Mentor to Support GDSII Replacement

The manufacturer says . . .

Murray Disman says . . .

Mentor Graphics Aids in Rapid Adoption of GDSII Replacement Format WILSONVILLE, Ore.--Sept. 30, 2002--Mentor Graphics Corporation announced plans to support the (yet unnamed) GDSII replacement format in the Mentor Graphics Calibre product family and the IC Station tool suite scheduled for commercial release as soon as the first quarter of 2003. The new format is a replacement to GDSII, which has been used in the industry to transfer physical design data for decades. This next-generation, nonproprietary, interchange file format will be first used for leading-edge technology nodes such as 130 nm, 90 nm, and 65 nm. Mentor Graphics played many instrumental roles in the delivery of the new format. The most notable were agreeing to take on the editing responsibility for the specification, the initial specification itself (Version 3 SLF), and assigning Mentor employees, including one of Mentor's top software engineers, Laurence Grodd, as technical editor, and several other individuals as liaisons to the SEMI NSF WG (SEMI New Stream Format Working Group). Mentor agreed to hand off all proprietary rights to Version 3 SLF and any subsequent versions that include numerous contributions from Working Group members from other EDA companies, semiconductor companies, and the mask industry. Mentor contributed a GDSII-to-SLF translator to Working Group members on four platforms, and also participated in both the Working Group and the SEMI Data Path Task Force. The new encapsulated-format specification offers substantial improvements in data-volume efficiency. To date, testing shows up to a 10–50 times reduction in hierarchical data volume from GDSII, and up to two or more times reduction in flat data volume from the MEBES mode5 format. With the new format, file-based flows become more efficient, even in multi-tool environments, by reduced read/write times and the ability to attain greater leverage from hierarchical processing. Because of inefficiencies in the GDSII format, the best hierarchical processing applications were handicapped at output time; with the new format the differences in data-handling efficiency of hierarchical EDA application software will become even more obvious. In addition, the format offers portability across platforms and a single physical file to hand off across organizational boundaries, such as from IC design groups to merchant mask manufacturers or internal mask-makers. "More than 40 people, representing over 20 companies, participated in the working group to develop this new interchange format. Mentor Graphics was involved from the inception, serving for many months as the editor of the draft specification, and incorporating many successive sets of modifications for the working group. Mentor devoted some of their most talented people to this project, and participated actively in every working-group session. We look forward to seeing a broad range of industry support for the new format," said Kurt Wampler, chairman of the SEMI New Stream Format Working Group. "Calibre's leadership in resolution enhancement such as OPC, entry in the mask-data preparation market, and interaction with leading-edge IC customers triggered the internal development of a next-generation data format some time ago," said Joseph Sawicki, general manager of Mentor Graphics Design-to-Silicon division. "We contributed our work in this area, without limitation, because we believe meaningful standards adoptions must be truly open. We are convinced this new SEMI standard format will improve the industry's ability to profitably build more advanced semiconductor products."

Everyone is lining up to join the parade. The EDA leaders, Cadence, Mentor, and Synopsys, have all agreed to support the replacement for GDSII, which is being developed by Semiconductor Equipment and Materials International's (SEMI) IC Design/Photomask Data Path Task Force. Mentor has committed to supporting the new data format in its Calibre and IC Station tool families. GDSII has been used for transferring mask-design data for over 25 years. It is therefore no surprise that the data format cannot keep up with today's requirements. Some claim that a mask for a 0.13 µm design can require up to 50 Gbytes of data, and that OPC and phase-shift requirements can push this towards a terabyte. That is an awful lot of data to handle and to transfer from the design group to the mask-maker. SEMI's task force has just delivered the Open Artwork System Interchange Standard (Oasis) as the replacement for GDSII. While Oasis met most of the goals set up by the task force, primarily a ten times reduction in file size compared to GDSII, SEMI stated that there was much work to be done. Even though the ten times reduction in file size is impressive, the mask-design complications introduced by sub-wavelength lithography raise the bar to another level. The reduction in feature size to 0.18 µm progressed with very little difficulty. Since then, I've seen nothing but problems that don't get completely resolved. Many companies bypassed the 0.15 µm node in a race to reach 0.13 µm—the point at which copper interconnections became mandatory. While a number of fabs are in "production" at 0.13 µm, problems with the copper/low-k dielectric have kept yields low. It has proven to be harder than expected to develop and produce stable low-k insulators. The dielectric constants are nowhere as low as originally promised. Signal-integrity effects reared their ugly heads at 0.13 µm. Device leakage and a host of other physical effects are awaiting the designer at 90 nm and below. Things aren't a lot better on the design end. Verification still dominates the develop effort, despite a rash of new verification techniques. Reuse of IP cores, while helping, has not been the answer. We seem to be stumbling forward with patchwork solutions that are barely managing to keep the boat afloat. To SEMI's credit, the organization acknowledges that Oasis needs additional work. It is also taking a longer range view to the mask-data problem by investigating what it calls a universal data model (UDM) that could be used in both IC design and manufacturing. At present, the UDM is based on the OpenAccess data base and an API. The OpenAccess data model was developed by Cadence and was given to the OpenAccess Coalition in an effort to establish an industry standard. The OpenAccess Coalition is being managed by the Silicon Integration Initiative (Si2) standards organization.

Home    Product of the Week    App Notes    Tech Notes    Newsletters   

		Click here to get your listing up.