As newer devices operate at ever-escalating power levels, controlling temperature during test gets tougher. These challenges may be difficult, but some solutions are developing.
Low-cost burn-in with test—an essential ingredient in producing highly reliable parts—faces new challenges as the power levels of semiconductor devices continue to rise. Achieving low cost requires burning in and testing as many devices in parallel as possible in a minimum amount of test floor space. This results in huge amounts of power and cooling in relatively small test systems.
Even more challenging is the need for new solutions that lower the thermal resistances between devices and the cooling medium. These solutions would support accurate individual temperature control of each device in the test system.
Why individual temperature control? The power consumed by high-power devices varies by up to 40% due to variations in the fabrication process and different operational modes of the device. Also, the airflow in a burn-in system can fluctuate by as much as 30% or more. This can lead to large temperature variations among the same device types in a single burn-in system.
If the device becomes too hot, it may be damaged while other devices may not be adequately burned in. To uniformly stress all devices, they must be kept very close to the specified burn-in temperature. This is best accomplished by individual measurement and temperature control of each device.
High-Power Burn-In With Test
Burn-in with test subjects devices to various input stimuli at temperatures generally ranging between 100°C and 150°C and at elevated voltages. It helps to eliminate infant mortality failures, allowing device manufacturers to greatly increase their part reliability and reduce overall costs. A typical high-power system will burn-in and test as many as several hundred devices in parallel, depending on the device power.
A high-power burn-in system requires the following functions:
- Programmable Power—Voltage regulators supply power to devices on the burn-in boards and provide voltage and current measurements to the system controller. High-power devices require sophisticated power supplies that can offer up to hundreds of amps for each device.
- Individual Temperature Control—This is the capability to monitor and control the temperature of each device on the burn-in board.
- Test Sequences—Pattern generation and control are used to direct test sequences and store test patterns.
- Test Output—Drivers/receivers produce the waveforms called for by the pattern generation and send them to the burn-in board. They also test return data from the burn-in board and log errors. This allows devices to be sorted into pass or fail at the end of a burn-in run.
- Device Testing—The devices typically are tested in sockets on a burn-in board. Burn-in boards interface with a voltage regulator, temperature control, and drivers/receivers. An oven or chamber contains the burn-in boards as well as the cooling mechanism(s) used to keep the DUTs at the desired test temperature.
Four principal challenges to high-power burn-in present themselves. As the power increases, it is necessary to:
- Lower the thermal resistance between the device and the cooling medium.
- Accurately measure each device individually.
- Control device temperature individually.
- Provide as much as hundreds of amps of current to each device at low voltages.
Solutions to these challenges allow devices to be properly stressed at uniform temperatures.
With the increase of device power, it becomes crucial to lower the thermal resistance. To make matters more complex, thermal resistance can vary widely depending on the thermal interface between the die and heat sink, the surface area of the device, the number of pins, the device and heat-sink surface finish, and the pressure applied to the heat sink. As shown in Figure 1, a small reduction in thermal resistance can make a large difference in the capability to control device temperature while dissipating more power.
Consider what is required to burn-in and test a device dissipating a maximum of 100 W where the cooling fluid temperature is 10°C. Assuming that the burn-in temperature must be between 100°C and 150°C, a sufficiently low thermal resistance is required to hold the device at 125°C while dissipating 100 W. The fundamental heat-flow equation can be expressed as:
Q = TD – TCF
RDHS + RHSCF
where: Q = heat flow in W
TD = temperature of device in °C
TCF = temperature of cooling fluid in °C
RDHS = thermal resistance between the device and the heat sink in °C/W
RHSCF = thermal resistance between the heat sink and the cooling fluid in °C/W
Fluid refers to the cooling medium used, such as air or liquid.
Solving this equation for the required thermal resistance yields the following:
100 = 125 – 10
RDHS + RHSCF
RDH + RHSCF = 1.15°C/W
As a result, the total thermal resistance can be no higher than 1.15°C/W. Since the heat sink to fluid can be 0.1°C/W or lower, the device to heat-sink thermal resistance must be less than 1.05°C/W. If the power increases to 300 W, the total thermal resistance required must be less than 0.38°C/W to maintain 125°C operation.
Lowering the thermal resistance is somewhat problematic. A larger device surface area allows more heat to be dissipated but runs counter to industry pressures to reduce device size.
