You are here : Stone Home > Info Center > Success Stories > Motorola


Updated:2010-06-08 17:54:37

Mark D. Wossum, Computer Integrated Manufacturing Manager

"Inspection with the aim of finding the bad ones and throwing them out is too late, ineffective, costly." So said W. Edwards Deming, perhaps the staunchest advocate of building quality into a production system.

Nowhere are his words more true than in semiconductor manufacturing, where microscopic variances and impurities can mean the difference between usable and useless product. Case in point: Motorola Inc.'s Metal Oxide Semiconductor Fabrication Plant 8 (MOS8, pronounced "moss eight"), in Austin, Texas.

At MOS8, our equipment operators get an unmistakable, graphic notification whenever a quality parameter- such as the thickness of the photoresistant coating applied to a silicon wafer- begins to veer out of control. Immediately, our QNX-based statistical process control system displays "SPC FAIL" in large, red letters on the operator's terminal. The operator and an engineer then work to put the process back on track.

What exactly is statistical process control (SPC)? In a nutshell, SPC lets you search for drifts in various parameters of any process, in order to catch changes that threaten the process and the integrity of the end-product. For example, if seven out of eight measurements on a parameter fall on one side of a historical mean, something is probably adrift in the manufacturing process. The following screen capture shows an example of the kinds of tests our SPC application performs.

A low-cost winner

Our SPC application runs 24 hours a day, seven days a week, and is part of Synapse, an integrated QNX-based manufacturing system within MOS8. Synapse SPC is completely mission- critical. If it goes down, our manufacturing goes down.

Synapse SPC is not only reliable, but sophisticated enough to have won an innovation award at MOS8. Yet it runs on standard PCs. The PC, in fact, helped launch Synapse SPC, since it let us create an inexpensive "proof-of- concept" to show to management. And better still, it lowered the cost of deploying the system.

A solid foundation

A mission-critical application like Synapse SPC needs a fault-tolerant foundation. To achieve fault tolerance, we rely on features of both the operating system, QNX, and the database engine, Birdstep Technology, a database designed for C and C++ programmers.

For example, because QNX is built as a network-distributed OS, our server processes are isolated. So, as a result, if a machine operator at MOS8 reboots or shuts down a workstation, he doesn't bring down the whole system.

As for RDM, it gave us tools- including source code and comprehensive function libraries for database storage and navigation- to build Synapse SPC to our exact specifications. Using this "open" database, we were able to incorporate features, such as a dual-database server, to fit the fault-tolerance requirements of semiconductor manufacturing.

Moving to Synapse SPC

Between 1990 and 1994, the value of world semiconductor production doubled. Likewise, production levels at MOS8 increased dramatically. As a result, we were quickly outgrowing our DOS-based standalone SPC application, which we ran at the site of each piece of equipment.

With the DOS system, there was no network, no centralized data repository. If a manager wanted to see data from a process machine, the manager had to walk over to that machine. Coordinating and sharing data from the 600 process machines we now use would have proved chaotic, if not impossible. We needed a networked solution that would give operators and engineers access to data from any computer, anywhere on the network.

We had another reason for a new solution: the growing recognition of quality as a competitive advantage. Customers were beginning to demand that quality be built into the manufacturing process, rather than be determined later by testing. One way to continually improve our manufacturing process was through an SPC system that would let us configure new measures of quality on an ongoing basis. (The following screen capture is taken from the system's configuration module, which lets operators create new SPC procedures.)

Delivering on time

Before creating Synapse SPC, we looked at one alternative- an HP-UX-based manufacturing system already in use at another fabrication plant. Unfortunately, the system lacked the configurability we wanted. And since it required workstation-level hardware, it would have cost at least 50% more than a PC solution. We were sure we could cut costs and provide more flexibility with the QNX-based system we envisioned, so we set about to prove it.

There was a problem, however. Because of rapid growth at MOS8, Synapse and SPC had to be developed in short order. And just as important, management needed quick proof that the concept would work.

Going with CSE

What to do? Well, because our end-users- machine operators and engineers- were nearby, we chose Concurrent Software Engineering (CSE) as our development method. In CSE, you do everything simultaneously: formulating requirements, designing, coding, testing, and even alpha testing.

Not only did CSE prove to be fast, but it avoided the "not what I wanted" and "what about this?" problems that often occur when using the traditional sequential or "waterfall" approach. CSE also helped avoid the trapdoor syndrome of shutting out further requirements because they weren't identified at the start of the project. (A caveat before I continue: While CSE worked well for us, it would have been much less practical if our end users had been at a remote site- immediate user input is critical to making CSE work.)

Rapid prototyping

To achieve the rapid prototyping that CSE requires, we used the QNX Windows Interface Editor (IEDIT), which saved us many programmer- hours with its ability to construct a window with objects.

Besides using IEDIT, we developed our own online tools for testing and validating the software. One of these, based on QNX's Ditto remote- diagnostics program, let us trace messages concerning system processes. From one terminal we could watch these programmatic messages throughout the system, and, as a result, it took much less time to find problems and debug. During the late stages of testing, these messages occurred too quickly to analyze immediately, so we simply saved them to files to inspect later.

The QNX environment helped us in still other ways. For example, having reliable networking built into the OS meant one less thing to integrate during the development phase.

How well did we do? Using two programmers, co-developer Phillip Harris and myself, we created a beta version- mostly from scratch- within seven months of first conceiving the system. Three months later, we extended Synapse SPC to the entire plant. That achievement says a lot about the utility of the tools that QNX and Birdstep Technology's database provided us with.

