Issue

Several robotic material-handling products have been introduced for test, assembly, and packaging (TAP) in semiconductor manufacturing with limited acceptance and proliferation. So we question the assumption that back-end automation will follow the model of the wafer fab. This assumption is based on extrapolation of industry marketing data and does not consider operations management principles or relative cost structures of wafer fabs vs. TAP operations. The bottom line is that individual back-end factories will only invest in automation strategies that deliver the appropriate return on investment (ROI).

Thus, the path to back-end automation is not necessarily straight. Myriad choices are available to factory managers who need to base factory automation decisions on operational objectives of their factories. Most importantly, TAP factory automation can be implemented in increments, achieving an aggressive ROI at each step. This methodology is contrary to the widespread belief that a factory is not automated until it "looks like wafer fabs that use robotic material-handling systems."

Click here to enlarge image

Figure 1. Alternate paths to TAP factory automation present additional choices to factory managers. The circled numbers represent required infrastructure: 1 = host system; 2 = manual material identification; 3 = equipment connectivity; 4 = integrated material identification; 5 = MES integration; and 6 = robotics. (Note: Distinctions of different levels of complexity for each infrastructure component are not shown.)

Figure 1 illustrates the increased infrastructure required to take incremental automation steps. Obviously, each automation step is accompanied by investments. Note that the path to factory automation starts with a split that affects the ability to modify or alter future tactical automation decisions. The choice is to commit immediately to automated material handling (AMH) or to take the incremental approach beginning with automated data handling (ADH).

The AMH path requires large infrastructure changes and inherently risky large-scale investment and may make other paths unavailable. These risks involve the large capital commitment and the uncertainty surrounding how the introduction of a robotic system will affect current operational procedures. To deal with the uncertainty of system-wide changes, discrete event simulation is often used to verify the proposed benefits of AMH. Other productivity-enhancing tactics shown on the ADH path may not be available through the AMH infrastructure.

Click here to enlarge image

Figure 2. Alarm reporting (a five-day summary is shown), tabulated through ADH, helps prioritize process improvement efforts.

The ADH path provides many choices for implementation. There are many opportunities for improving factory performance through the elimination of paperwork and manual bookkeeping. These may involve data associated with the plant and process equipment, or data associated with material or components being processed. Corresponding to these path choices are investments in plant infrastructure and software tools. From a basic host computer that collects equipment or material data to a full factory manufacturing execution system (MES), these investments can be matched to the needs of data management at the area, cell, or factory level. Unlike AMH, the investment in ADH is incremental.

Along the path of ADH is another key split: one path focuses on data about the process and equipment, the other on data about material being processed. The equipment data-handling path can lead to advances in equipment utilization, process consistency, and ultimately process control and recipe management. The material data-handling path, on the other hand, provides basic tracking information, routing, and work in process (WIP) control. The paths of equipment and material data may eventually merge, providing several alternate objectives of device traceability, automatic recipe verification and download, factory planning, scheduling, and adaptive test process changes based on real-time material information.

Factory managers are driven to improve their operations via, for example, reduced cycle time, assembly cost and space utilization, and high equipment utilization. Several specific steps to factory automation can help satisfy these drives:

• equipment monitoring and performance reporting;
• process recipe verification and management;
• lot tracking and WIP monitoring and management;
• device traceability;
• adaptive final test protocols;
• factory-wide planning and scheduling; and
• automated material transfer.

Achieving each of these steps requires specific data sets from a variety of automation products integrated in a bottom-up solution. The investment made in these product sets will depend on what requirements are targeted in a particular TAP cell, line, module, or factory, and which path has been chosen for preceding steps.

Equipment monitoring, performance reporting

Equipment monitoring and performance reporting means improving process-equipment performance via equipment data. ADH makes equipment data quickly and easily available, allowing the identification of opportunities for improvement.

Managers can use equipment data in many ways to improve performance. For example, real-time equipment status permits immediate dispatch of technicians at the onset of unscheduled downtime. This reduces the duration of the downtime. In addition, it is possible to compile equipment alarm and assist data sorted by occurrences or total downtime (see Fig. 2). Process improvements can be aimed at persistent problems identified by this data. Moreover, as improvements are made, managers can use the same reports to measure effectiveness (see table, row 1).

Click here to enlarge image

To gather real-time equipment status and generate equipment alarms and statistics reports, an equipment interface is required to translate internal equipment data for use by a centralized host. While SECS-I, SECS-II, HSMS, and GEM standards define the environment and behavior models for TAP equipment, interfaces are still needed for equipment-to-host communication. Even if a piece of equipment complies with all Semi standards, that equipment is not necessarily ready to plug-and-play with an external host.

