System Cooling Enhances Electronic Circuit Reliability

Effective thermal management must be employed at the component level and the system level (see "Shrewd Thermal Management Helps Defeat The Heat," ED Online 16767). The system may employ good circuit design practices and reliable components, but system reliability will suffer if appropriate temperature controls aren't properly implemented. Available devices, techniques, and software that provide thermal management at the system level include fans, liquid cooling, thermally enhanced circuit boards, and thermal analysis software.

Fans for electronic systems are usually attached directly to a motor's output, without gears or belts. Motors for ac-powered fans usually use mains voltage, while motors for dc-powered fans use low voltage, typically 24, 12, or 5 V. Fans for electronic systems usually use brushless dc motors to minimize electromagnetic interference (EMI). Table 1 lists the characteristics of fans employed in electronic systems.

A fan inside an electronic system draws cooler air into the case from the outside, expels warm air from inside, or moves air across a heatsink to cool a particular component. The use of fans and/or other hardware to cool a system is called active cooling. The width and height of these usually square fans are measured in millimeters, with common sizes including 60, 80, 92, and 120 mm. Round fans are also available.

Airflow ratings are in cubic-feet per minute (CFM) for a rotation speed in revolutions per minute (RPM). Fans with a higher CFM rating may produce less noise (in decibels, or dB). Some fans come with an adjustable RPM so they can produce less noise if the system does not need as much airflow. The type of bearing used in a fan affects performance and noise output (Table 2).

System packaging is important when employing fans for cooling. That is, components mounted on the pc board should not restrict air flow, particularly higher-power devices. Heatsink fins should be in line with the flow of air. In addition, if the cooling system uses air filters, they should be accessible for cleaning.

This heated liquid coolant is then pumped through the heat exchanger, which moves the heat from the liquid coolant to either the ambient air or, in the case of a liquid-to-liquid heat exchanger, to another liquid coolant. The cooled liquid coolant then flows through the pipes or hoses back to the cold plate. Most liquid coolants also use a small amount of additives to inhibit corrosion and lubricate the pump.

Cold plates come in different forms. One approach consists of copper or stainless steel tubes pressed into a channeled aluminum extrusion. A second technique uses flat aluminum or copper tubes with an internal fin that increases their performance. A third type consists of two aluminum or copper plates metallurgically bonded together with an internal fin.

Coolant compatibility with a wetted surface is a factor in selecting a specific cold-plate technology. A copper cold plate is compatible with water and most common coolants. Aluminum performs well with an ethylene glycol/water mixture, oils, and other fluids, but is not compatible with untreated water. Stainless steel is necessary when using de-ionized water or other corrosive fluids. Table 3 lists the possible problems that can occur with liquid cooling systems and the appropriate corrective measures.

Another liquid-based approach to combat heat is spray cooling with pinhole-sized openings that shower a semiconductor with droplets of a special liquid that won't damage electronic circuits. Cooling occurs in two ways. Heat is thought to disperse most efficiently when bubbles of vapor form as the coolant evaporates from the surface in a process known as boiling. The coolant also removes heat by simply warming as it flows over the semiconductor toward a drain. The liquid and vapor are then collected and recycled, and a heat exchanger with a slow-moving fan dissipates the heat they carry.

One type of thermally enhanced circuit board is the direct-bond copper (DBC) substrate commonly used in power modules. It consists of a ceramic base with a sheet of copper bonded to one or both sides of the ceramic material. In the manufacturing process, a copper-oxygen eutectic bonds to the copper and ceramic base. The top copper layer can be chemically etched in a manner similar to conventional printed-circuit technology. The bottom layer is usually soldered to a heat spreader.

One type of DBC uses alumina (Al2O3) as the ceramic base because of its relatively low cost, though it isn't a good thermal conductor and is relatively brittle. The aluminum-nitride (AlN) DBC is more expensive than alumina, but has better thermal performance. Beryllium oxide (BeO) can also be used for the ceramic base, but is usually avoided because of its toxicity.

A major advantage of DBC substrates is their low coefficient of thermal expansion (CTE), which is close to that of silicon. This ensures relatively good thermal cycling performance for semiconductors mounted on the substrate.

Continued on Page 2...

Another type of thermally enhanced circuit board, the insulated metal substrate (IMS), consists of a copper or aluminum substrate covered by a thin layer of dielectric material followed by a layer of copper. The dielectric is a proprietary polymer/ceramic blend that provides better electrical isolation and lower thermal impedance than a standard FR-4 board. Table 4 compares the characteristics of IMS and DBC.

