Many times, when a component supplier is asked to do an analysis on a failing component, the conclusion is electrical overstress (EOS). Unfortunately, this conclusion does not tell the user much about the cause of failure. EOS can be caused by an excess of voltage, current, or power. For best results, this conclusion should be followed up by further analysis to determine if the cause was excessive voltage, excessive current, or a combination of voltage, current, and power. The diagnosis of EOS does not give the designer any information that could lead to a design change or provide guidance for corrective action. It is only when further analysis is performed that a conclusion leading to correction of the problem can be determined.
EOS comes in many forms. Figures 1 and 2 show x-ray images of identical areas of a PCB assembly. Figure 1 shows a good PCB assembly, while Figure 2 shows an assembly that has been damaged.
In Figure 2, the copper conductor has been vaporized. In both figures these leads are connected directly to the outside world. In this case, the customer requested an analysis to determine why the assemblies failed. In order to vaporize the copper, (there were multiple places on the assembly that had been damaged with the copper vaporized) there must have been a lightning strike or a power-line cross to an unprotected network cable. The cause was external to the PCB assembly. We were able to tell the customer to add the necessary over-voltage protection to their network when the cables exit or enter a building. The relatively inexpensive protection devices can then eliminate much of the excessive power, although if the cause was a direct lightning strike it is unlikely that the power could be fully dissipated before it reaches the PCB assembly. (There is not much that can be done to protect equipment from a direct lightning strike except placing it inside a grounded metal container, and even then power can be coupled into the circuitry through the interaction of the magnetic field generated by the lightning and the cable itself.)
Figure 3 shows a different form of electrical overstress on a polymer fuse. A good polymer fuse (F4) and failed polymer fuse (near R36) can be seen in the image. This fuse has two resistive states. It is normally in a low resistance state which allows full current flow to power the device via power over Ethernet (PoE). It the current exceeds a preset amount, the polymer changes to a high resistance state, which stops the PoE device from operating. In this case, we were able to tell the designers that some PoE devices continued to operate when the fuse was in the high-resistance state. There was no indication that the fuse had gone into its high resistance state. This polymer fuse remained in its high resistance state causing a power overstress to occur. The damage to the fuse is extensive with some thermal damage extending to the PCB around (see Figure 4). This indicates that the power dissipation and heat generation was present over an extended period of time, possibly a few weeks. Neither the voltage nor the current exceeded the device’s specifications, but the combination of the two caused excessive power dissipation in the fuse due to the increased resistance of the device itself.
Figure 5 shows a damaged resistor. This damage was isolated to the conductive film and the glass protective coating. The damage to the glass is noted by the cracks in the glass, easily seen on the “7” marking and at the bottom of the “R” and “7” markings. The protective glass had been broken by the expansion of the resistive element and any gas developed during the EOS event. Since the damage was isolated to the resistive element and the thermal energy did not spread to the glass passivation or the ceramic, it can be concluded that the event was nearly instantaneous. The cause of this failure was traced to inrush current when power was initially provided to the circuit. The designer changed components to one that could withstand the higher input rush current at startup.
A simple conclusion of EOS would not have been of much help in resolving these failures. However, further investigation uncovered additional facts that were actionable for corrective actions. In each case the type of damage and the extent of damage to the components allowed a rough estimate of the time involved in causing the damage. It was nearly instantaneous in the case of the resistor and lightning strike and slower in the case of the polymer fuse.
