Publish Failure-Analysis Reports on the Web

We decided to set up a new failure-analysis reporting system that would

• get failure-analysis reports to key people in a timely fashion;
• make reports easy to access, and make older reports accessible;
• reduce the time needed to write reports; and
• reduce the costs of printing and storing failure-analysis reports.
  

  Based on these needs, we set the following goals for our system of Web-based software:

• Data from the lab’s sample-tracking system must automatically link to reports.
• A standard HTML layout will format reports.
• Clients will receive e-mail notice upon publication of reports.
• A manager must approve each final report before publication.
• Each report must be indexed and searchable.

We defined three basic sections for our reports: header information, a summary (or conclusion), and illustrations (if any). Defining the report this way let us separate the content from the format. We made the analyst responsible for creating the content, and made the computer responsible for formatting it. Because an analyst doesn’t need to spend time formatting the content, we reduced the time needed to prepare reports.

Our reporting system relies upon a sample-tracking database, in which we store information about each sample, such as device type, qualification stress (the accelerated tests that determine the lifetime of a part), failure modes, and customer information. The sample-tracking database makes use of the Web, so it was fairly easy to “connect” its information to our reporting system. We include hyperlinks in each report so clients can directly access the information in the sample-tracking database.

Our system, which we named the Deuce, comprises several software components. The tracking system and database use Action Request System (ARS) software from Remedy Corp. (Mt. View, CA). ARS communicates with a Web server using custom CGI scripts written in Perl, using the free ARSPERL library.

Creating a Report
An analyst starts a new report by clicking a button in our sample-tracking system. The tracking system creates a database entry for the report and links it to the sample. The computer then assembles a schematic view of the report. The analyst works in this view, adding report elements and information as needed.

The header section of the report provides basic information about a sample, as shown in Figure 1. (See www.tmworld.com/articles/09_ 1999_FA_report.pdf  for a report that includes illustrations.) The header information comes directly from the sample-tracking database. Rather than transcribing this information from the database, we set up our system to automatically import it to a report.

Figure 1. A typical report includes a header, the conclusions (or a summary), and illustrations (not shown). Our software provides a common format so analysts simply add information to a template.

After getting the header information, the analyst can choose another type of information item from a menu. Typically, the analyst works on the summary next, as it provides most of the key information about the tests run, the test results, and the conclusions about why a failure occurred. The failure analyst uses this section of a report to fully describe what took place in the lab.

The writer has no control over the layout of the information. He or she simply types information into the report for later automatic formatting. Our system always produces consistent, professional-looking reports, and the analyst doesn’t waste time playing around with the layout.

Check Interim Results
By relying on the Web to distribute information, we can easily communicate interim results to clients so they can see that we are making progress on analyses. The interim results also communicate information to other groups who are working in parallel on other aspects of the same problem. For example, our lab often works on design-debug problems with design teams outside the US. By using interim results, our lab and the design team can work together as effectively as two shifts.

During the course of an analysis, the collection of information grows to include many items, not all of which will appear in the final report. For example, although figures and captions—data items—usually appear last in a final report, an analyst begins adding them to a report long before the lab work and analyses are completed.

Also, when we relied on paper reports, an analyst would use Polaroid photos or an electronic image to show a client defects or failure modes. The photos took too long to mail, and the electronic images required the analyst and client to find a mutually acceptable electronic format as well as a transmission method available to both.

By using the Web to distribute reports, we have eliminated the need to coordinate image transmissions. Instead, images appear properly on a Web browser. And because the images are placed in a report form when an analyst acquires them, the images are available for anyone who needs preliminary analysis results. Producing the final report requires no reworking of the images.

Working with Images
As the analyst acquires curve traces, captures waveforms, and takes SEM (scanning electronic microscope) micrographs, he or she saves them on a local disk drive. Then, the analyst can load images into a report as data items by using a Web browser (Fig. 2). The browser display lets the analyst type caption text into a field associated with an image. The caption describes the findings and the image in more detail. At any time, analysts can use a Web browser to annotate the images they uploaded into a report. The screen in Figure 3 illustrates the annotations—text, arrows, lines, circles, and rectangles—that an analyst can add to highlight areas of interest in an image.

Figure 3. A simple image editor lets an analyst add text, arrows, and other items to an image. These items might call out specific points on a waveform, for example.

When uploading an image, an analyst must specify its source: an SEM, a TEM (transmission electron microscope), a curve tracer, or other instrument. The analyst doesn’t have to specify the format of the original graphics file because the image processing software we use automatically recognizes common graphics formats. The system can quickly convert images and optimize them using a free software package, Image Magick   ( www.wizards.dupont.com/cristy ). Image optimization and annotation automatically takes place through CGI (common gateway interface) scripts that we wrote in Perl, using the PerlMagick library that comes with Image Magick.

The CGI scripts automatically perform the following steps for each image as it is uploaded into a report (see “Optimize Your Images,” below):

1. If the image is wider than 500 pixels, the scripts reduce it proportionately in height and width to a maximum width of 500 pixels. This image size is more convenient for display on an 800-pixel wide monitor.
2. If the source is a photographic image, the scripts perform contrast enhancement.
3. The scripts apply a gamma correction of Xc = 1.6 to the image.
4. The scripts write photographic images to the Web server in JPEG format using medium-high compression. They put all drawing-style images on the server in GIF format.

Managers Publish Final Results
When an analyst finishes with a report, he or she clicks a link in the schematic view of the report. That click sends an e-mail message to a manager, asking for approval of the report. The manager can read and edit the report using the same schematic view the analyst used.

The manager can “publish’’ the report or return it to the analyst for additional work. Only managers have access to a special browser screen that lets them publish a final report. This system replaces the traditional sign-off on a printed report.

We index reports for searching using two methods. With standard Web tools, we index the HTML reports for word and phrase searches. We also use the query capabilities of the our system’s ARS software to search for reports based on criteria such as fab, responsible business unit, failure mechanism, and so on.

When a manager “publishes” a report, our system produces an HTML version of the report and places it on our Web site. Then, the system selects a list of report clients from a database of clients, based on the type of report and the clients’ interests, and sends each client an e-mail notice that the report is available on the Web. For example, if the report identifies a wafer-based failure mechanism for a device manufactured in Fab X, all clients interested in Fab X failures get selected. Other criteria, such as the assembly site and the responsible business organization, are also used to choose clients. Analysts add or delete names through a Web interface to keep the database up to date.

Each e-mail notice includes some of the header information, the summary, and the URL for the complete report. To reduce the flow of e-mail, the notification does not include the figures from the final report. Clients can click on a URL to read a specific report, or they can go to our lab’s internal Web site where they can search for reports based on publication date, business unit, and other characteristics.

To aid in the transmission of the report outside our company’s network, each report includes a link that performs a conversion of the HTML-formatted report into a PDF file. Clients can read the PDF file using free Adobe Acrobat reader software. T&MW

FOR FURTHER READING
You can find information about adding images to Web pages at www.photo.net/philg/how-to-scan-photos.html.
Information about gamma correction is available at www.cgsd.com/papers/gamma.html.

Dan Bodoh develops failure-analysis tools for Motorola’s NCSG Product Analysis Sciences group. He received a B.S. in computer science and engineering from the Milwaukee School of Engineering and an M.S.E.E. from the University of Wisconsin (Madison). E-mail: dan.bodoh@motorola.com.

This article is based on the paper, “Making the most of the Internet for Failure Analysis,” by D.J. Bodoh, Proceedings from the 24th International Symposium for Testing and Failure Analysis, pp. 332-338. Copyright 1998, ASM International, Materials Park, OH 44073.