The material used as a thermal interface has been the subject of much research, with each solution having positives (lower thermal resistance) and negatives (cost, weight, interaction with the system, potential residue, higher thermal capacity, flow rate, and lower working temperature). Heat spreaders can be used to lower thermal resistance for smaller devices, but they must be fixed to the device by the manufacturer. Applying more pressure to the interface between the heat sink and the device also can reduce thermal resistance.
The fin spacing, airflow rates, and the shape, thickness, and height of air-cooled heat sinks must be balanced to achieve optimum performance. The composition of the heat sink also is a critical component. Copper may have a lower thermal resistance than aluminum, but it also is substantially heavier and more expensive. And finally, a balance must be found between efficiency and cost.
Individual Device Measurement
Another challenge focuses on the accurate measurement of each device’s temperature. It can be advantageous to insert a temperature sensor in the heat sink so that it contacts the device but is thermally isolated from the heat sink. In other cases, it may be advantageous to embed the temperature sensor in the device. Three types of temperature sensors typically are used: the thermocouple, the resistance temperature detector (RTD), and the forward-biased diode embedded in the device itself.
The RTD is more convenient electrically but larger than the thermocouple. Since it requires more space on the device for temperature measurement, it reduces the amount of space available for heat flow to the heat sink. Use of the RTD also can increase the size of the test socket, where space is at a premium.
The forward-biased diode only can be used if the device has such a circuit built into it. This has been more common with high-end microprocessors. The forward-biased diode is the cheapest and potentially the most accurate alternative but not the most readily available. Currently, technology limits the measurement range of the embedded diode to about 145°C or less.
All of these factors must be balanced for an optimal design.
High Device Current
A system testing 100 500-W devices would need 50 kW of power just for the devices. In addition, the high current requires larger conductors and connectors to minimize the voltage drop and keep the conductors from getting too hot.
As transistor sizes are scaled down to maximize performance, the device operating voltage is lowered. With low operating voltages, the noise margin is narrowed significantly.
Solutions to Specific Needs
1 W to 35 W
For device power of 35 W or less, temperature can be controlled using individual heaters inserted into the heat sink and cooling air blown over the test sockets. Water-cooled air from heat exchangers is blown across all the devices. Individual heaters are pulsed on and off as needed. A temperature sensor inserted into the heat sink measures the device temperature.
20 W to 60+ W
When the power reaches the 20-W to 60-W range or even 100 W, one approach uses a small fan mounted above the heat sink of each device (Figure 2). A heater inserted into the heat sink pulses on and off to raise the temperature of the device. The fan is pulsed on and off alternately with the heater to control the temperature.
Cooling air is blown through the system and onto the devices by individual fans. A temperature sensor embedded in the heat sink measures the temperature of the device.
50 W to 200+ W
Above 50 W to 75 W, liquid-cooled heat sinks offer a solution (Figure 3). The heat sink contacts the device or a heat spreader mounted on the device.
Chilled water typically is used as the coolant. The spring-loaded heat sink is lowered to make contact with the device. The amount of force can be adjusted by varying the height of the plate on which the heat sink is mounted.
A temperature sensor is embedded in the heat sink to measure the device temperature. To further reduce the thermal resistance between the heat sink and the device, helium can be injected into the interface between the heat sink and the device. This lowers the thermal resistance by approximately 40%.
Device power continues to rise. A system has been designed to handle 400 W of power, and indications are that 600 W or more will be needed. These power levels demand a low thermal resistance. As seen in Figure 1, to keep device temperature at or below 100°C for a 600-W power requirement, the total thermal resistance would have to be 0.15°C/W or less.
While the devices are becoming more powerful, the relative size per watt is becoming much smaller. The increased power density makes temperature control more difficult. Yet the marketplace is demanding smaller devices and smaller system footprints to test the devices.
Future systems will need many test sockets in as small a space as possible. For that reason, there is a trend for some manufacturers to produce smaller, modular systems accommodating fewer devices. If a system fails, a smaller portion of the manufacturing capacity is affected. As cost becomes an increasingly greater consideration, the burn-in with test industry is striving to strike a balance between cost and capability.
About the Authors
Harold E. Hamilton is founder and president of Micro Control. He has a B.S.E.E. from the University of Nebraska and an M.S.E.E. from the University of Minnesota. e-mail: [email protected]
Jeffrey C. Urbanek is a technical writer at Micro Control. He earned a B.A. in English and an M.A. in journalism from the University of Wisconsin-Madison. e-mail: j.urbanek@microcontrol
Micro Control, 7956 Main St. NE, Minneapolis, MN 55432, 763-786-8750
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2002 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.