MH&W's Keratherm 86/77 provides electrically conductive bonding sites on a flexible, thermally conductive material that can be attached to nonlinear, heat-spreading structures (Fig. 2). Uses include the mounting of LEDs and other hot components on curved surfaces where FR4 and other rigid materials can't be used. It also replaces thicker, heavier pc boards where size and weight are restricted, such as in portable electronics.

Keratherm 86/77 is a flexible thermal circuit that conists of a thin layer of etchable copper bonded to a highly thermally conductive silicone film. The strong, supple material can be applied to curved and irregular heat-spreading surfaces with peel-and-stick attachment.

A single 86/77 flexible circuit can replace multiple boards, eliminating jumpers and connectors while improving reliability. Different copper thicknesses are available for specific power requirements. With the copper layer intact, the material also can be formed into a Faraday-cage design for EMI protection. Kertherm 86/77 contains no hazardous materials, such as flame retardants. Parts can be readily handled, fabricated and recycled.

In evaluating thermal analysis software for electronic systems, it is imperative for the user to have ready access to technical support from the supplier. The user should consider the modeling methodology, definition of a system for analysis, creation of a computational grid, solution and control features, and presentation of the results.

• Does the software represent all three modes of heat transfer - conduction, convection and radiation? If not, it's a non-starter! Heat transfer in electronics is a fully coupled problem involving all three modes simultaneously. Methods that prescribe heat-transfer coefficients at solid/fluid surfaces simply aren't adequate.

• What form of interface do you use to define the geometry of your system? Check out the user interface yourself. Find out how easy it is to create geometry, and focus especially on how easy it is to move items around after gridding is complete.

• Can you transfer complete system geometry from your MCAD software (i.e., true CAD integration), or must you transfer one part at a time? Look for a solution that not only allows you to transfer geometry, but also enables rapid simplification of the MCAD geometry where necessary. This will save considerable solution time.

• Does the meshing technique of this software lead to the most efficient, speedy solution for the kind of problems you need to solve?

• How quick and easy is it to create the computational grid? Look for software that creates the grid instantly around the objects and then allows you to refine the grid between objects as necessary. You should be able to move or resize an object after the grid has been generated, and the grid should adjust itself instantly.

• Is it possible to view variables of interest - the temperatures of critical components, for example -   while the solution is converging? Early indication of the solution can facilitate swift decisions, speeding up the design process.

• How does the software present the generated results?

• Can you output data in both tabular and 3D graphical form?

• Can you easily view point, planar, and surface plots of variables such as temperature or air speed?

• Can you present the results in a way which non-technical people will quickly understand (e.g., can you visualize the air flow through your system as a 3D animation)?

• Can you save images as JPG or GIF files for quick and easy incorporation into reports?

Continued on Page 3...

Although performance has increased, power efficiency is still critical in embedded computer systems because lower power cuts operating costs, reduces the necessary fan noise, and lowers cooling requirements. It is also important in battery-based systems where lower power consumption extends useful battery life.
The goal of lowering power has led to new circuit techniques, called dynamic power management (DPM), that reduce a microprocessor's average power dissipation by dynamically reconfiguring a system to lower power consumption during low-workload periods. The clue to reducing power dissipation rests on the system power consumption equation of CMOS ICs:

where:

P = system power consumption
C = total capacitance of all circuits that need to be charged during signal transitions
V = applied voltage
F = frequency

Therefore, C is a constant. Reducing the operating voltage or frequency, or both, will lower overall system power consumption. However, reducing the supply voltage may also reduce the operating frequency, so it's not exactly a linear relationship.

In principle, DPM identifies low processing-requirement periods and reduces operating voltage (voltage scaling) and/or frequency (frequency scaling) to reduce operating power consumption. This technique is called dynamic voltage and frequency scaling (DVFS). During these low power-requirement periods, idle circuits can be turned off to save even more power.

Proposed DPM solutions can be categorized as either predictive or stochastic. Predictive schemes attempt to predict a device's usage behavior in the future, based on past experience. Stochastic techniques make probabilistic assumptions based on usage pattern observations.

To be effective, DPM must account for the time it takes to change a power-supply voltage. Plus, the processor must be able to operate reliably when its supply voltage or clock rate changes.

The introduction of multicore processors will require a more sophisticated approach to reduce power consumption. A technique proposed by James Donald and Margaret Martonosi of Princeton University (2006 International Symposium on Computer Architecture, ISCA06) employs "a multi-loop mechanism that allows the operating system and the processor hardware to collaborate on a robust, stable, and effective thermal management policy." The researchers know of no other architecture work exploiting multi-loop control.