Abstract: Knowledge storage is the most important when troubleshooting a complex device. We want and need to understand some related issues. It includes the correct IC version number, where can I find relevant reference materials, and who really knows what happened to the client. Helping customers is our primary concern, and IC failure analysis requires a quick and correct response. But should we expect that the quality assurance department can test every parameter under all conditions during the failure analysis? No, it’s impossible, it’s too late. Most are guessing analysis. This may cause some people to be surprised, but people in the quality department also don't have crystal balls or mind reading. The only possibility for timely and effective IC fault diagnosis is to get the correct information about the IC failure from the client.
Failure analysis of the IC - it is a waste of time
We often hear that “concepts are real.†When the IC fails or the customer thinks it has failed, we must do a failure analysis. In order to be more effective, we must get accurate and relevant information. This is the only way to avoid guessing.
Let me tell you an episode that happened a long time ago. A device was returned due to failure and we did not get any relevant information. We have performed a series of verifications on this device, such as running automatic test equipment, bench test, X-ray inspection, and opening inspection. We immersed the device in a soft electronic environment and placed it under an electron microscope to observe the damage. We use a liquid crystal coating to measure the temperature. This device is intact. We did not find any cause of failure. So the quality department wrote it in the report, we want to know why this device was returned due to failure?
After about two months, we almost accidentally learned that when the device is heated above +60 ° C, the customer's product will fail. We did the failure analysis again. We tested this device at room temperature (+25 ° C) and found nothing. When the device is destroyed in the test, it no longer has any function. In the end, this is a return event and it didn't happen again. But there is a more important lesson here: there is no key failure data, we can only be a blind person, and we can only guess. We waste a lot of time and money. (See Appendix - IC Failure Analysis in China for another more personalized antique car story, grounding issues with another failed IC.
Thorough testing is futile for QA
Multiple damaged ICs can not confirm the root cause of the damage. The customer took a board from the general contractor back to the lab. In the lab, they took the IC off the board and declared the chip to be ineffective. It is likely that the customer will come to a conclusion that the root cause of the failure is due to the chip itself. They want us to do a failure analysis, but where is the invalid data? Was the situation at the time carefully recorded? How to prevent future failures? We turn back to the previous guess, not the actual view - it's hardly a good recipe for effective failure analysis.
In this case, the customer concentrated on three pins of multiple output devices. This is what we know: there are several departments and billions of certifications when the device leaves the factory; it works for hours before it expires. Will it be an initial failure or damage due to external operations? Is it already in the customer's circuit? Is it working in a reference circuit environment? Is the factory's ESD weakening circuit causing subsequent failures? Perhaps this is because the transporter has neglected ESD and caused chip damage? The list of possible factors seems to be endless.
It is not very helpful to receive the initial schematic from the customer. It shows nothing that causes the device to fail or show that it needs to be changed. The FAE needs to check the way the ground is handled. Is it correctly segmented? From the schematic, you can't tell the connection. We received more schematics of the page, but now the question is more than the answer. Why do customers only see three of the many outputs? Is any input or output of the device already connected to the pins on the board with a low impedance? Is the power supply and ground plane a low impedance connection? Is there a problem with ESD on the board? We are still guessing.
Effective failure analysis - troubleshooting is a crime scene
Now we want to ask, "How do you get the right information at the beginning?" Is it reasonable to expect QA to do every parameter under all conditions? Especially in the case of this failure we don't know anything. No. We will help customers understand why the IC will fail and correct it using the correct application circuit.
Obviously, this approach conflicts with ideas that the failure analysis should do right away. I have heard, "Failure analysis is often the first thing to do. Look at the inside of the IC before looking at the IC application circuit." I can't understand where this idea came from? I don't agree either. Failure analysis is not the first job to be done. Conversely, investigating “crime scenes†and failure events is the first step.
Information on fault location is critical, and like police investigators, we should do our best to protect the field data. The first thing is to look at the application circuit of the IC, ie where it fails. Such a simple thing, like solder splashing, may also be the real answer. The IC may be part of the work, not a complete failure. In fact, disassembling this IC may mask the real problem.
