junction temperature of IC
Semiconductor junctions — from the millions of transistors used in ICs to the large-area compound junctions that make high-brightness LEDs possible — can all suffer early failure due to increased heat. This becomes an even larger issue when feature size shrinks and current requirements expand. Even normal operation can create heat buildup, raising the junction temperature. Such elevated temperatures may increase the amount of defects within the junction, decreasing performance and shortening lifespan.
Therefore, an accurate temperature-measurement method for semiconductor devices is needed to prevent critically high temperatures. One technique is a simple junction-temperature measurement that can be performed using common test and measurement instruments. Results of this measurement can be used to monitor operating conditions for a given device. The ideal way to measure junction temperature is to monitor the device temperature as close as possible to the heat source. Current flowing through a semiconductor junction produces heat, which then flows through the junction material to the outside world.
Another method is to place a temperature sensor very close to the semiconductor junction and measure the sensor output signal. As the heat flows to the outside area, it would raise the temperature of the area and the sensor. Although a straightforward process, there are physical limitations due to the finite size of the sensor. In many cases, the sensor itself would be larger than the junction to be measured. It would add a large thermal mass to the system, as well as additional error to the measurement, thus degrading measurement accuracy. So this technique hardly helps most applications.
A better solution is to employ the junction itself as a temperature sensor. With most materials, there’s a strong correlation between the forward voltage drop of a junction and the temperature of that junction. The point at which a junction becomes nonlinear with respect to junction temperature depends on the material of the junction, as well as its design. It’s safe to assume linearity for most materials in normal operating environments up to 80° to 100°C . Nonlinearity can be determined experimentally by measuring the voltage at higher and higher ambient temperatures until there’s a deviation from linearity. This relationship is nearly linear for most devices. It can be expressed mathematically as:
TJ = (m ´ VF ) + T0 (1)
where TJ = junction temperature in °C
m = slope (a device-specific parameter) in °C/V
VF = forward voltage drop
T0 = intercept (a device-specific parameter) in °C
Therefore, at a given temperature ( TJ ), the semiconductor junction will have a specific forward voltage drop ( VF ). If we measure the VF at two different temperatures, we can calculate the slope (m) as well as the intercept ( T0) for a particular junction. Then because this relationship is linear, we can use Equation 1 to find the junction temperature under different conditions simply by measuring the VF .
Knowing TJ for different operating conditions and packages enables us to calculate thermal parameters such as thermal resistance for differing package types and designs. This is important when you design a particular operating condition to ensure the maximum lifetime of a device, because thermal effects are major contributors to early device failure.
In this test technique, the device-under-test (DUT) is placed in a temperature chamber and connected to the drive and measurement equipment. This may be a programmable current source and voltmeter. But other instruments can source a current and measure a voltage simultaneously. Commonly known as source measure units (SMUs), they can simplify the instrumentation considerably (Fig. 1) .
Next, connect the SMU to the device using a four-wire, or Kelvin, measurement technique. Four-wire voltage measurements decrease the error caused by lead resistance in a voltage measurement by sensing the voltage around the DUT, rather than sensing it at the input to the SMU. Figure 2 shows the details of a four-wire measurement.
Place the DUT into an environmental chamber and set the chamber to an initial temperature. The initial point is commonly measured at 25 °C , and then the DUT is allowed to reach thermal equilibrium. A dwell time can be determined experimentally. But for most packages, a soak of 10 minutes should be sufficient.
Once the junction reaches thermal equilibrium, a short-duration current is sourced into the DUT and the voltage drop is measured. The time duration of the pulse and the amplitude are very important. Delivering a larger amount of power (too much current or too long a pulse) may skew the results by heating the junction.
Many times, the junction-under-test is a silicon or compound diode. For these device types, a good starting point for experimentation is 1 ms of sourced current at a few milliamps of drive current. If you’re unsure, the self-heating of the junction also can be determined experimentally by using a source capable of very short pulses (less than 1 ms). You can then experiment by varying pulse widths and comparing the voltages of each pulse duration. Voltage differences on the order of 1 to 2 mV commonly indicate a 1 °C change in junction temperature. This measured voltage is the VF1 at TJ1 (25 °C ).
The temperature is then elevated to a higher value ( 50°C , for example), the DUT is allowed to reach thermal equilibrium, and the current pulse is again delivered. Voltages recorded at this temperature are labeled the VF2 at TJ2 ( 50° C in this example).
These steps can be repeated over a number of values, then plotted as voltage versus junction temperature (Fig. 3). Use at least three temperatures in the analysis to check for any discrepancies in the approximation. You can now calculate the slope (m) of the line as well as the intercept, using Equation 1:
TJ = (m × VF ) + T0
TJ2 – TJ1 = m( VF2 – VF1) (point-slope form of Equation 1)
m = (TJ2 – TJ1)/(VF2 – VF1) (2)
and you can then calculate T0 by extrapolation:
TJ2 – TJ1 = m(VF2 – VF1) (point-slope form of Equation 1)
By setting VF2 to 0, Equation 2 becomes:
TJ2 = TJ1 – m VF1
TJ2 in this case is equal to the intercept, or T0 .
T0 = TJ2 = TJ1 – m VF1.
Real World Example: High-Brightness LED
In this example, a new high-brightness LED die is being developed. The device is designed to carry more current than previous units, and we need to ensure high heat flux to minimize the junction temperature. This will certify adequate device lifetime for some of our more rigorous applications.
A common LED failure occurs when the bond wire connecting the anode or cathode to the LED die is broken. The common cause of breakage stems from temperature cycling of the bond wire, which results from elevated junction temperatures created by inadequate heat removal.
We place the LED die in an oven and follow the above prescribed test plan. We measure the following results:
VF1 at TJ1 (25 °C ) = 1.01 V
VF2 at TJ2 (50 °C ) = 0.78 V
m = (50 – 25 )°C /(0.78 – 1.01)V = – 108.70 °C/V
T0 = TJ1 – mVF1 = 25 °C – (-108.70°C/V) × (1.01V) = 134.79 °C
Therefore, the first-order equation describing the junction temperature versus forward voltage for this device is:
TJ = (-108.70°C/V) × VF ) + 134.79°C
We can now vary other aspects of the evaluation, such as operating current, environmental/case temperature, and packaging, and simply measure the V F to determine the actual junction temperature.
Sources Of Error
The largest source of measurement error comes from the measurement uncertainty of the temperature in the environmental chamber. This measurement is typically made using a thermocouple, which can have errors of ± 2 ° or more. Additional accuracy can be obtained by placing a more accurate thermal-measurement sensor, such as a thermistor or resistor temperature detector (RTD), near the DUT and using a separate digital multimeter to measure the temperature.
Uncertainty in the voltage measurement also adds to the error when calculating the junction temperature. Selecting an instrument with a high degree of accuracy and resolution on voltage measurements is the key to minimizing this error.
Errors in the junction temperature measurement can propagate to other thermal calculations, such as thermal impedance and resistance. Therefore, minimizing these errors is critical to obtaining accurate measurement results.