Results of a power supply failure survey

We here at Ridley Engineering recently conducted a survey with a group of almost 3,000 active power supply design engineers to discover the various reasons and circumstances they have encountered power supplies that failed. The experiences the engineers shared with us were very enlightening, and this article summarizes our findings.

Power Supply Failure Survey

Eight months ago, we created the LinkedIn^TM site “POWER SUPPLY DESIGN CENTER” [1]. As of January, 2014, there were almost 3000 members registered, many of whom have more than 30 years design experience with switching power supplies. Being a very active and vocal group, it has become a valuable resource for design help, as well as keeping up with the changes in the industry.

While power supplies are a mature technology, they continue to be plagued with field failures. When something stops working in an electronics system, the power supply is the usual suspect. In order to find out why this is still a problem in our industry, we went directly to the subject in question with the LinkedIn group:

“Why do power supplies Fail?” To keep the survey simple, five choices were offered for vote. The response was very positive, and a lively discussion followed that is still going on now. Figure 1 shows a bar chart of the results of the survey.

Click image to enlarge

   Figure 1: Survey Results for the Cause of Power Supply Failures    

Responses in order of dominance:

Power semiconductors. 36%. This is not surprising. The semiconductors are placed under extreme stress in power supplies and are usually operated very close to their maximum capability.

Magnetics. 26%. The majority of power magnetics are custom-designed with hand-wound construction. This can lead to many issues. In a future article, we will discuss these issues in detail.

Capacitors. 15%. All capacitor technologies have some fundamental flaws if they are not properly designed into the circuit. This will also be discussed in a future article.

Mechanical and PCB Issues. 11%.

Control Circuits. 10%.

Power Semiconductor Failures

While semiconductors topped the list above, this should not lead to the conclusion that they are poorly-built components. They are simply in the firing line of anything that might go wrong in the power supply design, and are almost always collateral victims.

Consider the circuit in Figure 2. Two fets are connected in series with a low impedance input source. This could be a switch and synchronous rectifier circuit for the front end of a buck converter, or the two fets forming a half-bridge switch pair from an offline source. With this arrangement, there is a tremendous amount of power capability available.

Click image to enlarge

Figure 2: Power Supply Circuit with Two FETs in Series with a Voltage Source.

For example, with a 12 V input and two 1-mΩ fets, if both parts are turned on at the same time, the peak power dissipation is 72 kW. When driving synchronous rectifiers or half-bridge circuits, the intent is to drive the fets with as close to 50% duty cycle as possible. Small amounts of noise or controller errors can lead to inadvertent overlap of the drives, resulting in a short circuit. The results are usually catastrophic for semiconductors. In this case, the semiconductors are not the cause of the failure, but the result of improper operation.

A follow-on survey was conducted to find out what causes the high incidence of failures in semiconductors. The results of this survey are shown in Figure 3.

Click image to enlarge

Figure 3: Survey Results for Causes of Semiconductor Failures in Power Supplies

Responses, in order of dominance, are as follows:

Operating beyond specifications. 67%.  Either electrically or thermally, this category dominates the observed failures. Overvoltage on the drain, or overvoltage on the gate are typical examples of how the devices can be electrically overstressed. Thermal failures of the semiconductors are caused by inadequate cooling, or an insufficient package size. Neither the overstress or the thermal failures are usually the fault of the semiconductor suppliers. This is a problem with the design of the power supply itself.

Inadequate specs on datasheets. 10%. The unspecified failures have a wide range of possible causes. For example, the reverse recovery characteristics of the antiparallel diode of a fet may be inadequately detailed for an application.

Other. 11%. In this category falls the serious problem of counterfeit parts. Several recent conferences sponsored by the Society of Automotive Engineers and others have highlighted the growing problem of counterfeit parts entering the supply chain. There are many horror stories, including brokers who ship reels of parts where the first few feet of parts on the reel will be legitimate parts and the rest will be counterfeit. This problem is multiplying. Engineers are often called in to fix “design problems” caused by inferior devices.

Summary 

The survey results in this article highlight the major causes of failures. Specifically, we discovered the causes perceived to be dominant factors experienced by a large group of engineers.

It is easy for newcomers to the field to be overwhelmed by the wide array of problems they face in design. We will continue conducting surveys within the peer groups available to us to uncover new sources of information and assistance in solving these problems. By discussing our experiences with others and sharing results, we can only hope to become better and more efficient design engineers.

Ridley Engineering

References

[1]       LinkedIn group “POWER SUPPLY DESIGN CENTER” www.linkedin.com/groups?gid=4860717

[2]       Ridley Engineering website www.ridleyengineering.com

[3]       Power supply workshops and training www.ridleyengineering.com/workshops.html