Thermal Characterization of Plastic Ball Grid Array Packages via Infrared Thermography

A methodology for performing characterization of Plastic Ball Grid Array (PBGA) packages via Infrared Thermography (IRT) was established. The thermal performance of various assemblies was characterized using IRT in conjunction with analytical methods to determine the individual contributions of the PBGA package and of the printed wiring board. Accuracy and repeatability of temperature measurements via IRT were found to be equal to those obtained by thermocouple techniques using standard JEDEC methods. Additionally, IRT was found to provide significant advantages relative to conventional measurement techniques including increased resolution, space and time domain profiling capability, ease of use, and reduced sample preparation and measurement time. Introduction Device surface temperature measurements necessary for thermal characterization are conventionally performed with thermocouples. Use of Infrared Thermography (IRT) allows greater flexibility and ease of measurement with accuracy that compares favorably to conventional techniques. We first established the validity of the IRT technique. A standard calibration, sample preparation and measurement procedure was developed using a commercially available infrared camera to measure surface temperatures. This method proved to accurately track the temperature of known sources. Next, IRT was used to characterize several types of Plastic Ball Grid Array (PBGA) packages on JEDEC standard Printed Wiring Boards (PWB). Data from conventional measurements of identical components were used to check the accuracy of the IRT method. IRT was also used to measure the thermal performance of PBGAs on an application-specific PWB. Thermographic data from standard JEDEC boards and the application specific boards were analyzed to determine the source of contributions to system thermal resistance. The results support a straightforward predictive model that accounts for the thermal performance of several PBGA package types, and for thermal impedance due to the PWB. Description of Samples PBGA Packages: The parts measured were 217 I/O packages with three different substrates. The standard 217 PBGA has two metal layers of 1 oz copper. The "multilayer" 217 PBGAM has two signal layers of 1 oz copper and two planes of 1 oz copper. The "multilayer thermally enhanced" 217 PBGAMT has two signal layers of 1 oz Cu and two planes of 2 oz copper. The PBGAs have a 23mm square (529 mm) body size with a 217 pin periphery ball array. The package has a 1.17 mm over-mold thickness and a substrate thickness of 0.36 mm, 0.54 mm, or 0.61 mm for the PBGA, PBGAM, and PBGAMT parts respectively. Each of PBGA, PBGAM, and PBGAMT packages were assembled with 5.7 mm square (32.5 mm) commercially available thermal die mounted in the center of the package. The part was heated by driving a DC current through a resistor that uniformly covers the surface of the die, while internal junction temperature was monitored by observing voltage across a diode under forward bias. Additionally, a group of PBGAMT devices were prepared with a 3.8 mm square (14.4 mm) thermal die, reflecting the characteristics of a nextgeneration product. These samples are referred to in the text as "Small Die MT" packages. JEDEC PWB: Each of the four types of test parts was mounted onto JEDEC standard thermal printed circuit boards. Measurements were taken on boards with one surface signal layer and two buried solid planes (also called "1s2p", or "fourlayer" boards). The two embedded planes are each 1 oz copper, while the top trace layers contain 2 oz. copper. The board itself was constructed from FR-4 material and measured 1.57 mm in total thickness. Application PWB: PBGA and PBGAMT type parts were mounted onto "six-layer" application specific boards provided by a customer. For the purposes of this study, devices mounted on these boards are referred to as "custom." Applicability of the "One Resistor" Model In the standard "Two Resistor" device thermal model (see Figure I-1), it is assumed that heat generated at the die flows through one of two paths: into the package case or into the system board. From case or board, heat is transferred to the ambient environment. At steady state, capacitive effects can be ignored and each element is considered as a thermal resistance (Θ), with units of C/W. Appendix I summarizes the subscripts assigned to indicate the location of each resistance. The Kirchoff rules used in electrical engineering also apply to thermal impedances, and their application leads to the following formula: ΘJA = ((ΘJC + ΘCA)*(ΘJB + ΘBA))/( ΘJC + ΘCA + ΘJB + ΘBA) The "junction-to-ambient" resistance ΘJA is commonly used as a measure of