When you work out, you get warm. When it is hot out to begin with, and you are going as hard as you can, you get very warm. The same thing happens to the machines built for the heavy lifting. Heat breaks down lubricants, ages gaskets and shortens the lifespan of components. Robots are expected to pay back their investors with constant uptime.
A single failure of one of these machines can bring an entire assembly line to a halt. At the printed circuit board level, we designers need to provide the most robust solutions to keep the wheels turning under the autonomous taxi or on the factory floor.
“Cooling takes many strategies, and one of the main ones is derating the components.”
The most effective thing we can do is to spread out the components. Each one has a thermal signature that creates the warmer colors on the infrared image where heat becomes visible. Thicker boards with heavier copper will go a long way towards heat sinking. What if we don't have room to spread out or we have circuits that are too dense for thick conductors? As the current density increases, the temperature rise follows. To manage this puzzle, we have to understand the paths that the thermal energy will take.
Image credit: Imaging1.com
Starting Off at the Device Level
Drilling down to the device level, we find two basic approaches. The first one is called Wire Bond. The bottom of the die is solid metal and is attached to the substrate with a thermally conductive epoxy. A robotic arm picks up the die and places it into the die cavity and then scrubs it back and forth several times to help set the epoxy compound.
Another machine has a spool of bare wire, aluminum or gold, and it reels it out through a special head that presses the wire onto an opening in the passivation coating of the die while applying heat. This joins the bond wire to the die in one of two ways.
Ball bonding fuses the tip of the wire and launches upward creating a higher looping wire bond cage but allows a wider launch angle. Wedge bonding presses the side of the wire onto the die getting a lower profile cage but with a narrow take-off angle to the die. Either way, the wire is reeled out, and the signal is completed with an attachment to the wire bond pad on the substrate. This method provides a wide conductive path for thermal dissipation from the die (or dice) to the substrate in much the same way as the thermal pad of the package interacts with the center pad of the PCB footprint. It is a straight shot to the ground plane(s).
Image credit: Amkor
Flip Chips, on the other hand, are more like a BGA pattern in miniature. The substrate acts as a go-between that increases the ball pitch and helps account for the coefficient of thermal expansion (CTE) mismatch between the die and the board. The under-fill helps prevent the two pieces from growing and shrinking at their natural rates which keeps them at a happy medium while also enhancing the thermal conduction path.
The Bigger Picture
Image credit: Analog.com
In either case, managing the overall system requires an understanding of the place(s) where the stress is the greatest. The junction between the ‘silicon’ and the lead frame is the hot spot. The above image lists some common thermal relationships. The gist of all of this is that as the environment warms up, the remaining headroom decreases accordingly. Some regulators use a thermal throttle such that power decreases as the temperature approaches the upper limit of safe operation.
Some of the heat will find its way out through the top of the device through radiation and convection through the air. Venting the enclosure to allow hot air to escape into the room will increase the effectivity of this path. The orientation of the board or boards within the housing will have some bearing on the natural air flow. Warm air rises, so the electronics at the top of the enclosure feel the heat dissipated by the circuits located below. The rising air also informs the location of the vents at the top and bottom of the enclosure.
The guru (Small g) Mind Trick:
At the PCB level, removing the blanket of solder mask on any hot traces or areas designated for heat sinking will allow the exposed metal to radiate more thermal energy. This is a free performance upgrade, unlike using a thermal fill material in the vias. Also, the exposed metal is ready to take a bus bar or other passive heat sinking.
Active cooling can be as simple as moving the air with a fan or attaching a heat pipe to the top of the device and using that as the main cooling path. Thermo Electric Coolers can be placed on the board as another active measure. They work like tiny refrigerators.
Back in my assembly days, we had a TO-3 transistor that ran much too hot for board mounting. The power transistor was mounted on a heat sink that was, in turn, mounted to the chassis with one-inch long standoffs. The transistor would be screwed into the middle of the heat sink, and the heatsink would be set to the side of everything else with the special hardware. Heavy-duty wire completed the circuit.
Image credit: Amazon
Over-Engineering (in a Good Way)
Cooling takes many strategies, and one of the main ones is derating the components. If the requirements call for a 5 Volt rating on a capacitor, using a 10 Volt rated cap in its place means you have headroom to spare. That component is not being taxed by the effort and will give a longer service life. Likewise, a quarter-watt resistor will handle an eighth-watt load all day long. Higher performance parts are generally larger, but the mass is part of what makes them more robust.
Connectors are, in some ways, like the junctions of the device but scaled to the next order of magnitude. The mains power is concentrated in that one spot. Locating the connector away from active components and treating the area around the connector as a heat spreader will help when the machine is asked to do the hard work. More is more when making the strong power connection.
Compare the cost of the parts with the expected degradation over temperature. Upgrading the weakest link to a part that is closer to the average raises the overall reliability. Perfect reliability engineering would result in an assembly where every single part fails on the same day. That day would be the one right after the warranty expires.
In the real world, we have infant mortality where something failed right out of the box and end-of-life (EOL) when the product’s various components have given all they have to give. So it goes, at the EOL, the failure rate is higher than can be justified for continued operation. The owner has hopefully been paid back on their investment over time and you, as the designer, have had time to come up with an all-new iteration.
Everyone is ready for the new and improved version. The goal is a never-ending stream of products going in one direction and money flowing in the other (your) direction. That can only happen if the reliability is there. The PCB Designer has a lot to say and do towards that end. Stay cool.