Issue



BGA Socket with Heatsink for High-power Devices


01/01/2004







BY ILA PAL

The classic function of a socket is to provide a connection mechanism from the IC to the circuit board with as little electrical load as possible. This allows the IC to function as it is soldered into the printed circuit board (PCB), but can be replaced by another IC to upgrade or test multiple ICs. With the advent of light-bulb-power-type ICs getting into the tens of watts, perhaps even 40 to 50 watts, the socket must accommodate removal of the heat due to this power or the IC will self-destruct.

Socket Overview

Two of the most common high-frequency contactors for a socket are elastomer and pogo pin with bandwidths of 6 to 20 GHz and 2 to 10 GHz, respectively, depending on the connection components selected. A socket consists of the connection mechanism that interconnects ball grid array (BGA) balls to the target PCB pads, socket body for precisely placing the IC over the connection mechanism, and a compression mechanism to apply the down force to compress the BGAs into the connection mechanism (Figure 1). In high-power applications, the compression mechanism must perform double duty — acting as a heatsink pulling the heat out of the IC. A backing plate may be needed to provide the rigidity for the connection mechanism to work reliably. This is dependent on the size of the chip, flatness required, PCB material and thickness (Figure 2).

Modes of Heat Transfer

Heat transfer can be defined as the transmission of energy from one region to another as a result of temperature difference. Heat conduction is caused by the property of matter that causes heat energy to flow through matter. Heat convection is caused by the property of moving matter (naturally or under force) to carry heat energy from a higher temperature region to a lower temperature one. Heat radiation is caused by the property of matter to emit and absorb different kinds of electromagnetic radiation.

A common heat sink system consists of a finned metal part forced down onto an IC that pulls the heat from the IC through conduction and then radiates the energy into the air. A fan is added to blow air over the heat sink to remove the heat through convection, which greatly reduces the temperature of the heat sink. The greater the velocity of the airflow, the lower the heat.

When designing a heat sink, two characteristics to be considered include thermal resistance from die to the IC heat spreader, and the size and material of the heat spreader on the IC.

Heat sinks are designed to transport heat away from the heat spreader on the IC to ambient air. Other characteristics to be considered for a heat sink include: thermal resistance from IC heat spreader (case) to the heat sink; XY area of the heat sink; whether the heat sink is finned or pin design (height of fins or pins); whether a fan is used and what its flow rate is; mounting relationship of fan to heat sink (vertical or horizontal); heat sink material (aluminum or copper for machined parts); maximum temperature to be allowed at die; ambient conditions of air and wall temperature and pressure (altitude); and heat sink color (black can be significantly more efficient than white).


Figure 1. A functional drawing of a socket.
Click here to enlarge image

null

Thermal Resistance

Thermal resistance is the critical parameter of heat sink design. The thermal resistance between a silicon-die junction and the ambient air determines whether a heat sink is adequate. Total thermal resistance can be calculated as:

ja = jc + ch + ha, where:
ja (junction to ambient thermal resistance) = (Tj-Ta) / P;
jc (junction to case thermal resistance) = (Tj-Tc) / P;
ch (case to heat sink thermal resistance) = (Tc-Ti) / P;
ha (heat sink to ambient thermal resistance) = (Ti-Ta) / P;
Ti = heat sink temperature;..Tj = junction temperture;
Tc = case temperature; .Ta = ambient temperature; and
P = power dissipation.


Figure 2. A BGA socket with a 30-W heat sink.
Click here to enlarge image

null

Thermal resistance is directly proportional to the thickness of the material, and inversely proportional to the thermal conductivity of the material and surface area of heat flow. A heat sink design can be a complex task requiring extensive math — finite element analysis, fluid dynamics, etc. Luckily, tools are available that quickly provide fairly accurate thermal models.

To design a 30-W heat sink, we used off-the-shelf heat sink design software.* Designing the heat sink is an iterative process with the possibility of changing a number of heat sink parameters. Part of the design is to make sure that the heat sink is manufacturable. For instance, changing the length and spacing of the pins can reduce the cost of the heat sink. Several variables (surface area of the heat sink base, reduced mass, structural rigidity for ease of manufacturing, fin geometry, cross-cut pins for efficient air flow, etc.) can be modified for optimal thermal design.


Figure 3. Expected heat distribution from the heat sink.
Click here to enlarge image

null

When two surfaces are interfaced, voids between the surfaces can cause poor heat transfer. This normally is solved with a thermal interface material that fills the gap between surfaces. Depending on the application and mating surface, a thermal interface material can be selected. Thermal grease can spill over intricate parts and cause contamination, which is especially dangerous in a fine pitch, tight tolerance socket. It becomes necessary to have a solid silicone pad for high-density interfaces, even though thermal resistance is higher for the silicone pad than the thermal grease.

In a case study, a top-mounted heat sink was used for an IC with an encapsulated die with a 16 x 16-mm XY dimension. The IC had a copper heat spreader with a dimension of 39.6 x 39.6 mm. The aluminum-finned heat sink pressed down on the IC's copper heat spreader with a 0.25-mm-thick thermally conductive filled polymer silicone laminate to fill voids between the two metal surfaces. This silicone laminate has a thermal resistance of 0.838708∞C sq. cm/W. With a size of 39.6-mm2, the temperature rise across this interface pad was ~1∞C. The heat sink was 60.5 x 60.5 mm, with 100 square pins that were 3 mm square and 18 mm high. A top-down fan was used with 8.67 L/sec capacity to convection cool the heat sink. Ambient air temperature of 25∞C and ambient pressure of 101.325 KPa were used. At a 30-W load from the die, the junction temperature was predicted to be 72.97∞C (Figures 3 and 4). The hottest spot on the heat sink was 46.27∞C, with an average temperature of 36.17∞C.


Figure 4. Expected heat distribution from the heat sink.
Click here to enlarge image

null

Conclusion

Thermal management of BGA socket is an all-inclusive method that involves material selection, design, analysis, optimization and verification of a cooling system for the purpose of producing reliable sockets for testing high-power devices. Currently, traditional methods of cooling components through conduction and convection are a satisfactory solution to many problems. In the future, cold plates, cold water circulation pipes and other enchancement techniques will be applied to lower temperatures.

*Q-fin

ILA PAL, principal engineer, may be contacted at Ironwoood Electronics Inc., PO Box 21151, St. Paul, MN 55121; (651) 452-8100; e-mail: mannan@ironwoodelectronics.com.