Over the years, IPC-D-356 has served as a popular and useful data transfer format for conveying bare board netlist connectivity from front-end CAD and CAM systems to the electrical test function. Unique in our industry, 356 has enjoyed grass roots support among a majority of test equipment companies and CAM software suppliers when other IPC data formats have struggled to gain acceptance.

Part of IPC-D-356's success stems from it's somewhat narrowly defined mission to serve as a netlist format without the complexity of a true CAD or graphically correct CAM format. This scope has served the industry well for a number of years, but as board designs, test systems and repair requirements have advanced, the most recent release of 356, Revision A, is falling behind only 4 years after it's release. As the industry and supplier base has consolidated, the requirements of the standard have also become more global.

This article describes a bit of the history of IPC-D356, some of the areas where the standard is likely to change, and the progress of the effort underway to update the standard. The IPC-D-356 Committee has been re-formed with a diverse cross section of participants from test, CAM and CAD to sort out ambiguities and consider yet another round of enhancements. Although this effort is just getting started, the willingness of individuals from companies around the world to participate has set the stage for a positive result. The committee is seeking direct involvement of programmers responsible for generating 356A input and output translators, since they are most knowledgeable about it's shortcomings. It remains a distinct objective that the standard be as easy to support as possible from a software perspective.

Before launching into the types of changes to 356 that are being considered, it makes sense to try to establish a historical perspective. When 356 was first developed in the late '80s, electrical test was rarely netlist driven. Board data, in the form of plated hole and SMD pad locations were garnered from whatever means a company had at its disposal. Often "bomb sighting" was used, a process of digitizing the coordinates of outer layers for the purpose of generating test fixtures. After fabrication of a suitable fixture, the board connectivity was "learned" on the tester; a process that became known as the "golden board" approach. It was far from golden. Such practices led to frequent test escapes, and a pronounced lack of correlation between test and the functionality envisioned by the designer.

IPC-D-356 was conceived by individuals from several companies as a simple ASCII netlist format that would enable front-end PWB design software companies such as Mentor, Cadence and Pads to generate suitable data for netlist test. As a member of the 350 Series of IPC standards, the 356 format was encumbered by an 80 column, fixed format that is somewhat difficult to work with by today's standards. In keeping with it's intended link from CAD to test, the original 356 contained board design information such as actual signal names, reference designators (component names) and pin numbers.

What occurred over the next four or five years came as a real surprise to the authors of the original IPC-D-356. The CAD companies failed to support the standard; instead, the emerging CAM companies took it up as means of conveying netlist information after Gerber netlist extraction. (Gerber Netlist extraction is a process of determining the connectivity of a board based solely on the graphical elements that make up the photoplotting instructions, a wholly non-netlist oriented data source!) The CAD based Signal Name and Reference Designator fields were simply omitted from the CAM output in favor of a non-descriptive net number.

Companies such as Orbotech, Lavenir and Infinite Graphics began outputting 356 for the purposes of test, and importing it as well, when available from CAD sources, as a reference netlist for the purpose of pre- and post-CAM netlist comparison. Despite the lack of true CAD company support, the original IPC-D-356 became widely accepted and served the industry well. By the mid '90s, however, the standard was showing signs of obsolescence. The industry needed better graphical support for downstream repair, and better support for emerging board constructions and test methodologies. A cross section of the industry, predominantly from North America, gathered in 1995 to create a Revision A and set the following objectives:

• Maintain support for CAD and CAM source netlist, while making IPC356A readers backward compatible, i.e. able to read original IPC-D-356 files.
• Enhance the standard to include trace segment data.
• Add a means to describe the many varieties of "via in pad" construction and blind/buried vias.
• Add structures for net adjacency to support flying probe test.
• Add a description of tester channel assignment so that 356A could serve as a bare board repair format as well as a netlist input format.
• Provide a means for supporting embedded resistors, capacitors and inductors.
• Embellish the standard with examples in an effort to ease the pain of adoption.
• Remove ambiguity and issues of interpretation.

With these good intentions, IPC-D-356A was approved by the IPC membership in 1997.

Several years elapsed before the industry took up the new format with any measure of conviction. Although backward compatibility had been maintained, it came at the cost of increased complexity. The new format proved more difficult to produce and work with than anticipated, largely because CAM companies were trying to make netlist sense out of the graphical elements that make up a modern padstack.

The changes between the original 356 and 356A were substantial, and as is often the case, it was only upon implementation did we learn the specific nature of it's deficiencies. Despite these deficiencies, all the major CAM and test software suppliers have begun supporting Revision A, and like the original, it has become widely adopted. 356A serves as the primary data input format for a number of test companies, and even the CAD companies have begun to get into the action! It seems the industry needed to use it on a more widespread basis before the difficulties were really understood.

