Real-time Thermographic Methodology with High-accuracy Temperature Monitoring Applied to Stacked Package of High-power Semiconductor Laser Diode

1Department of Electrical and Electronic Engineering, Chung Cheng Institute of Technology, National Defense University, Tauyuan County 33000, Taiwan, R.O.C. 2Materials & Electro-Optics Research Division, National Chung Shun Institute of Technology, Tauyuan County 32599, Taiwan, R.O.C. 3Department of Industrial and Systems Engineering, Chung Yuan Christian University, Tauyuan County 32023, Taiwan, R.O.C.


Introduction
Over the last two decades, an uncooled thermal camera has been widely used in the industrial, medical, and defense fields. (1) Generally, it is mainly applied to long-range target tracking, surveillance, and temperature measurement. Previously, the adoption of a highend uncooled thermal camera was limited owing to its expensive micro-infrared optics, the requirement of high spatial pixel resolution, and issues concerning optic depth resulting in the need for precise Z-axis positioning. In particular, for the measurement of temperature on a microzone of 50 µm scale or smaller, a thermal camera must be used with phase-locked microscopy and synchronized thermography. Therefore, this system has been very costly.
In this study, we used a commercial room-temperature-operated long-wavelength infrared imaging module incorporating a lens with a focal length of 19 mm and a lens sleeve designed to reduce thermal stray radiation, to shorten the distance between the lens and the object measured (from 50 to 2.85 cm). Hence, the spatial resolution approaches 8.5 μm, providing a zoom-in function to observe the temperature of a microzone containing the interfaces of a submount solder for a laser diode (LD). With the monitoring of the temperature of its housing and the operating temperature of the internal focal plane array (FPA) in the thermal imaging module under a controllable ambient temperature, the imaging module can be used with the linear regression approach for thermal radiation correction at different steady-state ambient temperatures. For example, in the microzone temperature measurement, the highpower semiconductor LD is regarded as the exciting laser trigger source. Also, the LD is the critical chip. Because of heat dissipation due to defects of the high-power LD, excess heat at the interface of the attached region cannot be removed efficiently, i.e., laser output would be decreased. In this study, a microzone temperature-monitoring platform with a real-time temperature calibration algorithm was proposed and applied to the package on the submount stage of a high-power semiconductor laser diode (HPSLD) to evaluate its power efficiency. (2,3) Because the size of the temperature monitoring scale is only 100 µm-2 mm, it is suitable for the above needs, that is, it is not necessary to use an expensive lock-in thermographic camera.

Experimental Setup and Temperature Monitoring Methodology
Using a commercial long-wavelength infrared thermal camera with an IR optic lens having an object distance from 50 cm to ∞, we design an optically stray-radiation-suppressing sleeve to shorten the distance from the FPA to the optic lens, for microzone imaging and accurate temperature measurement. The temperature accuracy approaches a value less than ±1 ℃. The thermal camera with the optically stray-radiation-suppressing sleeve is shown in Fig. 1.
According to Planck's Radiation Law, the radiant exitance or emissivity of an object is related to the surface of the material itself. Using Planck's spectral distribution, a thermal signal is converted to an electrical signal by a thermal detector to generate a thermal distribution pattern visible to humans. For the proposed temperature detection technology, the FPA based on a microbolometer is used to measure the thermal absorption that causes a change in resistance, and then a voltage or current (numerical intensity) signal is obtained. The voltage or current signal from the microbolometer is subsequently converted using the numerical fitting equation of Planck's distributed curve. The reverse substitution is conducted to obtain the actual temperature (T 0 ) as (4) where R, B, F, and O are the main thermal radiation regression coefficients. The actual temperature is obtained as where O indicates the offset value of numerical intensity, which leads to the parallel translation along the y-axis of the numerical fitting equation of Planck's distributed curve; R is the response of the thermal camera, which is related to the system sensitivity R s ; B is related to the effective wavelength sensitivity of the thermal camera; F is the nonlinear calibration factor, which is generally set to 1.

