Considerations, calculations and cautions regarding heat sinks and thermal interfaces.
By Mark Murdza and Erik Orwoll
Burn-in is a method of product qualification, and necessary in the manufacture of integrated circuits (ICs). The burn-in process subjects ICs to extreme temperatures, electrical currents and/or electrical signals that stress the ICs and force an infant mortality. This means that the manufacturer can weed out the devices that may fail early.
Recent trends in the IC industry are seeing processor speeds increase and package I/O densities decrease. As a result, device temperatures are on the rise – creating new burn-in challenges for major processor, SRAM and controller device manufacturers. To meet these higher temperature requirements, thermal management needs to be an integral part of the burn-in process.
If devices overheat during burn-in, they will surely fail. Often, the power input to the device – the thermal load – is pushed to the device's limits. Many times the manufacture of the die results in a device's ability to overheat at high speeds; if this thermal runaway cannot be controlled, the devices being tested will be “forced” into failure during burn-in. The result is reduced manufacturing yields, and the challenge for device manufacturers and socket suppliers is to find thermal management solutions to control these problems.
There are several questions that have to be addressed before a solution to the complex thermal management problem can be found. Until then, an appropriate thermal management solution cannot be implemented during the burn-in process. And there are several variables that need to be investigated further.
The first consideration is a thermal management strategy. The implementation of a thermal management solution can include a number of methods, such as the incorporation of fans, liquid heat exchangers or thermoelectric coolers. However, these devices can be costly to purchase and require investments to support them. Fans and thermoelectric coolers require extra wiring and hardware for the purposes of monitoring and power. Liquid heat exchangers necessitate the integration of fluid lines, so water can pass through piping and the unit absorbing the heat to pull the heat away from the system.
Heat Sinks
An easy solution to address thermal management, at least in how it's perceived, is by implementing heat sinks. Heat sinks maximize the surface area in contact with the cooling area, creating a conductive path that reduces die temperatures. Heat sinks are conductive devices that allow heat to travel through a material, rather than directly to the outside surrounding the device. The heat at the die moves through the heat sink at a rate that is dependent upon several things: the material, the surface area of the heat sink, the rate of airflow, the air temperature, the amount of surface area in contact with the die and the pressure at which that contact is applied. Essentially, heat sinks increase the surface area in contact with the cooling air, which results in an increase in heat dissipation. The mass with the larger surface area pulls more heat from the device than the device can dissipate alone.
Heat sinks can be designed and manufactured in many shapes and sizes, and with many thermally conductive materials – ranging from common aluminum to specialized polymers. The differences in thermal efficiency are dependent upon the material properties; while some materials are more efficient than others, many variables are involved when deciding which material is best-suited for an application. Aluminum is the most common material used because it is easy to manufacture, low in cost and thermally efficient. Also, aluminum heat sinks are generally manufactured through the use of extrusion processes; not only does this allow various heat-sink configurations, but it is a cost-effective method to manufacture them in high-volumes (Figure 1).
Heat Sink Concerns and Cautions
Caution must be exercised before using heat sinks because of the number of variables in the dynamic burn-in environment. Surface conditions, air velocities, oven temperatures and densities, and socket/device locations are contributing and complicating factors that must be considered.
One very important variable associated with heat sink design is interaction with the outside air. In fact, the temperature, velocity and flow of the surrounding air are the dominant factors that affect how a heat sink interacts with cooling air. Air passing over a surface pulls more heat out of the body than motionless surrounding air. In many cases, the cooler the temperature and the more turbulent the airflow, the better the heat dissipation. A turbulent airflow maintains an even, overall air temperature around the heat sink as the heat is pulled from the surface into the cooling air. If the airflow is laminar – that is, if it follows a straight parallel path – the air will always be hotter at the heat sink surface and coldest at a point farthest from the source. Turbulent airflow is more efficient in heat dissipation when compared to laminar airflow.
The geometry of the heat sink should promote – not restrict – a turbulent airflow (Figure 2). The heat sink design, as a result, must take into account socket orientation on the burn-in board, the orientation of the burn-in board in the oven, the airflow direction within the oven and the temperature gradient from socket to socket across the burn-in board. Air temperatures across burn-in boards may vary as air moves directly from one socket to the next, with minimal turbulence, carrying warmer air across the board and resulting in a less effective heat dissipation. This can also result in thermal failure at one side of the burn-in board, while the device on the opposite side of the burn-in board processes correctly.
Selecting a Thermal Interface
The location at which the heat sink and device make contact, the surface interface, is yet another important variable in thermal management design. There is a common misconception that the die surface mated directly with the heat sink is the optimal interface. Yet on a microscopic level, there are many surface voids that result in gaps between the heat sink and the die. These gaps create the opportunity for convective heat transfer rather than conductive heat transfer. Convection is defined as heat transfer between a surface and a moving fluid when they are at different temperatures; conductive heat is drawn across the object being heated. Conduction is more effective than convection with respect to heat dissipation between two bodies.
A thermal interface can be placed between the mating components to optimize the interface because heat is conducted across the medium instead of being convected across the voids (Figure 3). This interface must contain certain properties to be effective and appropriate for the burn-in and IC environments. Filling in the voids generates an increase in heat dissipation (i.e., it promotes conduction rather than convection) as heat moves from the heat source, through the medium and into the heat sink.
Many times, thermal tapes are used in IC applications. Tape selection is important. The tape must be compliant and able to withstand multiple cycles of applied forces and thermal variations, yet maintain its thermal efficiency and elastomeric properties. The tape must have high thermally conductive properties. The adhesive used to apply the tape must not contaminate or out-gas into the environment, and it must not leave residues on the die. Finally, the cost of the material should be compatible with the heat-sink/socket system.
Higher pressures, when applied to the die by the heat sink, also result in higher heat dissipation rates. And there is a balancing factor to this variable, as too little pressure will be ineffective, yet too much pressure may crack the die. The IC manufacturer should carefully specify the maximum allowable die pressure.
Proactive is Progressive
Many of the variables affecting thermal management require input from the socket manufacturer, as well as the user. Other variables that affect heat sink specifications include:
- Power usage as a function of time (peak and steady-state output)
- Die temperatures at burn-in
- Maximum allowable die temperatures that will not result in device failure
- Burn-in oven-temperature variations and the gradients across the oven
- Air flow analysis – characterization
- Power consumption variability, because individual processor speeds are typically inconsistent from device to device.
Compromises may need to be made when dealing with thermal management issues. For example, board spacing may need to be increased to enhance airflow, and oven temperatures may need to be reduced – both of which might reduce throughput and increase costs. Board layout and socket density may need to be reviewed, which could result in a need for more burn-in boards. Socket material changes, to those that are more thermally conductive, may also need to be pursued (remember, heat conducts through the socket contacts and housing, as well as into the printed circuit board itself).
After these considerations and calculations have been completed and the initial version of the thermal management solution has been implemented, full system characterization will be needed. Temperature readings need to be taken and analyzed at multiple locations throughout the system – including multiple readings at the die, heat sink, burn-in board and oven.
Finally, to fully and successfully optimize a thermal management solution, a partnership between the socket manufacturer and the user is paramount. Information, parameters, testing, validation and ideas need to be shared between these parties in order to find the appropriate solution to the particular requirement.
AP
MARK MURDZA is product marketing manager, and ERIK ORWOLL is engineering manager at WELLS-CTI. For additional information, contact Mark Murdza at 2102 West Quail Avenue, Phoenix, AZ 85027; 623-581-5330; Fax: 623-780-3987; E-mail: [email protected].