Developing equipment interfaces can be expensive and risky whether done by the equipment owner, vendor, or a third party. The writer must have a background in the TAP environment, knowledge of Semi communications standards, and an understanding of the host system and how the data will be used. Adding to the complexity is the maintenance required on an equipment interface as equipment is upgraded with new and enhanced features. Despite these difficulties, direct equipment communication is an absolute requirement for more elaborate factory control capabilities.

As stated above, the decision to integrate equipment to a central host should be based on specific benefits derived from direct equipment communications. In some cases, a factory's operational objectives can be met by monitoring and controlling only the material; equipment interfaces may not be necessary.

Process recipe verification, management

One of the fundamental challenges of TAP is maintaining an acceptable level of performance while assembling a wide variety of packages. Most TAP processing tools have a different process recipe for each product. The management of these recipes is a considerable task for any TAP factory. The objective for auto mating recipe management is to shorten equipment changeover time and to eliminate yield losses caused by running the wrong recipe (table, row 2).

Automated recipe functions differ based on equipment capabilities, host capabilities, and factory strategy. Some common recipe functions include electronic upload and download between host and equipment, and verification and comparison of a recipe resident on a piece of equipment to a recipe standard (also known as gold program) maintained on the host.

Off-line programming refers to writing recipes that can be downloaded by the host on an independent PC. This eliminates downtime while programming recipes on the equipment. More elegant systems can convert existing product drawings from CAD programs to recipes. This feature dramatically reduces the time required to write equipment recipes and improves accuracy.

Due to the lack of recipe standards in TAP, data required for such capabilities will be different for each equipment model on the floor and may, in fact, be proprietary to each vendor. Automated recipe management provides the most benefit for those processes that have the most complicated recipes. These typically occur in three operations: test, die attach, and wire bond.

Lot tracking, WIP monitoring

Tracking semifinished material within a factory means identifying and locating any part or component within a factory at a given time. Specific advantages include: process or material quality problems, in support of continuous process improvement, can be isolated, identified, and fixed faster; and production line conversions or process changes can be phased in faster.

Material identification facilitates even greater gains, including elimination of paper travelers and non-value-added labor needed to locate and inventory lots and to read and update lot travelers. The minimum investment necessary for the identification and tracking of semifinished material is a host computer and a means of collecting material identity and location from operator input by either terminal key entry or handheld code readers (table, row 3).

This approach provides data about the identity and location of all coded material to a point of inquiry. Factory personnel generate information about the identity, quantity, location, and time period of use. The task of hunting down lot travelers to make periodic inventories of WIP takes less time. In addition, this knowledge of material identity and location can begin the process of managing the amount of WIP at each station. Identifying and correcting process bottlenecks that arise suddenly from unscheduled downtime events downstream is much quicker than with a paper-traveler system.

Fully automatic material-tracking systems do not rely on operator data entry or hand wand readers. Instead, the material identification is done automatically as the material arrives and leaves each process station. With the advent of newer labeling technologies, the identity need not be limited to magazines or carriers. Unique identity codes on leadframe strips and even individual package substrates can be used to reduce tracking unit granularity to a single device.

WIP management

WIP management is distinguished from lot tracking and WIP monitoring by the ability to control WIP levels within the factory, thereby implementing material flow control. Two things are required to add WIP management: knowledge of basic machine status and a method to inform operators when and where to move WIP (table, row 4).

WIP management using human material handlers offers the same type of centralized material flow control that is offered by AMH. WIP management may be superior to robotics as an automation solution if labor reduction is not a primary operational objective for the factory.

Basic machine status required for WIP management is not the same as real-time equipment status described above. WIP management needs to determine whether or not each piece of equipment is allocated for production, and whether material is ready for placement or removal. This lower category of machine status may be supplied semi-automatically and managed through a central controller. In this way, direct equipment communication may not be required. Material movement instructions can be provided through monitors at a central location on the floor or through manual WIP-identification terminals already in use.

Click here to enlarge image

Figure 3. WIP management can provide short, predictable cycle times that can reduce quoted lead times. Each column shows average cycle time on the bottom and cycle time standard deviation on the top.

The primary benefit of material flow control through WIP management is cycle time reduction (Fig. 3). Reducing WIP, cutting batch sizes, and using WIP levels to control the material flow reduces cycle time. Equally important is the reduction in cycle time variability. The ability to operate with short, predictable cycle times allows shorter quotable lead times to customers and an advantage over competitors [1]. WIP management can also reduce or eliminate paper lot travelers on the factory floor because manual ID readers record the path of material as it moves through the process.

