Demand Grows For Reducing PCB Defects

Board manufacturers are boosting their investment in inspection, test and analytics to meet the increasingly stringent demands for reliability in safety-critical sectors like automotive.

This represents a significant shift from the past, where concerns about reliability primarily targeted the devices connected to printed circuit boards. But as SoCs become disaggregated into advanced packages due to the rising cost and technical complexities of scaling chips, the boards on which those various components are mounted are becoming increasingly dense. Lines and spaces between wires are shrinking, and the high-density interconnect technology that connects all of these components is becoming more sophisticated.

The result is a growing demand for the same kind of advanced monitoring techniques found in leading-edge chips. This is a must-have for automotive, as well as for medical implants, avionics, and some industrial applications. Yet even consumer products such as smart phones and audio/video control boxes have begun to require higher quality due to the sensitivity of high-density interconnect (HDI) technology, as well as concerns about how social media can spotlight failures. In response, electronics board manufacturers and assemblers have increased inspection and invested in more data collection, which can be used in data analytic platforms.

“There have been three big changes,” said Craig Hillman, director of product management for new and emerging technology at Ansys. “First is the increasing use of inspection at each stage of the PCB/PCBA manufacturing process — especially at the earliest stages, like stencil printing. Second, there is increasing sophistication of the automatic optical inspection (AOI) being used. And third, instead of relying on ‘go/no-go’ decisions, there is increased use of parametric data to predict the presence of abnormalities,”

System applications requiring HDI PCB technology also come with defect detection requirements that extend well beyond just opens and shorts.

“Over the past decade, we have seen a growing trend of new PCB technologies for smart phones, communications, big data servers, and also for automotive PCBs,” said Micha Perlman product marketing manager for via formation at Orbotech, a KLA company. “The defects in the line quality of the PCB may impact any type of electronics, and they are particularly critical for industries such as aviation, medical, or automotive, where failure in the field brings a high risk to users and their surroundings. Manufacturers of consumer products such as smart phones and notebooks also are paying more attention to these types of defects.”

Demands for long-term reliability also are having an impact on board manufacturers’ sensitivity to scrap costs. “Mission-critical applications are now not allowing for end-of-line repair of boards,” said Dave Huntley, director of business development at PDF Solutions. “Finding and isolating problems early in the process is critical to reduce scrap. Having assembly and consumable data that is stored and can be correlated to test data enables this.”

All of this data provides a more granular view of a device, which can have a big impact on reliability. It also can be used to identify and isolate problems earlier in the manufacturing process.

Efforts are underway to facilitate data sharing across disparate equipment, allowing data to be collected and correlated across multiple processes. So instead of just testing or inspecting for defects, data analytic platforms more fully enable engineers to reduce defects.

PCB defect types
Solder joints remain the top source of defects both in-line and in the field, but those are not the only problem areas. Understanding the defect mechanisms provides an engineering perspective of why a simple open and short electrical test does not suffice in defect detection.

Between the board and component, the solder ball connections need to be aligned, and the solder-paste needs to be contamination free. A contact can pass electrical test- simple shorts/opens, yet fail in the field due to a misalignment or material issue that grows worse in the field. Consider a TV set-top box, for example.

“The unit was in somebody’s home and failed. The cabinet environment in which it was used became hot enough that a solder joint failed,” shared Michael Ford, senior director of emerging industry strategy at Aegis Software. “A misalignment between the component and board pad occurred, yet it was just within the manufacturing tolerance. But when you started applying a lot of power through it under a certain usage condition, combined with the higher than ideal temperature, the failure occurred.”

So when does a manufacturing variation — a yellow flag — actually cause a failure? It’s nearly impossible to tell without feedback from the field.

“Most of the stuff we’re finding cannot be detected by functional tests,” said Anna-Katrina Shedletsky CEO and founder of Instrumental. “We look at things that may be difficult to actually test for. Consider part of the shield can’t snap down all the way.”

Yet solder defects still can get through, which Instrumental discovered at box build when it looked at an aggregation of data. The company analyzed 100,000 defects from 4 million customer units (Figure 1). “One of the interesting things is that soldering is number 4 in the top 10. So we actually see a prevalence of solder defects that make it all the way through end-of-line inspection, which represents what percentage could be escaping into the field.”

