Thermal imaging maps device heat dissipation
Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 5/1/2006
As the features in electronic devices get ever smaller, power consumption and heat generation become an increasing problem. An engineer with Dell Computer once explained to me why portable computers are now called "notebooks" rather than "laptops." "You can't put them on your lap anymore," she remarked. "They get too hot."
Semiconductor devices have undergone numerous design changes to reduce this effect. The development of dual-core microprocessors was partially driven by the need to increase the device's performance without overheating it.
At some point, it becomes necessary to assess the effect of generated heat on device quality. Merely measuring overall device temperature, however, is not enough. You have to map the heat distribution across the die surface to determine where the circuit is overheating, thereby permitting design or process modifications to minimize or eliminate the problem.
In fact, in a paper presented at the InfraMation 2005 conference, Volodymyr Malyutenko, from the Institute of Semiconductor Physics in Kiev, Ukraine, emphasized the necessity of creating such a high-resolution map of excess heat dissipation on a device's active area—which can be less than 0.1 mm2 (Ref. 1). Malyutenko contends that the infrared cameras that manufacturers use to measure overheating of printed-circuit boards (PCBs) offer only a static low-resolution picture and cannot pinpoint physical (thermal) reasons why devices fail.
Malyutenko proposes a transition from a static map to one that shows micromapping parameters across the device's active area and changes over time. He and his colleagues developed a high-resolution multi-spectral IR (<1 µm, 3–5 µm, and 8–12 µm) "vision" facility, applying IR microscopy to IR thermal imaging. They used the apparatus to perform heat (T-approach), light (L-approach), and emissivity (å-approach) tests on semiconductor devices. The researchers applied the technique to measure a variety of devices, contending that uneven light and heat distribution across the device active area are difficult to predict or avoid.
The T-approachTo create a thermal image of a semiconductor surface, you must measure the IR power (P) that the surface emits. This technique identifies hot regions on the device and graphs their patterns across the surface over time. The characteristics of these hot regions depends on thermal mass, thermal conductance, and the design of the heat sinks of the surrounding package. For example, the active-area temperature can significantly exceed the heat-sink temperature in places, which can cause lower-than-expected device efficiency as well as overheating at specific points on the device surface and can subsequently lead to device failure.
 |
| Fig. 1 (a) This single micro-emitter produced the image in (b), captured with an 8–12-um camera at different bias currents. (c) This image depicts the profile for a device whose resonator cavity was deformed by Joule heating.Courtesy of Volodymyr Malyutenko, “What is Hot in IR Micro Vision,” Proceedings of InfraMation 2005. |
Malyutenko and his team found that for mapping the hot spots precisely and accurately, two-dimensional thermal resistor arrays on electrically heated pixels provide the most successful IR scene projectors in the 3–12-µm range. Figure 1 shows these micro-membrane structures, which can measure object temperatures up to 500°C.
Unfortunately, the structures have many layers and tiny legs—used as a membrane's electrical contacts and mechanical supports—that present their own challenges. The sensing membrane sits 2.5 µm above the silicon surface whose temperature is being measured. The optical interference created by the resulting space permits the spectral emission band to be tuned. Thermal and mechanical stresses between layers dramatically reduce the IR power that the combination emits, making the IR "microscope" a much better tool than a conventional optical unit.
 |
| Fig. 2 (a) This image depicts excess LED temperature captured through a sapphire substrate. (b) The profile shows two “hot spots” on the surface. (c) This image shows local overheating of a degraded device.Courtesy of Volodymyr Malyutenko, “What is Hot in IR Micro Vision,” Proceedings of InfraMation 2005. |
Figure 2 depicts a typical heat pattern across an active device area. Figure 2a shows a pulse duration of 160 ms, a 25-Hz repetition rate, and a current of 200 mA. The temperature profile in Figure 2b—obtained by increasing the current to 800 mA and shortening pulse duration to 400 µs—clearly shows the two points of maximum temperature rise, corresponding to two symmetric "heat traps" near the crossbar of a contact fork. Although the local temperature spike at that point of more than 50°C appears reasonable, Malyutenko and his colleagues have estimated the maximum local heat density at a staggering 7000°C/cm!
Other techniquesIn his paper, Malyutenko proposes two alternatives to the T-approach. The first technique involves interband luminescence (L), which appears in the fundamental absorption spectral band and expands over the 3–5-µm and the 8–12-µm spectral ranges, permitting monitoring by thermal-imaging cameras. The paper proposes a way to provoke luminescence from semiconductors through electroluminescence or photoluminescence.
A thermal-imaging camera interprets luminescence caused by excess free charge carriers (electrons) as a dynamic increase in temperature. Similarly, a charge-carrier deficiency ("holes") appears as a negative luminescence, which the camera sees as a temperature drop. The accuracy of this technique depends on keeping the actual surface temperature constant so that the charge-carrier status represents the only variable.
Another technique takes advantage of the device's emissivity (å), which depends on the device's reflectivity and its transmissivity. In this case, the device emits IR power that depends on the spatial (charge-carrier diffusion length) and temporal (charge-carrier lifetime) evolution of the charge carriers. The researchers contend that this approach permits studying an electronic or optoelectronic device's properties as well as its performance.
When faced with the need to increase processing power while minimizing power consumption and heat dissipation, PCB manufacturers need a way to accurately measure the thermal properties of semiconductor devices both at localized points on the surface and as a function of time. Malyutenko and his colleagues have shown several approaches to this characterization that can predict device behavior and performance in the target products as well as indicate process changes that can minimize the thermal effects.
| Reference |
|