the total thermal performance of the system. Other common benchmarks include the correlation factors (Ψ) between Junction temperature and Board or Case temperature. The Ψ factors differ from true resistances in that they depend on the distribution of power, P, among the two parallel branches of the circuit: 0-7803-7038-4/01/$10.00 (C)2001 IEEE 2001 Electronic Components and Technology Conference ΨJB = (TJ – TB) / Ptotal ΘJB = (TJ – TB) / Pboard ΨJC = (TJ – TB) / Ptotal ΘJC = (TJ – TB) / Pcase Where Ptotal = Pboard + Pcase ; and with the ratio of Pboard to Pcase determined by the ratio of the resistances of the two paths. In many cases, the relative resistance of the path through the package case is so high that it is essentially an "open" circuit. In this situation, the "One Resistor" model holds (Figure I-2), and the thermal resistance ΘJA becomes simply the sum of ΘJB and ΘBA. Further, the thermal correlation factors ΨJB and ΨBA become valid approximations for the true thermal resistances, respectively, ΘJB and ΘBA. The value of the one-resistor model is that it allows for prediction of system thermal impedance ΘJA for any system consisting of a device with known ΨJB and a board with known ΨBA. As a consequence of Fourier's Law of conduction, ΘBA and ΘCA are inversely proportional to surface area. For small footprint packages mounted on relatively large PWBs, it is generally assumed that ΘCA is sufficiently large to warrant use of the one-resistor model. A primary goal of this study is to empirically test the hypothesis that the 217 PBGA systems under test will meet this criterion. Method Infrared cameras allow for passive, non-contact, real-time temperature measurement with a high degree of accuracy and precision. To perform quantitative measurements, the optical properties of the surface to be measured must be considered carefully. In order to properly correlate an object’s radiative output to its true temperature using Stefan's Law, the material’s emissivity (ε) must be known at the wavelength band that the camera uses. Emissivity is a scalar factor between 0 and 1 that compares the material’s actual radiative output with that of an ideal blackbody. Objects with inferior optical properties can be coated to allow accurate thermographic measurement. Emissivity calibration: The effective infrared emissivity of coatings was determined by applying the coating sample to a copper slab. A T-type thermocouple was epoxied inside a hole drilled in the slab. The sample was heated on a hot plate to several temperatures between 25°C and 125°C. At each setting the sample was allowed to equilibrate for over 15 minutes. An IR image was taken, and the emissivity adjusted until the IR reading matched that of the thermocouple. Once emissive coatings such as black tape were established, they were used to perform a secondary calibration of other materials. A coating is placed on a small region of the sample, leaving another region exposed. The sample is heated and allowed to thermally equilibrate. Then the emissivity of the sample material is determined by comparing the difference in apparent temperature of the exposed surface to that of the known coating. Generally, emissivity is a function of temperature and of wavelength. Due to the detection method, the camera cannot distinguish between the temperature and wavelength dependencies, and both appear as an effective temperature dependence. An ideal surface for imaging would display no temperature dependence. Such a surface is called a greybody. Not only must the emissivity of the surface be known, but ideally should be as close to 1 as possible. Surfaces with low emissivity also have low photon absorption, so the surfaces are likely to generate excessive noise due to reflection and transmission. To qualitatively monitor reflectance, observe the sample through the camera perpendicular to a surface. If the object is directionally reflective, an image of the camera’s cryogenically cooled interior optics will be clearly visible. Transmissivity can be measured, either qualitatively or quantitatively, by suspending the sample in front of a source of known temperature. We have found that some brands of vinyl electrical tape are appropriate coatings for steady-state measurements. These tapes have nearly ideal optical properties, low thermal resistance, and they are not electrically conductive. Certain black spray paints also perform well optically, and are useful for transient measurements due to their low heat capacity. IRT measurements taken with these coatings have been found to regularly agree with thermocouples to within 0.2 K. FIGURE I-2 FIGURE I-1 0-7803-7038-4/01/$10.00 (C)2001 IEEE 2001 Electronic Components and Technology Conference Temperature ε (paint) ε (tape)