Process control keeps faults in check
By using an optical inspection system to monitor your production line, you can reduce PCB manufacturing defects and cut rework costs.
Pamela Lipson, Imagen and Landrex Technologies, and Lyle Sherwood, SynQor -- Test & Measurement World, 12/1/2005 2:00:00 AM
|
|
READ OTHER DEC./JAN. ARTICLES:
Contents, Dec. 2005/Jan. 2006 DEC. 2005/JAN. 2006 FEATURES: The Best in Test Process control keeps faults in check Scan test drives yield Ensuring power supply accuracy The light and the distance |
Most electronic assembly lines produce some defects, or faults. For example, solder-paste printers may deposit excess, smudged, or insufficient quantities of solder, and component pick-and-place equipment may improperly place or orient components. Such faults reduce first-pass PCB-assembly yields, sometimes over short periods, but more typically over extended times (Figure 1).
To catch these defects, many manufacturers use automated optical inspection (AOI) equipment to identify problems that workers then fix at the end of the assembly line. This tactic is fine when a company can absorb the cost of rework and field failures, but for PCB manufacturers who compete in a market with small profit margins, these costs can be unacceptable.
PCB manufacturers would be wiser to use AOI to help line managers acquire quantitative inspection information that they can use to adjust their processes and reduce defects in the first place. First-pass production of PCBs with fewer defects cuts costs, saves time, and reduces the need for rework employees. By using an AOI system to perform process monitoring, manufacturers can:
-
identify quality problems that develop during the startup of a new production line,
-
maintain high-quality production on an existing line, and
-
shorten setup times required by changes.
|
| Figure 1. A typical first-pass PCB-assembly yield pattern shows an increase in defects over time. |
For example, one manager who used an AOI system to acquire and analyze data from 1 million PCBs, assembled during a five-month period, decreased defects tenfold. AOI data let the manager adjust production variables as they occurred and before they could cause defects. These slight process corrections consistently kept the number of defects low.
AOI data can also help a line manager stabilize the assembly of a new PCB type so PCBs are assembled properly the first time. We have heard from managers who report a 50% reduction in the time needed to stabilize production of a new type of PCB as a result of acquiring and analyzing inspection information.
Catch the drift
To uncover the root cause of defects and take corrective action, a line manager must determine whether defects occur at random or as the result of production-system problems. Using an AOI system, a manger can gather quantitative optical-inspection data that illuminates patterns and pinpoints the source of problems.
Many engineers know about optical-inspection attribute data that produces "go/no-go" decisions, but they may misunderstand the need for variable data, which comes from quantitative measurements of solder volume, part placements, and so on. To provide useful results, an AOI system must supply variable data with a tenfold increase in repeatability and accuracy over that found in assembly equipment, such as a pick-and-place machine. A post-placement AOI system installed to inspect for the placement drift of a 0402-size SMT device, for example, must offer a precision of 100 µm and a repeatability of ±10 µm. That ratio represents a Gauge Repeatability and Reproducibility (GR&R) value of 10%.
|
| Figure 2. By relating defects to a specific machine, line managers can make corrections and increase yield quickly. |
During the past seven years, we have used the Optima 7200 AOI system from Landrex Technologies to profile the PCB-assembly processes and analyze the post-placement processes at many companies. The analyses encompass many types of PCBs, such as those used in cell phones, PCs, and servers. The Optima 7200 operates as a post-placement inspection system that provides attribute data, such as missing part, wrong part, and wrong label. The system also precisely measures component positions to furnish variable data. Careful analysis of the data, at several "levels," can quickly improve first-pass PCB production yield.
Until recently, though, many companies that assemble PCBs have not collected this type of data. Data collection and analysis take time, talent, and money. Often, production-line managers and AOI experts spend so much of their time trying to keep a production line up and running that they have little time to gather data for subsequent systematic analysis. Newer AOI systems, though, offer high-precision measurements that will help line managers translate accurate inspection information into lower PCB defect rates and higher profits.
Our work has confirmed that the use of optical inspection of PCBs during production can reduce defects. These five examples illustrate ways in which PCB manufacturers used AOI to increase yield:
-
Machine-level analysis. Even a simple analysis of AOI data can reveal "macro" defect patterns. During a case study, a production-line manager inspected 1 million PCBs assembled by three separate parts-placement machines. After analyzing assembly defects—based on attribute data—over several months, the manager found one placement machine caused 71% of the defects (Figure 2). When the production staff adjusted the machine to reduce the defects, the first-pass yield increased. In this case, the AOI system tracked specific defect types and their sources.
