Photomasks, although a significant part of IC manufacturing, have always been perceived as playing a small role in the industry because they are considered to be a subset of the design process and because the tapeout process involves relatively few people. Historically, photomask costs have represented a comparatively small portion of overall chipmaking expenditures. Since old paradigms are hard to break, it is therefore no surprise that robust systems have not been implemented to optimize photomask ordering. The lack of a modern electronic handshake has hampered the ability of chipmakers and photomask suppliers to shift from manual to automated ordering processes. Consequently, the speed and accuracy of order placement depends, in most cases, on the data-entry abilities of mask house and fab personnel.

What has been the impact of not investing in automation? The answer can be found in the "1999 SEMI/SPIE Mask Industry Quality Assessment" survey, in which photomask manufacturers cited job planning and collective administrative functions as the primary causes for returned material authorizations (RMAs). The category of job planning includes setup errors, programming, jobdeck preparation, and data-fracturing errors, while the category of administration includes sales orders, missing information, procedural errors, etc. Figure 1 presents data on mask returns from 1995 through 1999. Companies surveyed included DuPont Photomasks, Photronics, IBM, TMC, DNP, Infineon, TSMC, Compugraphics, Align-Rite, Toppan, and Innova.

Figure 1: Data highlighting reasons for mask returns, 1995–1999. (Chart courtesy of SEMI/SPIE, "1999 Mask Industry Quality Assessment.")

Standard turnaround times in the industry are falling rapidly. Twenty-four-hour turnaround times (or less) are becoming the norm as time-to-market becomes increasingly important to chipmakers. Given that trend, most photomask suppliers would probably agree that order-entry and job-planning steps are, or are rapidly becoming, major bottlenecks.

Throughput bottlenecks in semiconductor processes, as in many other multistep manufacturing processes, can arise periodically because of production anomalies that occur at random intervals. At Photronics, however, the job-planning bottleneck was not a passing issue, but a long-term problem that would not disappear. The seemingly permanent bottleneck was caused by rapidly increasing order specifications, design data complexity, and a growing workload shouldered by a finite number of employees. Problems arising from manual operations were compounded by the manual quality-assurance checks that were performed before orders were released to manufacture.

Chipmakers join mask suppliers in paying for the lack of automated processes. Often, photomask suppliers are contacted by customers and asked to verify specifications or instructed to halt photomask manufacturing entirely because of order errors. In such cases, chipmakers must pay a quick-turnaround premium to maintain shipment schedules. Defective photomasks have a negative effect on the industry as a whole, regardless of whether they reach customers or not. To confront this issue, the industry has attempted to adopt guidelines to address both cycle-time and quality issues associated with photomask ordering and manufacturing.

SEMI P10 is one of many guidelines formulated in the hopes of achieving standardization in the semiconductor industry. Although this specification has existed for years, there has been a recent drive to implement it. The goal of the specification is to facilitate the automatic generation and downloading of photomask orders. In theory, SEMI P10 enables any qualified photomask supplier to download a complete order generated by a customer. It allows the customer to supply billing, inspection, registration, and other types of general information about a photomask order, as well as jobdeck and measure file information. The system has been implemented primarily by larger chipmakers with substantial IT resources. Not coincidentally, such chipmakers face the most demanding turnaround times and see that SEMI P10 gives them a strategic advantage over their competitors.

SEMI P10 has been formulated in conjunction with industry representatives. A task force composed of chipmakers and photomask suppliers meets to discuss industry requirements and to anticipate technical requirements for the months following the meeting. Ideas are exchanged and recommendations made. SEMI members are afforded the opportunity to cast ballots to accept or decline task-force recommendations. Results are then released by SEMI. If approved, recommendations become part of the specification. Table I presents an example of a SEMI P10 specification.

The industry must begin to invest resources into streamlining the photomask-ordering process. It is safe to assume that unless industry representatives allocate time and effort to improve the situation, both quality and cycle times will continue to suffer.

Research shows where improvements are needed. In its current form, SEMI P10 is a good start. This specification functions well and benefits the mask maker and its customers. However, it must be supported by the semiconductor industry as a whole, for if each photomask supplier attempts to introduce its own brand of customer integration, cycle times and quality will be hindered rather than improved. For the most part, the SEMI P10 standard can be comprehended by both photomask suppliers and chipmakers alike.

