Control Chip Temperature During VLSI Device Burn-in

To compensate for these power-dissipation variations, you can provide independent temperature regulation for individual devices during burn-in. Such regulation keeps you from damaging good parts with high dissipation and ensures that you adequately stress devices having inherently low dissipation.

The traditional bathtub curve (Fig. 1) represents semiconductor device failure rates as a function of time. The initial failure rate is high, but devices that survive the first few hours—the infant-mortality period—operate reliably until they reach the end-of-life stage many years hence.

The burn-in process uses power and temperature extremes to compress the infant-mortality period, forcing failures to occur quickly, thus saving time and money. The more extreme the power and temperature, the sooner infant mortalities occur, reducing the required burn-in time. Ultimately, increasing voltage and temperatures will not only compress the infant-mortality period but also weaken good devices that will survive the burn-in yet fail well before their anticipated end of life.

Semiconductor companies spend considerable time and money determining the optimum burn-in temperature for production devices. An effective production burn-in system forces all devices to stay close to that optimum temperature. It compensates for variations in the dissipation characteristics of the devices undergoing simultaneous burn-in as well as for variations in each device’s dissipation in response to varying electrical inputs during the burn-in-and-test cycle.

Burn-in Strategies
Burn-in strategies include static, dynamic, and burn-in with test. Static burn-in systems apply extremes of voltage and temperature to each device but do not exercise the device. Thus, static burn-in—the least expensive of the burn-in strategies—does not stress all the potential failure mechanisms.

Dynamic burn-in systems exercise the device inputs and properly terminate the outputs in addition to applying extremes of voltage and temperature. With dynamic systems, electron charge transfers occurring at the exercised device’s circuit nodes initiate failure mechanisms that would escape static burn-in.

Burn-in-with-test systems test devices while stressing them. They provide test vectors to a device and compare actual device outputs with expected outputs while the device under test (DUT) operates at its voltage and temperature limits. Burn-in-with-test systems can identify devices that fail to meet spec under marginal conditions but that would pass a post-burn-in room-temperature test.

Burn-in-with-test systems also verify that a device under test gets exercised—that is, the device is powered up and test vectors are applied. Keep in mind that burn-in sockets—fragile high-pin-count components subjected to the repeated insertion/extraction cycles of production burn-in—are themselves prone to failure. Just a few bad socket pins could prevent test vectors or supply voltages from reaching the device undergoing burn-in, resulting in your shipping or using parts that haven’t been electrically stressed.

In one approach to high-power burn-in, an operator plugs devices into sockets on one side of a burn-in board (Fig. 2). A clamshell or other press fixture then brings a heat-sink assembly (Fig. 3) into contact with each device.

Figure 2. This burn-in board accommodates 24 DUTs. A burn-in-with-test strategy ensures that supply voltages and test vectors pass through the fragile burn-in-board sockets to the DUT.

Figure 3. A heat-sink assembly attaches to each DUT on a burn-in board. A sensor provides temperature information to a control subsystem, which regulates device temperature by controlling heater duty cycle. Thermal foam ensures good thermal contact between DUT and heat-sink assembly.

The heat-sink assembly contains a spring, temperature sensor, and heater. The spring holds the temperature sensor tightly against the device package to ensure good thermal contact. The control circuitry monitors the device temperature and supplies the proper heater power to maintain the device at the required temperature.

Maintaining Die Temperature
The die inside the DUT is the active component whose temperature is most important. The high-power burn-in system monitors and controls the package temperature, which in turn controls the die temperature in accordance with the thermal impedance between die and package. This thermal impedance is typically less than 0.5°C/W, depending on package, die, and die mounting method. Although this thermal impedance varies by device type, it’s usually uniform for a given part number.
You can calculate the package temperature (TP) required to maintain a specific die temperature (TD) from the thermal impedance (q) and the DUT heat dissipation (P):

T_P = T_D – (q x P)

For example, the package temperature required to provide a die temperature of 150°C for a device with the thermal impedance equal to 0.25° C/W and a heat dissipation of 10 W is as follows:

T_P = 150°C – (0.25°C/W x 10 W) = 147.5°C

Thus, controlling the package temperature to 147.5°C will maintain the die temperature at 150°C at 10-W dissipation.

Air Temperature and Velocity
A burn-in test chamber accommodates multiple burn-in-board/heat-sink-assembly combinations and provides a uniform airflow. The optimum air temperature and velocity are functions of device power, the thermal characteristics of the heat-sink assembly, and the required package temperature.
The air temperature and velocity must ensure that the heater can control the device temperature over the full potential range of heat dissipation. If the air stream carries away too much heat from the heat-sink assembly, the heater will not be able to maintain the package at the desired temperature. On the other hand, if the air stream carries away too little heat, the device will become too hot even with the heater turned off.

The quantity of heat transferred to the air stream is proportional to (T₁ – T₂)Ö V, where T₁ is the heat-sink-assembly temperature, T₂ is the air-stream temperature, and V is air velocity. Figure 4 illustrates airflow rates and air-stream temperatures for three rates of heat flow out of a heat-sink assembly.

Figure 4. Many combinations of air temperature and velocity will remove a given amount of heat. Each of the three lines here represents the various combinations of air temperature and velocity that will provide the required heat flow from a heat-sink assembly.

Air temperature and velocity are generally set so that the heater runs at half power (50% duty cycle) when the device is operating at nominal power. This choice centers the heater output relative to the device power range. As the heat given off by the device increases, the control circuitry senses the temperature increase and reduces the heater power, allowing the package temperature to settle back to the setpoint temperature. Similarly, as the device dissipates less heat, the heater will dissipate more.

The total heat dissipation of the heat-sink assembly is nearly constant as the power dissipation of the device varies under test. Thus, the air temperature and velocity, once appropriately set, need not be changed during testing. The temperature-control mechanism compensates for variations in the air temperature and velocity at a particular device in the test chamber. For example, slightly warmer air at one device location would result in slightly less power to the heater for the corresponding device.

Control Software   
A high-power burn-in system’s software can provide individual device temperature control and monitoring. In the Figure 5 example, the display provides color-coded information about each device; text codes provide details. The display also shows global temperature settings for the burn-in board as well as actual minimum, maximum, and average temperatures and heater duty cycles. T&MW

Figure 5. Burn-in control software can provide information about each DUT on a burn-in board. Here, green indicates satisfactory temperature performance; red indicates a temperature anomaly; and blue indicates a disabled heater. The 266.3°C temperature reading for DUT 22 could indicate an open resistance-temperature-detector (RTD) sense line.

Harold E. Hamilton is founder and president of Micro Control Co. He has a B.S.E.E. degree from the University of Nebraska and an M.S.E.E. degree from the University of  Minnesota. Charles H. Morris, technical writer at Micro Control Co., has a B.S.E.E. degree from the Georgia Institute of Technology.