-
Part-level analysis. Sometimes, the root cause of a decrease in PCB-assembly yield relates only to a few parts. During a two-week period, we analyzed the AOI data from a run of PCBs. This attribute data indicated the proper vs. improper positioning of components on a PCB. The Pareto chart (Figure 3) shows that three parts—the MICRO8 ball-grid-array (BGA) package, the 0805C-CT tantalum capacitor, and the EMT3-x small outline transistor (SOT)—caused the most defects. When line managers adjusted the production equipment to properly place these three components, PCB-assembly defects declined by 80%.
|
| Figure 3. A Pareto chart plots defect types from largest to smallest numbers so managers can solve problems in order of how they affect quality. |
-
Part-level analysis on one board. During inspection, we acquired variable data for the positions of the same components used to plot the Pareto chart in Figure 3. An analysis of the data revealed details about the expected position of the components and the positions at which the AOI equipment found them. In this case, we reviewed all the part-placement information for one type of PCB. A scatter-plot diagram (Figure 4) shows the difference between the expected and the actual part locations. (The scatter plot shows PCB reference designations rather than package types.)
|
| Figure 4. A scatter plot shows the difference between expected and actual part position as measured by AOI equipment for one board. |
-
Each red dot on the chart represents a component on the PCB, and each rectangular label calls out a specific component class. The inner and outer rings indicate the 4-mil and the 8-mil process limits, respectively. From the plot, you can see five component classes—H11, U2, J5, U4, and C33—cause the most defects. Production equipment placed one or more of these parts more than 8 mils away from its expected position. If you look closely at the plotted data, you will see components labeled J3, U6, U5, and H115 do not yet cause gross defects during PCB assembly because they fall within the 8-mil process-control limit. But component-placement equipment has put them more than 4 mils away from their expected positions. Without any corrective action, placement of these components may drift outside the 8-mil limit, which will cause assembly defects. The chart in Figure 4 also shows that the component-placement equipment puts the majority of parts off center. If managers do not correct this centering problem, they soon may experience many more defective PCBs as placements drift farther from the center "target" position. Although not plotted here, analysis of data acquired from several of the same PCBs produced the same variable-data "signature," which indicated a system-level production problem.
-
Part-level analysis across boards. Although the plot in Figure 4 suggests a centering problem, information such as this may not tell the whole story. Further investigations showed the component-placement equipment always placed an SOT "off pad" in what seemed like a random manner. When we examined the Äx and Äy positional data for the SOT on many PCBs, however, we saw something different: The production equipment placed the SOT at a constant displacement in a radial direction around its target position. To production managers, this "signature" indicates a bent nozzle used to place the component. (A bent nozzle will place a part a fixed distance from its intended location, but at a different radial position that depends on the rotation of the nozzle at the time of placement.) After we identified the root cause of the defect, the line manager quickly solved the problem.
-
Board-level analysis across assembly lines. One common hypothesis holds that two production lines outfitted with the same equipment and set up to build the same PCB should yield similar results. To test this hypothesis, we ran an experiment that placed two identical AOI machines on two seemingly identical production lines that assembled the same PCB design. Based on the acquired data, we characterized the manufacturing process on each line.
|
| Figure 5. a) Production Line 1 showed many defects, but most part placements took place within the 4-mil target. b) By comparison, Line 2 showed fewer overall defects, but poorer parts-placement accuracy, which may cause more defects in time. |
-
The two production lines, Line 1 and Line 2, produced different defect profiles that resulted from different causes. Overall, Line 1 produced about three times as many defects as Line 2. The plotted data for a series of boards shows equipment on Line 1 (Figure 5a) placed the majority of the parts well within the 4-mil tolerance window. In contrast, Line 2 (Figure 5b) produced a few defects just outside the 8-mil tolerance window, but all parts showed a high variability of placement. Line-2 production equipment placed the majority of components 2–6 mils off center, so this line does not adequately control the component-placement process. Although Line 1 may seem more problematic, because of the number of far-ranging defects it produced, you can expect Line 2 to produce more problems in the long run due to the wide variations of placement differences. But this cloud has a silver lining: You can use the variable-process data to catch and correct large and subtle defect trends before you end up with a lot of incorrectly assembled PCBs.
These examples show how production-line managers have taken a strategic approach to process control. By combining their knowledge of manufacturing processes and quantitative trend information from inspection equipment, they can keep PCB-assembly lines under control, which in turn reduces defects and manufacturing costs.
AOI systems simulate human brain
02/01/2007Camera concepts come clear
11/01/2004False failures still vex AOI
09/30/2002Multi-line production alters test economics
08/31/2001Scan for ESD-induced errors
07/31/2006
-
FLIR offers IR camera for under $3000
-
Don't let the economy compromise quality (Guest commentary)
-
Danaher speeds up restructuring, acquires life-sciences businesses
-
Agilent’s Cover-Extend technology eliminates need for physical test points for in-circuit test
-
So many combinations: Testing a switch-matrix board
View All Webcasts
