-
- News
- Books
Featured Books
- smt007 Magazine
Latest Issues
Current IssueIPC APEX EXPO 2024 Pre-show
This month’s issue devotes its pages to a comprehensive preview of the IPC APEX EXPO 2024 event. Whether your role is technical or business, if you're new-to-the-industry or seasoned veteran, you'll find value throughout this program.
Boost Your Sales
Every part of your business can be evaluated as a process, including your sales funnel. Optimizing your selling process requires a coordinated effort between marketing and sales. In this issue, industry experts in marketing and sales offer their best advice on how to boost your sales efforts.
The Cost of Rework
In this issue, we investigate rework's current state of the art. What are the root causes and how are they resolved? What is the financial impact of rework, and is it possible to eliminate it entirely without sacrificing your yields?
- Articles
- Columns
Search Console
- Links
- Events
||| MENU - smt007 Magazine
Step 2 Process Control
December 31, 1969 |Estimated reading time: 21 minutes
By Robert Rowland and Tom Woody
The path to defect prevention starts with a good process quality plan. Process quality determines product quality, whose main goal is to limit and reduce variations toward creation of a healthy process. This plan should outline the methods used to monitor products and processes.
Figure 1. Defect categories. The pie chart is an effective way of displaying such data.
Many companies are reactive to the incidence of product defects though they may think of themselves as proactive because most of their efforts go into counting defects rather than measuring processes. In fact, counting defects is reactive because the damage already has been done. Measuring the process is more proactive because it is a method that helps prevent defects. To be sure, companies with good quality control plans do count defects. But more importantly, they measure their processes and focus on defect prevention. Following is a discussion of process control concepts and tools, inspection methodology, and quality metrics, beginning with a quality plan separated into four levels:
- Preventive Action and Awareness. The process is in or out of control. It is either capable of producing acceptable product or not. If the latter, one proceeds to the next three levels of activity.
- Failure Analysis. The skills and tools are available to effectively analyze and investigate typical process and product problems.
- Problem Solving. Structured problem solving skills and tools (e.g., cause-and-effect diagrams, Pareto charts, etc.) are available to isolate the root causes of problems.
- Corrective Action. With such causes uncovered, permanent solutions can be found and implemented.
Basic Process ControlProcess control will help achieve and maintain a desired level of process capability, stability and repeatability. Process control is more than statistics; it is a detailed sequence of events (Table 1). If the events are followed, process control will be achieved.
All assembly processes have some variation. Statistical process control (SPC) provides tools to measure the process and monitor the variation, of which there are two forms: common cause and special cause (as defined by Deming). A common cause variation is naturally inherent in a stable process, i.e., while not perfect, a stable process nevertheless is repeatable a key requirement for measurement. In contrast, a special cause variation is the result of a specific action such as an execution error (e.g., the documented process was not followed). Thus, unless a process is repeatable, the result (as displayed on an X bar R chart) will reflect operator variation rather than process variation. With an out-of-control process possibility in mind, engineers sometimes attempt to make a process perform perfectly before measuring it to prevent getting a low "score." However, such a step is somewhat self-defeating; SPC should be used to stabilize new processes and to improve existing ones.
Finally, basic process control must be based on facts obtained from accurate measurements and observations. Control charts, which simply are histograms spread out over time, record and display the data. Variable data (e.g., adhesive dot diameter) are process focused and obtained by measurement while attribute data are obtained by observation and are product focused. Typically, collecting and analyzing attribute data are reactive actions while collecting and analyzing variable data are more proactive steps.
The ToolsProcess control relies on statistical tools for measurement, feedback and analysis. Commonly known as the seven quality control (QC) tools, they are essential parts of a process quality plan because they can be used to collect, display and analyze data:
Check sheets, used to collect data, also can display data. They are simple, easy to apply and effectively show the frequency of various results.
Pareto charts display defect categories in descending order on a histogram, useful in separating the vital few from the trivial many.
Cause-and-effect diagrams focus attention on the most likely causes of a defect and their potential interrelations. Attention typically is focused on four categories: methods, materials, people and equipment.
Flowcharts illustrate the various steps in a process using standardized symbols. A flowchart clearly illustrates the entire assembly process from start to finish.
Histograms are bar charts displaying statistical distributions. They are helpful in showing data dispersal, which makes it easy for comparison and analysis.
Scatter diagrams plot numerous data points to display cause and effect between two different, yet interrelated, characteristics of a process or product.