Fig. 1: Top 10 defects in production (not skewed to new products). Source: Instrumental

With HDI technology comes an increase in open and short defects. Yet the need for high-quality traces drives metrology and inspection needs.

“Defects in quality such as line nicks, dish-down defects or abnormal pads are enormously critical in industries where safety and reliability are mission-critical, such as aviation, medical and automotive,” said Orbotech’s Perlman. “These defects in quality have far greater potential risk than open or short defects.”

HDI permits tens of thousands of tiny vias between layers, which translates into an equal number of opportunities for defects.

“Due to the high stress of the component assembly process or during the end-device’s operation, latent defects introduce high risk for future failure,” Perlman said. “Therefore, laser via inspection, relatively rare a decade ago, is more common today. It is in high demand from car electronics manufacturers that require high reliability and safety. Several automotive PCB experts recently told us that a significant part of the quality issues they currently deal with are due to HDI boards malfunctioning as a result of bad inter-layer connections through the laser vias. This is occurring despite the fact that the PCBs had successfully passed all the electrical tests prior to shipment.”

Inspect, collect, test
The board manufacturing process, and components-to-board assembly, are largely mechanical processes. To ferret out defects, board manufacturers rely upon visual inspection and end-of-line test.

Optical inspection has been part of PCB production for more than three decades, whether looking at images for proper solder joints or detecting issues on the inner layers of the board prior to lamination.

“Inspection has evolved from 2D cosmetic defects to 3D inspection and metrology,” said Subodh Kulkarni, president and CEO of CyberOptics. “From pass/fail reports to quantitative measurements, manufacturers now can not only conduct AOI and solder paste inspection (SPI), but also attain coordinate measurements with in-line coordinate measurement machine (CMM) capabilities that are much faster than with a traditional CMM system. The need for metrology data continues to increase so they can not only detect critical defects, but also measure critical parameters.”

Imaging systems for metrology purposes have become more common. “Direct Imaging (DI) ensures accurate registration between the pattern to the Z dimension connection through the vias,” said Perlman. “In addition, DI’s pattern imaging quality is very accurate and stable along the board, with minimum edge roughness. This supports the need for precise line imaging and etching for the demanding impedance-controlled transmission lines.” He added that AOI for laser vias has been implemented more and more by manufacturers.

While inspection has been the bread and butter of PCB quality control, manufacturers have begun to take a closer look at the data from solder paste inspection, pick-and-place, and other steps prior to electrical test.

“Historically, sensor data has focused on go/no-go parameters,” said Ansys’ Hillman. “Going forward, sensor data that is more parametric can be fed directly into reliability predictions. For example, currently systems exist that can measure the volume of solder paste depositions. Sensor data that can provide real-time measurements of solder volume can be used to predict fatigue lifetime. This would be greatly beneficial and would allow the manufacturers more discretion on what is good and what is bad.”

Getting access to that equipment data has motivated the Institute of Printed Circuits (IPC) to develop a standard called the Connected Factory Exchange (CFX). “This is an IIoT-based data exchange mechanism,” said Aegis’ Ford. “We have defined the exact data fields and what they mean for every type of technology, such that all machine communication is digitally modelled based on their operation and capability. Basically, any measurement or event that occurs, either visually or electronically, is transmitted through CFX, being reported on a real time basis in a single ‘plug and play’ language environment.”

This enables engineers to respond to real-time analytics, and it provides a larger data set they can analyze to understand yield and field failures.

Board electrical testing provides a key gatekeeping role for interconnect health. Beyond simple connectivity checks for opens and shorts, an opportunity exists to take parametric measurements — frequency, voltage, current, capacitance, resistance, and settling times, which provide insights into performance variation and can be tied back to manufacturing performance. The parametric information, coupled with system test data/field data, can better inform engineers about their manufacturing performance indicators.

This level of data has been lacking due to a focus on connectivity, decreasing access to signals, and test equipment limitations.