Temperature Calibration of Thermal Camera Module
Because of the difference in the processes for the microbolometer and FPA, which are affected by the operating temperature, the signal output from a readout integrated circuit (ROIC) contains some numerical error that results in an inaccurate temperature measurement. Thus, using a proper calibration methodology, the accuracy of temperature measurement must be improved. In other words, ambient, shutter surface, and indoor FPA temperatures are monitored to determine the thermal radiation regression coefficients R, B, F, and O. (5,6) The thermal camera module based on the microbolometer is placed in a temperature-controllable box (the temperature ranges from 0 and 50 ℃). The two flat plated blackbodies are put in front of the thermal camera module. The radiant exitances from the two flat plated blackbodies tuned at two different temperatures for the derivation of R and O in Eq. (2) of the thermal camera module in the temperature-controllable box are determined for the accurate calibration of temperature. (7) The following are the calibration steps. 1. Two blackbodies of low and high temperature ranges (0-50 and 50-500 ℃) are used for high-and low-temperature-point calibrations. 2. The temperature-controllable box is set up with temperatures in the range of 15-45 ℃ at intervals of 5 ℃.
3. At each testing temperature interval, the overall ambient temperature is in the steady state after the target temperature is held for 20 min. Then, the nonuniformity is processed. 4. At the same time, the temperatures at the surface of the optic shutter and the indoor FPA are recorded. 5. For the low and high temperature ranges, 30 and 90 ℃ blackbodies are set up and V D is recorded automatically over 2 min (120 data sets are obtained). The obtained data are averaged and shown in Fig. 2. Although the average counts show some deviation, the curves are generally monotonic.  It is obvious that the accuracy of temperature measurement is very high and may be better than ±1 ℃. It is also interesting to note that the maximal errors are larger for lower blackbody temperatures than for higher blackbody temperatures. This result could be explained by the measurement of lower target temperatures (in this case, blackbody temperatures of 20 and 50 ℃) being more easily disturbed by the ambient temperatures of 15-45 ℃. Figure 9 shows the experimental setup for monitoring the temperatures of multimetal solders of the LD bar and MCC stage. The thermal camera module with the above temperaturecalibration methodology is used and mounted on the triaxial optical positioning platform. Owing to the very high power coherent exitance of the LD bar upon serious damage of the microbolometer FPA behind the IR lens, the optic path is off-axis by 30°, which includes the normal direction of the focal plane of the thermal camera. The optical imaging distance between the objective and the IR lens is about 2.85 cm. The digital imaging data and spatial       intensity are transferred into NI-PXI through the RJ-45 port of Giga-Ethernet for microzone functional thermal imaging and real-time calibration associated with high-precision temperature measurement. Figure 10 shows the thermal image of the high-power LD bar corresponding to the schematic diagram of the Type 1 package. There are two positions where the temperatures are monitored: the beam surface of the LD marked by "+" and the junction below the hard solder, i.e., the topmost surface of joint multimetal layers, marked by "×". In this research, we use two kinds of joint mounted structure for the package of the high-power LD with the peak wavelength of 808 nm. One comprises Cu/AlN/Cu multimetals soldered with Au 80 Sn 20 , as shown in Fig. 10; the other comprises submounted Cu 10 W 90 sandwiched with two 5-µm-thick Au 80 Sn 20 solders defined with the Type 2 package, as shown in Fig. 11.

Real-time Temperature Analyses on Two Types of Submounted LD Bar
For Type 1 package, the temperature of the beam surface on the edge LD bar approaches 43 ℃ as the driving DC current increases to 34 A. In contrast, the temperature of the Cu/ AlN/Cu metal joint is only 24.5 ℃, which is close to the ambient temperature of 22.5 ℃, as shown in Fig. 12. The higher DC driving current indicates a larger temperature difference between the beam surface on the edge LD bar and the Cu/AlN/Cu metal joint. This means that once the lasing process begins, the heat absorbed by the LD bar is not conducted effectively into the Cu/AlN/Cu metal joint mounted in the water-cooled MCC module, resulting in the decreasing efficiency of LD power. However, for the Type 2 package shown in Fig. 13, the temperature of the beam surface on the edge LD bar remains at only 47.7 ℃ as the driving DC current increases beyond 90 A. The temperature of the joint on the sandwiched Cu 10 W 90 approaches 44.5 ℃, which is higher than the ambient temperature of 23.0 ℃. These differences in temperature between the beam surface of the LD and the joint on the sandwiched Cu 10 W 90 are almost the same. This can be explained by the submount package providing the better heatconducting junction for the removal of heat generated from the high-power LD bar. The peak wavelength suddenly shifts from 803.25 to 804.75 nm when the DC driving current exceeds 55 A. Furthermore, a driving current higher than 85 A causes the peak wavelength to shift to 805.45 nm. It is concluded that the hop mode appears in the single-mode LD owing to the changes in operation temperature. A high power of 86.6 W and a peak wavelength of 805.25 nm can be obtained under the DC driving current of 90 A. By real-time microzone temperature monitoring, the changes in the temperatures of the beam surface of the high-power LD bar and its joint submount are resolved well. The temperature-driving current curve provides the optimal choice of metal joint materials and the process of Type 2 package on the junction of the high-power LD bar.

Conclusion
An optical zoom-in and stray-radiation sleeve was slipped onto an IR lens with a constant focal length, and the proposed correction methodology for the regression coefficients R, B, F, and O of a microzone temperature-monitoring thermal camera was used for real-time temperature measurement. In our research, we have used a commercial room-temperatureoperated long-wavelength infrared imaging module incorporating a lens with a focal length of 19 mm and a lens sleeve designed to reduce thermal stray radiation, to shorten the distance between the lens and the object measured (from 50 to 2.85 cm). Hence, the spatial resolution approaches 8.5 μm, providing a zoom-in function to observe the temperature of a microzone containing the interfaces of a submount solder for an LD. The accuracy of temperature measurement is higher than ±1 ℃. Applying the proposed microzone temperature-monitoring thermal camera for the evaluation of the package of an HPLD, we clearly observed the real-time temperatures of the junction of multimetal joints and the beam surface of the LD. In the future, by realizing heat distribution and thermal crowding on the junctions of the submount and LD to optimize the joint materials and packaging process, a high-performance package for the highpower LD bar or array will be obtained.