Cost-effective burn-in-with-test is essential to produce high reliability components. To ensure uniform burn-in of all components, a test solution must control the temperature, voltage, power, test sequence, and inputs/outputs to the DUT. Of these, temperature becomes significantly harder to control as the power level of the devices increases.
When burning in relatively low-power components, pressing a heat sink against the components and blowing air over the devices in a temperature-controlled chamber may be effective. As the power of semiconductor devices rises, individual temperature control becomes very important to ensure that all devices are uniformly stressed.
Power consumed by high-power devices can vary by 40% or more due to variations in the fabrication process and different operation modes of the device. The thermal interface between the DUT and any cooling sink can vary significantly, and the airflow in a burn-in system can fluctuate by 30% or more.
All of these variables can lead to large temperature variations among the same device types in a single burn-in system. Some components may be damaged while others are inadequately burned in.
Figure 1 is a thermal schematic representing the flow of heat during the burn-in process on a packaged die. An important factor not detailed in this schematic is how well the device is packaged. Most burn-in systems accept prepackaged silicon components. The package must allow for conduction of the heat from the silicon to the packaging material. This also needs to be consistent from device to device to enable accurate temperature control.
Figure 1. Thermal Schematic of a Packaged Die
The remaining areas of concern are detailed in the schematic. To maintain proper temperature control of each semiconductor device during the burn-in process, a thorough understanding of each of these factors needs to be gained. An effective burn-in solution must address:
� The interface resistance between the DUT and the heat sink applied to the face of the device.
� The resistance to heat flow down through the socket pins into a burn-in board (BIB) and into the chamber cooling air.
� The resistance between the heat sink design and the cooling fluid. This cooling fluid can be air, water, or some other medium.
� The accuracy of the temperature measurement electronics.
� The effectiveness and uniformity of cooling fluid delivery in the chamber.
Interface Resistance
As the power of devices increases, an effective burn-in solution needs a low resistance to heat flow from the device to the cooling fluid. Most often, the way to achieve this is to press a heat sink against the device. Ideally, the resistance to heat flow across this interface needs to be very low and repeatable from device to device.
The interface resistance is affected by numerous variables. Among these are the materials used, the flatness of the packaged device and the heat sink, the force applied, how centrally this force is applied, the area of the interface, and how clean the interfaces are maintained.
Materials
Exhaustive research has been done on thermal interface materials and the benefits and disadvantages of each. The advantages of low thermal resistance and void-filling capability are traded off against the disadvantages such as cost, interaction with the system, residue, lower working temperatures, and effectiveness.
Many applications benefit from a properly selected interface material. Testing on any proposed interface material is essential since application-specific data can vary greatly from published numbers.
Flatness
Both the packaged device and the heat sink must be very flat. Whether using an interface material or not, maintaining consistently flat surfaces greatly improves performance and consistency. Any voids created between the device and the heat sink will be filled with air, which has a poor thermal conductivity.
One way to improve this performance is to inject another medium to fill these voids. Introducing helium, which has six times the conductivity of air, will help overcome flatness problems.
Figure 2 shows a flatness measurement for a water-cooled heat sink, which is used for greater than 200-W components. This heat sink is lapped to a flatness of half a micron and has eight holes for injecting helium into the interface. The large hole in the middle is for a temperature sensor. Heads with this type of interface can achieve resistances of 0.1�C/W on a 22�mm x 22�mm interface at a relatively low interface pressure of 10 psi.
Figure 2. Flatness Measurement of a Water-Cooled Heat Sink
ForceThe greater the force applied between the heat sink and the device, the lower the thermal resistance. The amount of force may be limited by the heat sink design, the chamber design, or the device packaging. Whether the force is being applied by the device socket or the burn-in chamber, the force must remain consistent from location to location. A variation in this force will adversely affect the accuracy of the burn-in temperature. Centrality of ForceForce must be centrally applied. The interface resistance on a 22 mm x 22 mm heat sink can be doubled by shifting the applied force 3 mm to 4 mm from the center of the package. Size of the InterfaceThe greater the interface area, the greater the amount of heat that can be passed through the interface. This often conflicts with pressures to reduce device size. CleanlinessClean interface surfaces are going to produce much more repeatable results since contamination particles will introduce additional voids. Heat Flow Through the Socket and BIBLarger components can have a large number of pins that produce a low resistance path for heat to flow into the BIB and into the chamber cooling fluid. Understanding the amount of heat flowing into the board and applying techniques to accommodate it are important. � A device with 200 pins could have a resistance to cooling air in a chamber of 11�C/W. A device with 2,000 pins may have a resistance of 1.5�C to 2.0�C/W. This low resistance might allow up to 80 W of power to flow down into the BIB with a device-to-air temperature difference of 140�C. Applying heat sinks to the bottom of the BIB will be necessary to remove this amount of heat. � Many low-power BIBs have backer plates added to the underside to protect the components that may be present. As device powers increase, this can create a problem with heat generation in the BIB. Flat plates may need to be added to the bottom of the BIB in medium-power applications and large heat sinks added in high-power applications. Controlling the conducted heat into the BIB is essential for temperature control. Heat Sink Considerations
Over-designing a heat sink for a given application can be as detrimental to controlling device temperature as an undersized heat sink. When a heat sink is too large, the chamber may not be able to maintain temperature on individual devices, and these components will experience less than the required temperatures. A number of steps must be followed to properly design a heat sink:
1. Device maximum and minimum power must be known.
2. Using the maximum power (Q), we can determine the total resistance (Rt) required from the DUT to the cooling fluid given the cooling fluid temperature and the burn-in temperature:
Rt = ?T/Q
3. Using the schematic representation in Figure 1:
Rt = (Rds + Rsa)Rba/(Rds + Rsa + Rba)
where: Rds can be obtained via data sheets for the interface material or, preferably, by experiment and Rba can be determined from testing or from experience. Therefore, the required heat to cooling fluid resistance Rsa can be determined.