Control charts display attribute (counted) or variable (measured) data over time and between actual and imposed upper/lower limits. They clearly reveal abnormal patterns, trends and cycles that may affect process capability. There are six basic charts:· X bar displays the variation in the average of a measurement series· R displays the variation in the range of a measurement series· C displays the variation in the number of defects· U displays the variation in the number of defects per unit· P displays the variation in the fraction defective· Np displays the variation in the number of defective units.
InspectionSome level of product inspection is necessary and desirable. However, evidence is clear that in the long run, mass inspection does not improve product quality. Inspection is a screening process that endeavors to find unacceptable products that need repair. Mass inspection should be avoided unless an event has occurred that necessitates 100 percent product screening. For every-day inspection, it should be based on a reasonable sampling plan. Deming frequently urged mass inspection abandonment. Instead, he encouraged stable and repeatable processes that are monitored statistically.
Defining inspection methods and tools is important. Too often it is assumed that people know how to inspect and, as a result, most create their own methods, resulting in inspection variation. Inspection should be regarded as a process to be defined and documented like any other process. Once accomplished, inspectors should be trained and certified. In practice, inspection can be separated into three basic models:
· When every component on every board is checked (mass inspection)
· When specific components on every board are verified (reduced mass inspection)
· When specific components on specific boards are examined (sample inspection).
Figure 2. FPY vs. DPMO. On the basis of the close correlation between practice and theory, a conversion to DPMO as a primary metric for PCB defect measurement was established.
Ideally, a company progresses from the first model to the last. Unfortunately, manufacturing operations more likely start with the first model and remain there. Printed circuit assembly complexity affects inspection as the average number of components per board continues to increase, rendering it more difficult to properly inspect every component on every board. In fact, complexity is a main reason mass inspection fails to improve product quality. One solution is to reduce inspection fatigue by decreasing the number of components to be inspected. In theory, it is desirable to detect and fix all problems immediately; in reality, it simply is not possible.
Focal-point inspection is an example of progressing through the models. Based on historical data (i.e., Pareto chart analysis), components demonstrating a higher defect rate are inspected more frequently. This approach ensures that troublesome components get more attention. Components with low defect rates are inspected less frequently or not at all, improving the overall detection process.
Inspection vs. MonitoringInspection is product focused; monitoring is process focused. A good process quality plan should include both. However, the long-term goal should be less inspection and more monitoring. This is desirable because product inspection is reactive (the defects have occurred) while process monitoring is proactive (the defects can be prevented), which is a value-added concept. The more difficult task, however, is creating a balanced inspection and monitoring strategy that does not rely on mass inspection.
Figure 3. DPMO vs. manufacturing date. The chart clearly shows how effectively the primary metrics are monitoring a board build operation.
Defect CategoriesDefect codes, such as solder bridge, solder open, insufficient solder and missing component, are used to simplify data collection and analysis. At one company*, these codes are further classified into defect categories, pinpointing areas needing additional improvement. A pie chart, as shown in Figure 1, is an effective way of displaying the data. There are four defect categories:· Process capability means that the process cannot achieve the desired result. For example, an adhesive dispenser that consistently will not create the desired dot size.· Process execution, in which the process is capable, but overall execution is poor. For example, a machine operator fails to follow the written procedure.· Defective material, e.g., leads on a fine-pitch component do not consistently meet the coplanarity specification.· Design for manufacturability means that a board is not designed for assembly ease. For example, the fiducial target is designed incorrectly.
Process Control MythsMyth #1: Process control and SPC are the same. Process control is a concept; SPC is a set of tools that supports the concept. Before SPC can be an effective tool, the process must be stable and repeatable (but may not be perfect).
Myth #2: Creating and displaying charts and graphs is process control. Displaying charts simply to impress management and customers is a waste of time. Using charts and graphs to prevent defects and solve problems is part of process control.
Myth #3: Mass inspection is an effective process control tool. Mass inspection is not process control. Rather, it simply is an attempt to sort acceptable and unacceptable product. A process quality plan should eliminate the need for mass inspection.
The Art of MeasurementMetrics is the science of measuring performance. For example, for first-pass yield (FPY), a metric reveals how well a process or operation is performing, usually against a desired result or expectation. Therefore, metrics are important because a company must possess the capability to measure its performance. Metrics should be reviewed and analyzed frequently daily in some cases, weekly in others. The information gathered is not just for management but should be shared with engineers, technicians, supervisors, operators, etc. To be of value, metrics must be based on facts and truly reflect actual performance. During analysis, management must promote an atmosphere of openness and honesty with no fear of criticism or retribution. It is very important to ensure that all individuals and shifts collect data using the same format and actually respond to their findings.