With this experience, a critical mass of test, CAM and CAD suppliers gathered at IPC Expo in San Diego this past spring to consider the future of 356. In a surprising unanimous poll of the participants, it was decided the committee should move forward with the development of a Revision B . This was particularly interesting in view of the fact that several proponents of GenCam and ODB++ were present and expressed interest in seeing 356 continue as a "test-specific" netlist format. It was generally felt that 356 should retain the simplicity and test focus that made it successful in the first place.

IPC-D-356B Methodology
During the coming months the committee will conduct a series of conference calls to determine the scope and focus of Revision B. Interested parties are encouraged to get involved! This far, we have active participation from four companies in Europe, 10 in North America and interest/comments from two in Asia. Although test, CAM and CAD suppliers are all represented, CAD is somewhat under-represented, again.

Proposed changes to the standard will be broken into sections and participants will create "straw man" versions for peer review. When the general form of the changes become clear, one or two focused meetings will be held to determine the text content. Hopefully, the committee can attract the participation of the software developers who actually write 356 input and output translators to these face to face gatherings.

Although it is still early in the process, the following has already been decided:

• The group felt that maintaining simplicity of the standard is important.
• 356B should not necessarily be graphically perfect, however, it should represent a substantial improvement over 356A.
• The objectives are to clean up ambiguities and add missing features. While there is notable interest in making changes to the fundamental record structure (and continuation records), no one suggested scrapping and starting again. Those companies that have already invested a substantial amount of time and effort will look at changes from a work/benefit standpoint.
• 356B readers NEED NOT be able to read 356A files since the version command is available to distinguish them. This revision is not necessarily backward compatible.

It is a little early to draw conclusions from the effort thus far. It would appear that with such a large group of participants, it will be a challenge to satisfy everyone with one solution.

Proposed 356B Changes
The proposed changes to 356 span the entire range from complex graphical support to removing ways that the existing standard can be interpreted. Below are some of the more important and popular proposals:

• More Test Point Shapes - 356B currently supports only round and rectangular pad shapes. Some means of expressing ovals, rounded rectangles and D shaped pads is desired. Complex apertures and polygon pads will also be considered. ·
• Test Point Attributes - some additional information about test points would be useful for the test department. For example gold plated test points that require a special contact method could be identified, although it is not clear that such information will be readily available from upstream sources.
• Tester Automation Support - automatic alignment and some tester automation require particular types of data to be conveyed, beyond simple fiducials that are supported already. This is critical as more automated test and sequential test is performed. Requirements for non-board features such as dog-bones to test for complete scoring or routing are also desired. ·
• End Point Method - Several forms of end point optimization exist and it is much easier to read a 356A file if the method were stated in the file.
• Polygons, Contours and Scratch/Paint - A significant problem exists with the increasing use of polygons, contours and special apertures as a way to support sketched pads and complex shielding patterns. Some means for supporting this is needed, if even just the outline and polarity of the contour for improved graphical representation during repair. This would also prevent CAD/CAM suppliers from needing to "explode" polygons into huge numbers of drawn positive segments with the corresponding explosion in file size.
• File Organization - It is not uncommon for 356A suppliers to have ignored or misinterpreted file organization requirements. These sections need to be clarified and better examples provided.
• Complex Record Structure - Complex records (via in pad constructions) are of particular interest because they must be tested using different approaches depending on the size of pad on the surface, and the size and location of holes within (or touching) the pad. While most participants appreciate the logic in the current scheme, the group will look for ways to better express complex records and to make their processing more straightforward from a coding perspective.
• Soldermask Support - Although two methods for describing soldermask coverage exist in 356A, there is some dissatisfaction at how these methods are used, particularly with regard to complex records.
• Multiple Test Points on a Record - Some discussion is needed about how the standard supports multi-pass testing of an individual board. Examples are needed.
• Inner Layer Pad Data - Some companies feel inner layer pad data is becoming a necessary part of the electrical test data. It is currently permitted, should it be embellished and encouraged?
• Multiple Blind Vias in a Pad - This construction is becoming more common and is not well supported.

On balance, the proposed changes are constructive and will improve the industry's ability to support the kinds of board constructions and test methodologies we face today. In itself, the gathering of such a wide group of industry perspectives has already begun to lead to a better understanding of how to interpret 356A. It remains to be seen if the objective of "simplicity" can be met with all the new requirements that the industry has heaped upon the standard.