“In today’s PCB manufacturing lines there is a lack of good rich parametric test data,” said PDF’s Huntley. “For example, in-circuit testers (ICTs) are not logging most of the test data that they could. The reason is that it impacts test time significantly, which would create line bottlenecks. Also, in PCB end of line (EOL) testing, the testers are not sophisticated. There are wide variations across testers in measurement accuracy. This leads to increased scrap, and poor quality of parts shipped.”

After board-level manufacturing comes the test box build, and that can cause new problems. The mechanical process of enclosing the fully assembled board into its container can result in a physical change, which has an impact on the electrical performance. Consider the delicate antenna arrays found in products with wireless functionality, for example. Often referred to as “princess arrays,” a minor change of position can crater WiFi performance. If engineers fail to provide a RF/Wifi test at system-level test, these can turn into customer returns.

Connecting data between production steps
Data collected at each production step has value. In fact, connecting data all along the supply chain can be quite powerful. But there are some obstacles — lack of standards, data availability, connecting data via product traceability, and meaningful data that connects manufacturing steps.

With CFX providing a standard for data, factory management systems enable engineers to find the weaknesses in the board manufacturing and assembly process. That results in improved yield and quality.

“In the last two years, things are beginning to change in a new and different way,” Ford said. “Rather than being dependent on physical inspection to expose defects, all of which cannot be guaranteed to be found in most cases, we are now using six sigma-based trend analysis of contextualized data to augment that inspection or test process. Thus, at an earlier stage you can identify and eliminate potential causes of defect from the variation in all of the different processes, which reveals things that were previously hidden through optically or electronic test in isolation.”

Connecting data throughout the supply chain enables engineers to solve problems at the source rather than through end-of-line inspection.

“We’ve created a data platform so that our customers can ingest traceable product data,” Shedletsky said. “Data includes images from various stages, test station data, and performance test like ICT test. The platform then enables different stakeholders on the engineering side or the manufacturing side to get the information they need to actually solve the problem in the first place.”

For complex supply chains the source of an identified problem often can be upstream from where it is detected. Combining that data into one analytics platform enables engineers to solve the problem at the root cause, resulting in increased yield, reduced scrap, and improved quality.

However, data is not always consistent and available. “Most often data is shoved into closets (data lakes) without much sorting,” said Sam Jonaidi, vice president of automotive solutions at OptimalPlus, which is part of NI. “Data is also stored on the machinery to the extent of HDD storage available. Worst case, data is used only to make real-time decisions and then discarded. We prescribe a comprehensive data engineering discipline where the entire data universe is mapped and high-value data elements are bestowed into a data harness with a precise and monitored governance.

Both SEMI and IPC have developed standards for product traceability, which assists with connecting data across the manufacturing supply chain. Yet tracking raw materials also has value in faster root cause analysis of production issues.

“This could be as simple as keeping track of consumables like solder paste lots used, all the way to analyzing end-of-line test data to find semiconductor process issues that escaped the semiconductor testing process,” said Huntley. “It has been shown that by having all this data collected and correlatable, that reliability case avoidance can be improved by 50%.”

Conclusion
With both old and new PCB technologies, manufacturers have upped their investment in data collection, which in turn fuels the need for analytic solutions to fully leverage the data. Collecting and storing the appropriate data can lead to detecting defects, and comprehending their root cause can lead to reducing defects.

This represents a significant shift in this part of the electronics industry. “Ten years ago, PCB manufacturing was mostly oriented towards getting the right build recipes into the lines through the implementation of MES (Manufacturing Execution Systems),” said Huntley. “The focus was on making sure the machines were set up properly to put together all the pieces of the puzzle that ended in a completed board. The communications were more single-directional from MES to machine. Today, in the age of machine learning and AI, the flow is bi-directional. However, there remains a huge gap compared to semiconductor manufacturing. The huge amount of data that is now collected is not well stored or collected in a unified, consistent, smart way to be able to do cross operational correlations.”

Board manufacture engineers can’t change what they don’t measure. When it is measured, engineers should be able to analyze it.

Related Stories

The Need For Traceability In Auto Chips

Better Analytics Needed For Assembly