For a valid failure analysis, we need to look at the customer's schematic and collect all the failures, why it will fail. Yes. This process may encounter customer confidentiality issues. This is a common problem, which is why there is a confidentiality agreement. This is also the case with FAE as the factory's eyes and ears around the world. The FAE can view the customer's equipment and evaluate their schematics, wiring, and other conditions of the application. To protect the customer's secrets, the FAE only needs to send the relevant part of the customer schematic design to QA. Now, QA will use these trusted fault data for final processing.
Successful result
Back to our previous story, local FAE customers pay close attention to this failure problem. When there are more schematics in hand, we can see that an op amp is connected to an output pin, but due to the series resistance of 10kΩ, It should hardly work. By using a common ground, non-separatingly connected to a star point, the power supply noise is directly coupled, although the decoupling capacitors are connected to other power supplies. The minimum decoupling capacitor is 0.1μF, and the typical self-resonance of a surface-mount 0.1μF capacitor is about 15MHz, which is higher than its function as the frequency of the inductor and causes it to stop functioning as a capacitor.
There are two lessons from it. First, the decoupling capacitor is bidirectional. If a noisy power supply is coupled to a clean supply, the noise will contaminate the clean supply. Second, the same thing happens on noisy grounds: noise will contaminate clean power. A noisy power supply should be paired with a noisy ground wire, and a clean power supply must be paired with a clean ground wire. Cross-contamination can damage power and ground. The self-resonant frequency of the appeal makes it an inductance, that is, it no longer conducts or attenuates high-frequency energy.
in conclusion
We took a lap to repeat our initial view: knowledge is king in solving an IC failure problem. From the beginning of an investigation, no one can have a local FAE and customers to check the problem side by side is more valuable. The FAE must carefully examine the entire system, the layout of the board, the schematics and applications, and then pass the data back to QA. Only with these accurate and detailed data can we solve the IC failure problem. Without such data, QA can only be forced to guess the "crime scene."
Appendix - Failure Analysis at the rear
Here is a related example of why a proof of knowledge is king on a failed circuit. Without complete failure data, it is impossible to get an accurate FA. This example didn't solve an IC failure problem at first, but soon the IC failure problem was rolled in.
A friend has an old Model A Ford® car that was built between 1927 and 1931. He installed a radio bought from a local auto parts store, but it didn't work. He took the radio back to the store and changed it to a new one. After installing it again, the result still didn't work. After trying the "bad" radio three times, the merchant returned the money to him.
He told his antique car club members about his experience. They told him that the Model A car had a positive ground, so the radio reversed its power supply. When the radio is expected to be connected to the positive voltage, it is actually connected to a negative voltage. If the power supply on the semiconductor is reversed, the semiconductor will smoke.
The story of this Model A continues. Based on the knowledge of the positive ground, our friend bought an expensive custom DC-to-DC converter to reverse the supply voltage. For testing, he connected the battery to the DC-DC converter and radio on the workbench, and the radio started working. But when he embeds everything into the car, the fuse blows. Finally, he had to ask for help from an engineer friend.
The base of the Model A is connected to the positive terminal of the battery (in terms of today's electronics, it is equivalent to the negative supply). After 1956, the cars in the United States were all grounded with a negative iron. The negative terminal of the battery was connected to the base to generate a positive power supply. The consumer goods that are bought today in car stores assume that the car is a negative ground. Figure 1 below works on the workbench because the radio is not bolted to the base of the car.
Figure 1. This unit can work on the workbench because the base of the dotted line is not connected to the radio.
There is no isolation in the DC-DC converter to save cost; in fact, the positive input and the positive output are connected to the ground of the base. Therefore, when the device is on the workbench, it works because the base of the broken line portion is not connected to the radio. Once the radio is connected to the car, the radio is shorted to the power supply and the fuse is blown.
Suppose you are a radio company technician and are arranged to perform a failure analysis on these returned radios, and the local merchant will only tell you what he knows: "I don't work after loading it." Then you turn on the radio and find a lot of burnt Bad device. What is causing these problems? If there is no more specific performance data, you can only guess. As we said, any QA needs a complete failure story to come up with an effective fix for the device.
Shenzhen Niimoo Innovative Technology Co., Ltd , https://www.niimootech.com