Customer reaction to SEMI P10 has been extremely positive. As a result, Photronics has decided to use it as its preferred means of introducing orders into its global manufacturing network. In addition, it has developed CyberMask, an order-processing software package used to complement the specification's automation potential. The software was developed by the company's IT organization to automate the manufacture of photomasks and remedy quality and cycle-time issues. The system is versatile enough to handle not only the preferred SEMI P10 format, but other customer data media as well. A sample screen capture of the software is presented in Figure 2.

Figure 2: Sample screen capture of the order-processing software.

To develop the software, systems analysts, working in conjunction with job-planning personnel, documented the tasks required to manually prepare an order for manufacturing. Specifications for each task were created, and differences among customers and among the company's manufacturing locations worldwide were taken into consideration.

The system uses electronic information provided by mask users to automatically process and prepare orders for manufacturing. It reduces the need for manual intervention during job planning, avoiding costly delays and transcription errors associated with manual processing. Critical steps, such as the modification of design data provided by photomask users, are automatically and consistently performed so that higher mask quality and improved cycle times result.

The order-processing software application runs on a server that receives and processes orders for manufacturing. The software is segmented into distinctive functions that correspond to the manual job-preparation tasks documented during the analysis process. The system's core functions (highlighted in Figure 3) include: 

When evaluating production issues, organizations focus on root causes. If the semiconductor industry begins to invest in IT solutions, remedying cycle-time and quality issues will no longer be dependent on applying human Band-aids. For too long, the industry has relied on work instructions, corrective actions, and confusing or outdated revisions locked in three-ring binders. To make matters worse, the prospect of the $1 million photomask set is fast becoming reality. Expenditures at that level will force the industry to reevaluate how it manages the ordering and manufacturing of photomasks.

Industry standards such as SEMI P10 and the order-processing software discussed in this article offer system alternatives to an outdated means of transacting business. The software has led to quality advances and helped to reduce front-end data preparation cycle times by 85% or more. A comparison between manual and software-automated cycle times is shown in Figure 4. Hundreds of production orders involving the manufacture of more than 4000 individual mask layers have been processed using the software. Initial results have been extremely positive, yielding no customer RMAs or internal rejects. Customer satisfaction has been strengthened because the system can implement customer-requested changes systematically, ensuring quality output and timely deliveries.

Figure 4: A comparison between manual and software-automated cycle times.

Acknowledgments

The authors would like to thank Peter Jones, Mary Spano, Ed Mills, Norm Miller, Ryan Vo, Doug Lewis, Rob Gianazza, Steve Priest, Tony Baxter, Rich Meskill, Rick Gray, Justin Bradley and Mike Harrison for their contributions to this article and their work on the software discussed here.

Edward Suttile is the manager of the Corporate Optimization STandardization (COST) department at Photronics in Brookfield, CT, which is responsible for the development and deployment of the SEMI P10 and CyberMask projects. He has been with the company since 1985, working in the area of microlithography. Throughout his tenure he has held numerous manufacturing and analyst positions. In 1997, he was promoted to corporate manager of the COST department. (Suttile can be reached at 203/740-5323, or esuttile@brk.photronics.com.)

Charles Croke is a project manager at Photronics, where he is in charge of a customer integration initiative which includes the SEMI P10 download. He has held both manufacturing and sales positions in the photomask industry since 1993. He received an AS in liberal arts and sciences from Mattatuck Community College in Waterbury, CT, and is studying at Teikyo Post University in Waterbury. (Croke can be reached at 203/740-5214, or ccroke@brk.photronics.com.)

James Morrison is a project manager at Photronics, where he leads the CyberMask development and deployment initiative. He began working at the company in 1997 as a corporate production analyst. His responsibilities have included the Y2K-inspired roll-out and training of personnel for the MES to be used at all of the company's global manufacturing facilities. In 1992 he received a BBA with a focus on industrial management from Western Connecticut State University in Danbury. (Morrison can be reached at 203/740-5329, or jmorrison@brk.photronics.com.)