Taking control

When you mention semiconductor manufacturing, people are apt to conjure up images of high-tech workers dressed in space suits, laboring in a spotless, climate- controlled factory. In MOS8, those images come close to reality. But even so, they only begin to suggest the steps we take to control contaminant levels and other environmental factors that can affect quality.

For nearly every machine and process, we perform statistical tests. For example, when a machine called the coater lays down photoresist on the silicon wafer, we monitor the film's thickness. Then, when circuitry is photographed onto the chip, we use a focus machine to measure the picture's sharpness, in microns. Later, when a chemical bath removes the conductive or insulating material that isn't part of the circuitry, we use a test wafer to measure how many angstroms of material are removed per minute.

Synapse SPC takes in all this data through 140 networked computers that tie together about 50 types of manufacturing equipment- some 600 process machines in all. In most cases, data enters SPC automatically through drivers written with the Semiconductor Equipment Communications Standards (SECS) protocol, though some data is still entered by hand.

SPC system architecture

If you look at the accompanying diagram, you'll see that Synapse SPC includes several modules to collect, analyze, and report data.

Configuration - Lets the user specify a new SPC procedure. This is an important capability, since user configurability is crucial to our goal of continuous quality improvement. With this module, the user can construct a chart or graph to be displayed in connection with the new test procedure, and build a form for data entry. The user can then test the chart online with manually entered data. IEDIT plays an important role in this module. For example, it lets us incorporate useful graphics, such as a dial that shows when a piece of equipment needs preventive maintenance. And it also lets engineers store charts as binary large objects (BLOBs) in the database.

Data Entry - Handles all data entry. Also performs the SPC execution (statistical analysis of trends) and displays the results of the analysis.

Auto Data Transfer - Lets users share data with other QNX-based applications in the Synapse system, including our equipment-management application.

Reporting - Processes data from the database to provide summary reports and other information. The user can print the reports or display them on screen.

Extraction - Lets the user take raw data from the database, convert it to a text file, then export it to other systems.

RDM builds fault tolerance

Any SPC system is ultimately concerned with collecting, storing, and analyzing information. And for that it needs a database. Our problem was that an off-the-shelf database solution wouldn't provide all the features our application demanded, such as replication and a high degree of fault tolerance. And so we turned to Birdstep RDM Server, which gave us tools- including libraries for database control, definition, and manipulation- to build the database to our exact specifications.

A dual-server approach As the accompanying diagram indicates, we implemented the dual-server approach to build replication, fault tolerance, and online backup capability into the database. The servers consist of two PCs running duplicate copies of RDM.

The database has, in fact, three levels of fault tolerance. To achieve the first level, we group related database updates into a unit that succeeds, or fails, as a unit. This prevents records from getting out of synch- a primary cause of database corruption.

To achieve the second level of fault tolerance, we mirror all database changes from the first server to the second. If the first server fails, the second server takes its place in a matter of seconds.

To achieve the third level of fault tolerance, we maintain a physical backup of the database on the secondary server. In addition, whenever a message is echoed from the first to the second server, we store a complete record of the transaction in a message log that could be reconstructed if the secondary server fails.

The secondary server also allows "hot online backup," or backup while updates continue to be made to the database. This is achieved through a backup log on the primary server. While the secondary server is being backed up, the primary server echoes messages to this log instead of to the secondary server. When the secondary server comes back online, it "catches up" with the primary server by receiving messages from this backup log.

Eliminating the overhead

Besides helping with fault tolerance, RDM also helped maximize performance. It did this by allowing us to design the database using elements from the pointer-based network model, which predates the relational model. In the network model, records are grouped into directly linked "sets," an approach that can avoid much of the overhead and redundancy of the relational model. The net result: increased speed and efficiency.

A suite of QNX applications

I've described Synapse SPC, but in fact Synapse consists of several QNX-based apps. For example, it includes an equipment- management program that tracks the status of each machine and maintains the chemical recipes the machine can use.

There's also a recipe-download program, which automates equipment configuration. This program accesses information on the mix of chemicals to be used by a particular machine in different stages of manufacturing, and then loads this recipe directly into the machine.

Yet another important application is the Standard Operator Interface. This GUI program provides information on the state of specific manufacturing lots through a serial link to a VMS-based logistics program. It also lets operators run the recipe-download program.

The quest for quality

We're continually upgrading SPC, both to implement our own enhancements and to take advantage of the latest features of QNX. For example, we're now evaluating the Photon microGUI, which we may use to connect the Synapse system to heterogeneous networks. The offices at MOS8 use networked Macintosh computers, and with Photon, we could display our QNX Windows applications on these machines. Synapse SPC may already do its job well, but like the manufacturing process itself, it will benefit from constant improvement.


Premium Related Products
semi precious stone1
semi precious stone1
Price:USD56-99 / square meter
China Newstar Stone Co.,ltd
semi precious stones picture
semi precious stones picture
Price:USD56-99 / square meter
China Newstar Stone Co.,ltd
semi precious stones
semi precious stones
Price:USD56-99 / square meter
China Newstar Stone Co.,ltd
Browse by: Stone Manufacturers | All Products | For Buyers | For Sellers | Suppliers | International | China | Site Map
About Us | Stonebtb Service | Contact Us | Our Partners | Video | Exchange Link | Terms of Use
Copyright 2007-2019 All rights seserced