FPY vs. DPMOAs mentioned, one measurement frequently used is FPY. Traditionally, the success of a surface mount manufacturing line is gauged by this measure. The sequence of operations typically is from paste print through in-circuit test (ICT), which includes adhesive dispense, component placement, reflow soldering, through-hole assembly, wave soldering, cleaning and second assembly (press-fit connectors, swaging, addition of other components, wires, etc). FPY is defined as a measure of how many printed circuit board (PCB) assemblies were completed successfully (with no defects), divided by the total number of assemblies produced. Mathematically, this is stated FPY% = [good boards/tested boards] x 100.
In a typical high-mix factory with medium- to high-complexity boards, FPY can range from 50 to 80 percent or more depending on equipment age and type, inspection methodologies, and process control techniques used.
Figure 4. The quality star one system for communicating how selected metrics are affecting production quality.
One problem with FPY measurement is that it only deals with the number of good vs. bad boards; the number of defects associated with the bad boards is ignored. This is a concern for the high-mix board manufacturer when it comes to deploying resources to fix yield problems with various products. It also is possible to have an FPY number that appears at first analysis to be acceptable, perhaps more than 90 percent, while rework and labor costs are out of proportion because a few boards have several defects per board. This can occur, for example, if a placement machine feeder changeover is performed incorrectly, or if solder paste printing is inadequate for a short period. While affecting only a handful of boards, the number of defects produced in a short time can be considerably greater than the average per hour.
The trend recently has been to move from FPY as a measure to a more realistic metric known as defects per million opportunities (DPMO). For this metric, the number of opportunities equals the number of parts (SMT or through-hole) plus the number of solder joints (for components only; vias or test points are excluded). The DPMO calculation provides a means of normalizing the defects to board complexity, i.e., a board with fewer opportunities, for the same DPMO, will have fewer defects, whereas a highly complex assembly with many more opportunities will have more defects for the same DPMO number (DPMO = [defects/opportunities] x 106). The relationship to FPY is inversely linear: As the number of opportunities increases for a fixed-process capability, FPY will decrease proportionately.
When this metric was first encountered in 1999, it was decided to test the theory against actual practice on a select number of assemblies in production. Boards that had at least six months of production history, were relatively stable and in volumes of at least 500 units or more were selected. Figure 2 depicts the result. The P/N as shown is plotted in increasing DPMO to the right, showing a correlation with the theory. In this example, the average DPMO for the factory is approximately 55 over a range of 20 products that represents 65 percent of the total volume. On the basis of this correlation between practice and theory, some companies* have converted to DPMO as a primary metric for surface mount operation.
Long-term Trends Bar ChartsOne way to communicate the overall trend on a metric is to use a bar chart vs. time, but on a compressed horizontal scale as shown in Figure 3. The idea is to summarize yearly or quarterly data via a single bar and to show more current information on a monthly or weekly basis. In this way, the chart can provide data over a long period while more recent data can be shown in a manner to provide nearby history without being too busy.
For example, on a specific product, the conclusions from the chart would be:1. Earlier in 2000, DPMO is below target (which is good).2. Then, mid-year, something happens to cause DPMO to exceed the target.3. After corrective action, DPMO starts trending back toward the target.
Of course, there may be many explanations for the problem. The chart, however, shows very easily, over time, the actual performance and, with annotation, it can be a powerful tool for communicating results to all involved.
Dashboard Metrics and the Quality StarMeasuring performance is one thing; effectively communicating the data and trends to all affected participants so they can use the data is quite another. For example, the "quality star" method comprises five basic metrics that monitor a company's health and manufacturing operations. These include:
1. DPMO.2. Dead on arrival products (DOA) products that do not function within the first 30 days (when turned on by the customer).3. Warranty failures products that turn on initially but fail at some point during the next six months and are returned for repair.4. Functional test yield an FPY calculation in which products are tested and either pass or fail functional test.5. Quality audit also an FPY calculation includes products audited against visual criteria, such as IPC-A-610 and cosmetic workmanship standards.
If performance is good on all these fronts, then internal goals as well as customer expectations for quality, reliability, cost and performance should be met.