4. Using the chamber and socket cooling fluid flow data, a heat sink now can be designed to match the application.
5. A device at the low end of the power spectrum then needs to be checked to ensure the chamber can maintain the temperature of these devices. If not, a heater may have to be added. Make sure the chamber can supply the required power for the heater and place the heater where the heat can reach the device instead of vanishing into the cooling fluid.
6. Ideally, analysis of the design should be performed using a simulation software package followed by a prototype.
Temperature Measurement
Minimizing temperature measurement inaccuracies is very important. Ideally, this would be zero percent of the total allowable temperature accuracy budget. Unfortunately, this is not practical due to a number of factors.
Temperature measurement can be accomplished via a thermocouple, a resistance temperature detector (RTD), or a forward biased diode embedded in the device itself. Thermocouples and RTDs typically are embedded into the heat sink but thermally isolated from it. A forward biased diode may be the cheapest and potentially most accurate solution but it must be designed into the device and frequently is not done.
Embedding the temperature measurement device into the heat sink has the potential to affect the temperature reading. The same concerns in trying to maintain a good interface between a heat sink and a DUT apply to the interface between an RTD or thermocouple and the device.
Also, while ideally the measurement device is isolated from the heat sink, it is cooled slightly by a heat sink that is cooler than the device. This is especially true when using liquid-cooled heat sinks and high-power devices. The combination of the contact resistance and cooling effect of the heat sink leads to a difference between the measurement temperature and the actual temperature.
A calibration fixture may have to be developed to create a correction factor that is used to determine the actual temperature of the device based on the measured temperature and the heat sink temperature:
TD = TM + CF x (TM � THS)
where: TD = device temperature TM = measured temperature CF = correction factor THS = heat sink temperatureRTD vs. Thermocouple
RTDs and thermocouples each have their advantages and disadvantages. RTDs are more convenient electrically since they provide a direct temperature measurement. Each time a thermocouple is terminated, the junction temperature at that point must be compensated. RTDs can use stranded wires, which gives greater flexibility than the solid wires of a thermocouple.
Thermocouples are lower cost, smaller, and take up less space in the heat sink. RTDs also have a higher and less repeatable interface resistance to the device. This variation can greatly increase the amount of the temperature error budget taken up by the measurement device. Thermocouples can be made into much more durable assemblies since RTDs are constructed of thin, fragile ceramic.
Burn-In Chamber Layout Considerations
When considering a chamber for a given burn-in application, the movement of the cooling fluid past the parts is very important. Ideally, the chamber should be able to control the flow of cooling fluid into each device, and devices upstream should not affect downstream devices.
If cooling air is simply directed from one side of the BIB to the other, heat sinks on the BIB will have to be oriented parallel to this flow. A device on the leading edge of the BIB, which is generating heat, will raise the temperature of the air presented to the downstream devices and make temperature control very difficult.
A possible solution is to vary the design of the heat sink from one edge of the board to the other. This is not a simple task and would increase the cost and complexity of the BIB.
A much better solution is to present the air from above the device. If the air be presented to the heat sink individually from above, either via an air-controlling tray in the chamber or by indiviual fans mounted to the actual heat sink, each device can be presented the same temperature air. Also, the heat sinks can be oriented so that heat introduced by upstream devices flows by the sides of the downstream devices and not over the cooling fins of the heat sinks as shown in Figure 3.
Figure 3. Airflow Circulation on a BIB
Another chamber factor that needs to be considered prior to designing a heat sink is how much air the chamber can supply. Using the approach in Figure 3 varies the amount of air available to each device when the number of devices increases. This will be true when using fans attached directly to the heat sink as well. An insufficient amount of air being delivered by the chamber will starve individual fans of air.
Conclusion
As component powers increase, simply applying a heat sink to the devices and blowing air over them will not guarantee a successful outcome. Individual temperature control must be done to guarantee that all devices are equally burned in without damage to some and insufficient testing to others. This is accomplished by:
� Determining how a device will be tested while designing the device.
Many factors can be considered in the early design phases of a device. Repeatability of the die to package thermal resistance is probably the most important of these. The flatness of the final package also is very important to achieve a low and repeatable package to sink resistance. Embedding a forward biased diode into the device may aid in accurate temperature measurement when trying to achieve burn-in later.
� Fully understanding the characteristics of the chamber that will be used to accomplish the testing.
To properly control burn-in, the thermal circuit must be fully understood, and many factors must be considered such as how much heat will be transferred through the board, how the airflow is delivered to the device, and how much air is delivered to the device. How best to measure the temperature then can be considered.
� Developing tools to validate and verify designs.
Tools may be needed to measure the effectiveness of interface materials or flatness of interface surfaces. To design heat sinks that are not oversized yet can handle the amount of heat for a maximum power part requires adequate software tools. Calibration fixtures also might be needed to develop correction factors.
To achieve a cost-effective, reliable burn-in solution, it is important to understand all the factors involved and correctly apply the available tools and solutions.
About the Author
John T. McElreath is the mechanical engineering manager at Micro Control. He received a B.S.M.E. from Michigan Technological University and has spent his career designing machinery for the assembly, inspection, and test of electronic components and assemblies. Micro Control, 7956 Main St. NE, Minneapolis, MN 55432, 763-786-8750, e-mail: [email protected]
FOR MORE INFORMATION
on high-power burn-in
www.rsleads.com/710ee-185
October 2007