Signal Integrity

Power-aware Signal Integrity and EMI/EMC On High-speed Digital Chip-to-Chip Links

Signal Integrity header image 2

“O” is for … ?

Posted March 3rd, 2012 · 9 Comments · Editorial

Share

Posted by Colin Warwick

Occupy? Offspring? Occlude? According to an article Which Data Transfer Format is Best for the Industry? the “O” in “ODB++” stands for Open as in Open DataBase. So, kudos to the Mentor Graphics’ Open Door program for opening the new, improved ODB++ Solutions Alliance. But I’m wondering if this new attempt to add a little openness to ODB++ is in response to the relaunch last year of the competing computer aided design to computer aided manufacturing (“CAD-to-CAM”) data format, namely IPC-2581 aka “Offspring” backed by IPC and by Mentor’s two main competitors in the enterprise printed circuit board (PCB) tool market: Zuken and Cadence… Long story…

Nyitott ajtó (Open door) by Albert Neuhuijs 1844-1914, oil on canvas, 1905, Teylers Museum, Haarlem, NetherlandsNyitott ajtó (Open door) by Albert Neuhuijs 1844-1914, oil on canvas, 1905, Teylers Museum, Haarlem, Netherlands.

How did we get here?

As everyone knows, inside every electronic gadget is a PCB that connects the semiconductor and other components together. Almost all printed circuits boards are designed virtually in a CAD system. To physically realize the design, the computerized design information must be transferred to a photolithographic CAM system. Since the CAD and CAM systems are generally produced by different companies, they have to agree on a CAD-to-CAM data exchange format to transfer the data. Over the years, there have been several formats. First there was Gerber, named after Heinz Joseph Gerber’s company, Gerber Systems. On August 27, 1980 the first edition of the Gerber Format: a subset of EIA RS-274-D; plot data format reference book was published as a specification to drive their CAM photoplotters. As the title of the manual implies, it was a proprietary subset of EIA RS-274-D, an industry standard for numerically controlled machines. These were the days of 7-bit punched tape data storage, which drove it to be a very terse standard. It took a skilled operator to infer the design intent from those few ASCII rows.

Since Gerber defined exactly what’s in and what’s out, Gerber was a proprietary standard, even though it was derived from an industry standard. The advantage of a proprietary standard is that things can move quickly: no red tape and fractious committee meetings. And to the proprietor there’s the big advantage of being able to call the shots. If partners and users of the standard have some way to benefit, an nice little ecosystem builds up around the proprietor, like App developers around Apple’s App Store, or Windows applications around Microsoft Windows API. In contrast, industry standards move more slowly, but they can offer a level playing field.

New kid on the block

In 1995 a CAM company, Valor Computerized Systems, Ltd., came up with a much more complete CAD-to-CAM data exchange format first called ODB. The ++ part — evocative of C++ — was added for the 1997 revision. ODB++ covers the specification of not only conductor layer artwork and drill data, but also material stack up, netlist with test points, component bill of materials, component placement, fabrication data, and dimension data. How would the Gerber camp respond?

By this time Gerber had sold the CAM part of its business to Barco and with it the Gerber standard. On September 21, 1998 the new owner published “extended Gerber” in RS-274X Format User’s Guide. Each Gerber file still specified a single conductor or mask layer image but ambiguities inherent in the terse older version were cleaned up. This was nicer than standard Gerber, but still nowhere near as complete as the new kid on the block, ODB++. Workarounds were found: the drill data could be specified in Excellon. The netlist could be specified using IPC-D-356. Layer names, material stack up could be provided piecemeal in informal text files or drawings. So now we had two workable, proprietary standards with varying degrees of completeness. There was a need for a industry standard, one that was complete, neutral, open, unencrypted, and public. Maybe a third standard, IPC-2511 (“GenCAM”), which resulted from a donation of technology by Teradyne/GenRAD to IPC, could fill the need? In 2002, after a two year effort by the National Electronics Manufacturing Initiative to build a consensus, a compromise format based not on GenCAM but on ODB++ and XML, ODB++(X), emerged. But without complete supply chain support, it was mostly ignored. Companies carried on with extended Gerber (owned by Ucamco by this time) plus the workarounds and Valor’s version of ODB++!