Device traceability

Traceability of finished goods in the field refers to after-the-fact identification of the facility, materials, and processes used to manufacture the product. Because traceability relates only to identification of product after it leaves TAP, it is generally a customer requirement (table, row 5). The objective is to provide a product history in the event of field failure, to reduce product liability cost, and to minimize recall expenses [2]. Reasons why traceability of material data is necessary include:

• the ability to record serial numbers of components or lots for warranty, field service, and recall use;
• the recording of what specific process tools, process recipes, and operators were on the production line when the product was manufactured; and
• the recording of data and lot codes for failure analyses of returned or recalled parts.

To implement lot or device traceability, connectivity to the various process systems is desired, but not required. Equipment connectivity allows automatic archival of data pertaining to each specific tool. This data may include process tool configuration and recipe identity; the identity of process consumables and special tooling; metrology or inspection results; and assembly cell or module, lot, unit, shift, and operator identity. As a minimum, these data may be captured manually via operator terminals or handheld readers. No matter how this data is captured, a continually maintained database is required to map product serial numbers to the production data recorded for each device.

As with material tracking, the device traceability process may be automated using unique identity codes on leadframe strips or individual package substrates. When identity codes are located on strips, industry standards are applicable (see SEMI standard T9). In addition, identity codes on strips, carriers, or magazines necessitate a mapping of device identity and location within these tracking units when individual device traceability is required.

Adaptive final test protocols

The ultimate gain from ADH is to incorporate material information into downstream TAP processes to take advantage of a priori knowledge of the material by adjusting or adapting processing parameters, thereby saving time or improving quality (table, row 6).

One obvious opportunity is in final test. Wafer probe data can be saved, updated, recalled, and eventually used to prescreen devices at final test or to allow more optimum test protocols. The identity of each device can be maintained through either a unique code at each die site or via a material-tracking and substrate-mapping system that uniquely identifies each assembled device from die attach through mold, trim-and-form, and mark.

Factory-wide planning, scheduling

After implementation of WIP management, higher-level factory improvements can accrue from the addition of production planning and real-time scheduling capabilities. In a production system with automated flow control, a scheduler can release lots into the system for processing and also select lots to be processed at each station.

Schedulers make decisions based on current factory status, including equipment status, WIP levels, order backlog, order due dates, and other information relevant to operations. This information is then put into mathematical models that consider the relative importance of different scheduling decisions. A scheduler outputs directives telling when to release and process each lot.

The scope of factory planning is larger than production scheduling. Factory planning software makes decisions regarding staffing, procurement, and subcontracting. Both planning and scheduling can be incorporated into enterprise resource planning (ERP) or MES systems that may already be in place. If the ADH path shown in Fig. 1 is taken, it is a matter of integrating the lower-level host to an upper-level MES to gain full top-down, bottom-up connectivity within the factory. With this investment, management can meet another key factory automation objective - linking the factory floor to the outside world (table, row 7).

Automated material handling

Without material management software tools to direct an AMH transfer system, robot-based or linked equipment can only perform a limited function. This functionality - moving material from one process tool to another - has significant benefits: the elimination of labor and errors caused by mishandling or mixing WIP.

In some assembly operations, these benefits are sufficiently valuable to justify investment not only in the transfer equipment, but also in modifications to factory infrastructure (table, row 8). These secondary investments may often exceed the cost of the handling systems because there are as yet no industry standards for physical and control interfaces, and many installed semiconductor assembly process tools are not yet "automation-ready."

Complete turnkey solutions from one supplier can surmount these obstacles. Investment in new equipment can ensure a smooth implementation. However, the integration of legacy equipment in existing factories is difficult and expensive (see "Top-down vs. bottom-up: Two approaches to assembly integration" on p. 80).

If implemented without material management software tools, AMH systems cannot significantly improve product cycle time or reduce WIP levels. Robotic transfer times are often not much faster than a person. Getting material to a process station faster will not reduce the backlog at that station. Thus, the benefits of automated material transfer are not as pervasive as once perceived.

Conclusion

We suggest that incremental TAP automation implementation provides a migration path offering aggressive ROI with less risk than all-inclusive automation. Detailed knowledge of each factory is required to select one automation strategy over another. Today, there are many options for TAP factories to improve operations through large- and small-scale automation implementation.

References

1. R. Burda, Proceedings of Semicon Taiwan Packaging Seminar, Nov. 5, 1998.
2. K. Boardman, Semicon West Proceedings, 1996.

Authors

Gerard H. Michaud is an engineering manager at Kulicke & Soffa Industries Inc., Factory Systems Group, 2101 Blair Mill Rd., Willow Grove, PA 19090; ph 215/784-6000, fax 215/784-6284, e-mail [email protected].

Richard Burda is a factory operations analyst for Kulicke & Soffa Factory Systems Group, e-mail [email protected].