Ten Steps to “Defect Engineering”
Last month, I introduced the Ten Steps to Defect Engineering or Technical Systems Management program and promised to expand on it this month because I consider it a valuable tool, a systematic method with which to attack contamination control problems. It could also be easily expanded into a project management tool for getting the message across to management, as well as a very important tool in your own career advancement. This month, I will go through each of the steps in some detail. They are:
1. Process Outline Flowchart
2. Project Targets
3. Technical Targets
4. Measurement Systems
5. Process Improvements/Timelines
6. Yield, Quality, Reliability, Customer Satisfaction, and Market Dynamics
8. Roadmap of Future Requirements
9. Implementation Plan
10. Iterative Cycling
Process Outline Flowchart
The first step is to construct a complete flowchart of the process. For example, let`s look at the etching of silicon oxide in semiconductor processing. The process flow is considered in a series of three flowcharts. The first flowchart (Fig. 1) considers the basic steps of the oxide etch process. The second flowchart (Fig. 2) considers the materials used in the actual process, and a third flowchart could consider the environmental factors affecting the process. Additional flowcharts can be used as complexity increases.
The second step is to establish fairly broad project targets that outline the major goals of the project. In the case of a semiconductor manufacturing project, the major goals would include: plant size, yield targets, wafer size, cost targets, deadlines, and checkpoints. This section would require more detail for a totally new startup than for changes to an existing facility.
The third step is the generation of technical control targets. These are targets that define the level of control required by a particular manufacturing process. If selected correctly, they will define the minimum control limits necessary to achieve the desired manufacturing yield, quality, reliability, and cost. Targets that are too loose will not achieve product goals; targets that are too tight make a product too expensive and, therefore, noncompetitive.
Most targets are based on previous experience, current manufacturing results, supplier input, information from technical associations, and literature studies, coupled with engineering judgment. Targets are a starting point and will change with time. In high technology, targets can usually be broken into five categories: product, materials, tools, environment, and handling.
In addition to general targets, specific industries have unique targets to give direction for overall control. In the semiconductor industry, there is defect density or electrical fault density per unit area of silicon wafer (defects per square centimeter). General targets are tracked and compared to defect density results; then yield, quality, reliability and cost are used to track goals and set new targets.
The targets in Step Three must be evaluated with respect to the measurement equipment currently available in the marketplace. This will indicate if the targets are sufficiently aggressive. If equipment exists to readily measure all of the targets for a high technology product, this may be a clue that targets need to be tightened. On the other hand, defect control targets for circuit boards and modules may be considerably less stringent than those for semiconductor chips, and topographical feature may be much larger. Different technologies require customized inspection equipment.
The fifth step is to define timelines for process improvements. The required process improvements are derived from a direct comparison of the targets in Step Three with actual measurements from Step Four. Targets that are tighter than the “actuals” indicate immediate areas to be worked on. Areas where improved measurement techniques may be required can be determined if the equipment does not have the necessary sensitivity. The timeline is simply a graphic representation of targets and actuals over an extended period of time, usually one to two years. Timelines afford evaluation of long-range vs. immediate targets. They force long-range planning to keep up with product evolution.
Yield, Quality, Reliability etc.
The sixth step is to consider yield, quality, reliability, customer satisfaction, and market dynamics and track their trends using the targets in Step Three. In very complex processes involving hundreds of steps, it is difficult to get exact correlation, but trends do emerge. Efforts often have to be focused on general trends and catastrophic failures. Faster and more accurate measurements are essential, but progress in the development of inspection tools is often offset by increasing product density.
Yield measurements and failure analysis results are used to pinpoint and adjust critical targets. Total target updates are often implemented with the introduction of product advances, then measurements are again used to set priorities and correct targets.
Customer satisfaction and market dynamics should be examined regularly, because as technology and products rapidly evolve, it is easy to spend all your efforts on improving a particular product only to find that the market has disappeared! This is yet another reason to develop a total systems approach, so that you will be prepared for change–and positioned to react to it with flexibility.
The seventh step is feedback: the use of data and experience to fine-tune the process and meet the targets set out in Steps Two and Three. The system of relating yield and failures to Step Six targets is one example of feedback. This is one of the slower and more difficult feedback loops, largely due to lack of data and correlation. Using specific targets from Step Three and relating them to in-line specifications on individual process steps is easier to apply and track. Handling damage and foreign material on specific process steps can be tracked against tooling, process, and material changes and handling improvements. Specific changes such as the introduction of improved filtration can be directly related to foreign material on product or monitor wafers. Improvements in instrumentation and measurement accuracy can afford tighter process controls.
Roadmap of Future Requirements
The eighth step is the development of a future product or process roadmap. This is really an extension of the process improvements and timelines described in Step Five. The ability to meet timelines over longer time periods (two to five years) allows you to focus on long-range problems. The roadmap is extremely important and should be given special attention, for it guides long-range plans and provides a basis for strategies, development plans, and resource plans. Often, the line organization, working on current product improvements, is too busy to focus on long-range needs. The long-range roadmap contains current targets and short-range timelines, but it focuses on extending current targets to future program targets, drawing timelines to meet this roadmap. It insures that the longer timelines are supported by resources and checkpoints.
The future product roadmap is generated by using past product history, comparing it to current product targets, and predicting future needs. In the case of semiconductors, this offers an extremely aggressive learning curve that shows no sign of slowing down. Product density is plotted to predict future requirements; then targets are scaled to meet these targets. Roadmap planning is also the ideal vehicle to bring all of the various players in the project together as a team.
The ninth step is the implementation plan. It insures that as improvements are implemented to meet the roadmaps, these improvements are tested for manufacturability. Tool plans, capital plans, and resource plans all come together to achieve improvements on the manufacturing line. This insures that manufacturing can meet the needs of new, higher density products when they are introduced, and it makes use of advanced future improvements to enhance current products.
The tenth step is iterative cycling throughout all of the first nine steps whenever changes or improvements occur at any individual level. As improvements and innovations are introduced throughout the process, advances are measured and targets tightened. This can be driven by new measurement techniques that allow tighter control, or it can come about through a new or improved process, or the need to tighten targets for a new product generation. In all cases, it is enhanced by a systems approach that examines and understands where you are today and predicts whether changes will have a positive or a negative effect on product yield, quality, reliability, and cost.
A well-defined defect engineering or technical systems management approach provides the method and means to implement and maintain a flexible and competitive product. It also provides professionals working in these disciplines a better view of the “big picture.” n
Harold D. Fitch is president of Future Resource Development, a consulting firm in Burlington, VT, specializing in cleanroom education and problem-solving. He conducts international training seminars for CleanRooms` shows and seminars.
Figure 1. The first step in defect engineering is to complete a flowchart of your manufacturing process. The example shows the basic steps of a semiconductor oxide etch process.
Figure 2. Often, defect engineering will require completing more flowcharts as production complexity increases. This flowchart shows materials used in the actual etching process.