectors, cracking components, and creating air gaps in heat sink mounts. The result is worse heat flow, increased temperatures, increased sensitivity to vibration, and heat-related intermittent electronic failure.
To find the high-stress points in the simulation, use transparency for the smaller values. This shows a problem with the screws as the PC board material expands.
The graph above shows a relative measure of the amount of creep. The creep increases with pressure, time, and temperature. FR-4 ismuch more susceptible to creep damage at temperatures above 105°C. To provide a margin of safety, design for FR4 board temperatures at or below 95°C.
I use a rule of thumb: I can usually get away with 1 Watt per square inch of PC board area. In the next simulation, I reduced the power to this level and found that temperatures will be about 95°C. The Von Mises Stress in the PC board is below the FR-4 yield strength, and deflections are within the limits to prevent cracking of capacitors with package size 1206 and smaller.
FR-4 is a difficult material to model accurately. It is anisotropic, meaning that its properties change depending on the direction that you measure them. The fiberglass weave conducts heat better than the plastic, making the heat conduction directional. This detail is not modelled in Fusion 360. There is no substitute for real-world temperature testing!
Higher temperature substrates are available, and there are better PC board mounting systems available than just screws and standoffs. There are not hard-and-fast design rules for thermal design, but costs tend to go up for more exotic solutions. Flexible mounting schemes can provide relief for thermal stresses. Forced air cooling and heat sinks can also improve PC board power density.
Data for FR-4 material creep comes from: http://sottosgroup.beckman.illinois.edu/papers/nrs015.pdf
Fusion 360 information is at http://www.autodesk.com/products/fusion-360/overview
Belleville washer design calculations for Roger’s board material are at: https://www.rogerscorp.com/documents/635/acs/How-to-Avoid-Creep-on-Board-Assemblies.pdf
To learn about capacitor cracking, see: http://www.avx.com/docs/techinfo/CeramicCapacitors/cracks.pdf
For printed circuit board thermal calculations, see: https://www.tempoautomation.com/blog/smt-resistor-thermal-design-and-layout
To learn how Belleville washers also help with PC board vibration and this simulation, see: https://www.tempoautomation.com/blog.