An example of a quality star is shown in Figure 4 (gold is "good," blue is "bad"). The "radar" graph in Microsoft Excel produces the chart and is based on a set of agreed minimum and target goals for each metric. The minimum would produce a quality star with a point that is completely blue, as illustrated in Figure 4 for quality audit. Meeting the target would produce a quality star with a point that is completely gold, as illustrated by warranty, test or DOA. Partially meeting the target will produce a quality star with a point that is blue and gold, as in the DPMO metric. For example, the DPMO target may be 65, which if achieved, would find the point of the quality star filled as shown in Figure 4. If, however, the actual result is 80 (which is poor compared to the target) then a partial fill will result.
Similarly, the other parameters can be set up to graphically represent the desired range of minimum-to-target performance. Once the target goals and minimum acceptable performance is communicated to employees, the quality star becomes a tool to communicate daily, weekly, monthly, quarterly or annual performance data to all.
SummaryQuality improvement begins with a plan, a set of objectives against specific criteria, tools by which to make measurements, a means to make changes for the better, and a way to communicate results. This article outlines the basic components of such a quality system for producing high-quality PCBs and a means to demonstrate progress against performance targets in terms all affected personnel can understand.
- RadiSys Corp., Hillsboro, Ore.
REFERENCES
- "Best in Class Process Quality Benchmarks," CEERIS Report, March 2000.
- Peter Mears, Quality Improvement Tools and Techniques, McGraw-Hill, 1995.
ROBERT ROWLAND, process engineering manager, and TOM WOODY, quality systems manager, may be contacted at RadiSys Corp., 5445 NE Dawson Creek Dr., Hillsboro, OR 97124; (503) 615-1100; Fax: (503) 615-1115; E-mail: rob.rowland@radisys.com and tom.woody@radisys.com.
___________________________
ABCs of SMTAccuracy: The difference between the measured result and the target value.Automatic Test Equipment (ATE): Equipment that automatically analyzes functional parameters to evaluate product performance.Bed-of-nails Test: A method (fixture) employing vacuum activation and an array of spring-loaded pins to make electrical contact with specific circuit nodes on a board.Boundary Scan: A self-test method that permits testing via a built-in four- or five-pin test bus accessing input and output pins.Defect: A deviation from the normally accepted characteristics for a device or unit.Fault Profile: A description of the type and frequency of electrical defects most likely to be found on a populated board assembly.Functional Test: Automatic equipment for testing a finished PCB via application of inputs and the sensing of outputs through the board's edge connector.In-circuit Test (ICT): A system that measures component values on a loaded PCB before power is applied.
___________________________
Getting Started In ESD
To effectively combat and prevent electrostatic discharge (ESD), one must use the right equipment in the right way. Thanks to a range of powerful closed-loop ESD protection, monitoring and ionization equipment, it now is possible to treat ESD as a process control problem.
By Simon Hawkins
ESD is a familiar and underestimated source of board and component damage in electronics assembly. It affects every manufacturer, regardless of size. Although many believe they are producing product in ESD-safe environments, in reality ESD-related damage continues to cost the world's electronics industry billions of dollars annually.
What exactly is ESD? ESD is defined as a discharge (flow of electrons) to or from a charge (deficit or surplus of electrons) that formerly had been static (immobile). A charge is immobile under two conditions:
1. When it is "trapped" on a conductive but electrically isolated object, e.g., a metal screwdriver with a plastic handle.
2. When it resides on an insulative surface (e.g. plastic), over which it cannot flow.
However, if an electrically isolated conductor (screwdriver) with a sufficiently high charge is brought close to an integrated circuit (IC) at a different potential, the charge "jumps" causing ESD.
ESD occurs very rapidly with great intensity and typically will produce enough heat to literally melt the interior circuits of a semiconductor chip at a microscopic level (under an electron microscope the appearance is like small bullet holes blown outwards), causing instant and irreversible damage.
Worse still, in only one in 10 cases will this damage be catastrophic enough to cause a total component failure at final test. The other 90 percent of the time, ESD damage will cause only a partial degradation meaning a damaged device could pass final test completely unnoticed only to be doomed to an early field failure when shipped to the customer. The result is most reputation damaging and an expensive place for a manufacturer to rectify any manufacturing fault.
The main difficulty with controlling ESD, however, is that it is invisible at the level that can damage electronics components. A relatively large charge build-up of approximately 2,000 V is needed to produce a discharge that can be heard as a "click," while 3,000 V is required to be felt as a small electric shock, and it takes 5,000 V to be seen as a spark.
Yet common devices such as complementary metal oxide semiconductors (CMOS) or electrically programmable read-only memory (EPROM) chips, for example, can be damaged by an ESD potential difference of only 250 and 100 V respectively, while an increasing number of sensitive modern devices, which include Pentium processors, will be killed by just 5 V.
The problem is compounded by the everyday activities that cause damage. Walking across a vinyl factory floor, for example, creates friction between the floor surface and shoes. The result is a net electrical body build up charge of somewhere from 3 to 2,000 V, depending on the relative humidity of the local atmosphere.
Even the friction created by the natural motion of workers at a bench can create 400 to 6,000 V. If the workers have handled insulators during the unpacking or packing of PCBs in foam-lined boxes or bubble packs, the net charge build up on the surface of their bodies can reach approximately 26,000 V.
As the main source of ESD damage, therefore, all staff entering the electrostatic protected area (EPA) must be grounded to prevent any charge build up, and all surfaces should be grounded to maintain everything at the same potential to prevent an ESD occurrence.
The primary products used to prevent ESD are wristbands with curly cords and dissipative work surfaces or mats both of which must be grounded correctly. Additional accessories such as dissipative footwear or heel-straps and suitable clothing are all designed to prevent personnel building up and retaining a net electrical charge as they move around the EPA.
During and after assembly, PCBs also should be protected from ESD arising from both internal and external transportation. A range of board packaging products is available for this task and includes ESD shielding bags, tote boxes and mobile trolleys.
Although the correct use of the above equipment will guard against 90 percent of ESD-related problems, to achieve the final 10 percent demands reaching the next tier of protection: ionization.
The most effective way to neutralize assembly equipment and surfaces that can generate static charges is by using an ionizer a unit that blows a stream of ionized air over a work area to neutralize any electrical charge build-up on insulative materials.
It is a common fallacy to believe that because a wristband is worn at a workstation, any charge carried by an insulator in the area, such as a polystyrene cup or cardboard box, will be dissipated safely. Insulators, by definition, do not conduct electricity and it is impossible to discharge them except by ionization.
If a charged insulator remains within an EPA, it will radiate an electrostatic field that will induce a net charge onto any nearby object, thus increasing the risk of ESD damage to products.
Although many manufacturers attempt to ban insulative materials from their EPAs, this practice is very difficult to enforce. Insulative materials are too much a part of everyday life from the foam cushion operators sit on for comfort to the build information they are referencing in a plastic cover.
With the use of ionizers, manufacturers can accept the fact that that some insulators will appear within their EPAs. Because ionization systems continuously neutralize any charge build up that may occur on insulators, they are a sound investment in any ESD program.
Ionization equipment in standard electronics assembly comes in one of two basic formats:· Benchtop (a single fan)· Overhead equipment (which essentially is a series of fans together within a single overhead unit).
Room ionizers also are available but currently remain the domain of clean room environments.
The choice depends on the size of the area to be protected. A benchtop ionizer will cover a single work surface, while an overhead ionizer will cover two or three. Another benefit is that ionizers also prevent dust from electrostatically adhering to product, which can degrade cosmetic appearances.
No protection program is complete, however, without regular testing and monitoring of the ESD equipment's effectiveness. Leading ESD control and ionization specialists report instances of manufacturers employing failed (and therefore useless) ESD equipment without any knowledge of the failure.
To combat this situation, in addition to the standard ESD equipment, ESD suppliers offer a wide range of constant monitors that automatically alarm if the performance of an item falls out of specification. The monitors can be used as standalone units or linked together in a network. Networking software also is available for automated data acquisition and the real-time presentation of system performance, related to both the operator and the workbench.
Monitors can simplify an ESD program by eliminating many routine tasks, such as ensuring that wrist bands and curly cords are tested properly on a daily basis, ionizers are balanced and correctly maintained, and grounding points on benches are undamaged.
ConclusionThe first step to combating ESD is to appreciate how small details can cause irreparable damage if overlooked. An effective program demands not only the use of effective ESD-protective equipment, but also tight operational procedures to ensure that the behavior of all factory floor staff is ESD-safe.
Although many manufacturers use automated wristband testers, it is all too common for operators to either bypass the test or fail it altogether because, for example, wristbands are too loose. Many operators then will try to pass the test by simply gripping the tester closer to their wrist with their other hand.
However, the good news is that ESD is avoidable. The time and money invested in both the right equipment and improved safety procedures will be rewarded by a corresponding yield increase.
ACKNOWLEDGEMENTThanks to Semtronics for help in preparing this article.
SIMON HAWKINS, European product manager, may be contacted at Semtronics, a Div. of Metcal, 12151 Monarch St., Garden Grove, CA 92841; (714) 799-9810; Fax: (714) 799-9764.