IPC launched the IPC-2581 “Offspring” effort to merge ODB++(X) and GenCAM around 2004 but progess was slow at first. Then the need for an industry standard grew more urgent and intense when Mentor Graphics acquired Valor (and ODB++) in 2010. Mentor’s main rivals in PCB CAD, Cadence and Zuken, joined the effort. Mentor declined saying “our strategy is based around ODB++.” To avoid the slow start that GenCAM and ODB++(X) suffered, a supply chain consortium gathered together to promote the standard.

So here we are

Gerber is still the most widely used format. Ucamco owns it and revises it from time to time. The latest version, revision H, was released as a freely downloadable PDF in January 2012. The ODB++ spec is available but only if you meet Mentor’s Open Door criteria which include:

  • Provide a commercially available software product that complements a Mentor Graphics product offering to provide a more complete design solution to our joint customers.
  • Demonstrate a customer need for this integration through references from mutual customers.
  • Provide a recommendation from a Mentor Graphics product division or demonstrate the incremental value of this integration to both Mentor Graphics and the partner company.

Gulp. Not exactly vendor neutral. So my point (and I do have one) is that if Mentor chooses not to place ODB++ into the public domain (which is their right), the industry needs to get behind IPC-2581 Offspring in order to move forward. What do you think? Please leave a comment below.

Tags: ···

9 Comments so far ↓

  • Julian Coates

    Thanks for the acknowledgement of the release of the latest incarnation of the ODB++ partnering program! Indeed, the “O” does stand for Open, and if you look at the list of partners on the site you will note that all the main movers and shakers behind the IPC 2581 consortium are also ODB++ partners. Why? – customer demand. See the latest press release on http://www.zuken.co.jp/ for illustration. The acid test, I suggest, for the openness is whether any applying-partner gets rejected; I can report that no partner has been rejected from the ODB++ Solutions Alliance – all have been welcomed and the format definition made available without hesitation. I invite you to join the Alliance as an individual, where you will see many resources freely available. If you then decide to join in order to make a commercial ODB++ interface you can apply via the Open Door channel. Will be glad to work with you.

  • Colin Warwick

    Thanks, Julian.

    But if no-one is ever rejected, surely the spec could be openly published like Ucamco publishes Gerber? Isn’t that the test of a truly open spec? This would open up the spec to people who can’t apply at all now because they can’t swallow Open Door’s T&C’s.

    Cheers,

    – Colin

  • Colin Warwick

    Recent development: Mentor announced they would support IPC-2581, creating the possibility of a choice for all designers no matter which EDA tool they use. May the best format win!

    http://pcdandf.com/cms/designnews/8879-mentor-to-support-ipc-2581

  • Gary Carter

    IPC-2581 Revision A was published last year. We successfully exported our design from Cadence SPB 16.6 in IPC-2581 format, imported the IPC-2581 data into WISE VisualCAM and constructed our pallet, then exported the IPC-2581 and fabricated a bare board with a supplier that uses WISE software.

    At Apex 2013 Frontline PCB Solutions announced IPC-2581 support in Genesis 2000 Version 10. We are excited to be working with Frontline and Sanmina, our to prove the flow through Genesis is production ready.

    We are developing internal capability to consume IPC-2581 for assembly in our factory (primarily FUJI SMT) and expect to have this completed shortly. Several enhancements were added to IPC-2581 to support this capability.

    IPC-2581 Revision B is due to complete the review and approval process in June 2013. This release adds new capability for defining the PCB material stack-up including the ability to exchange stack-up definition between design, simulation, and fabrication tools. It also provides enhanced support for “Specifications” that further reduces reliance on paper documentation in the product realization process. See http://www.ipc2581.com for details!

  • Colin Warwick

    Thanks, Gary. That’s exciting news.

  • Aubrey Sparkman

    While I would definitely prefer an open IPC standard format over a vendor controlled format, I have to ask a question important to EM simulation tools. Does either format handle stackup and materials information better than the other? IE does either (or both) include dielectric loss tangent? Does either (or both) allow a Dk and Df for the metal layers?

  • Colin Warwick

    Hi Aubrey, Good points. We run into this issue a lot. Neither format was primarily intended for EM modeling in a field solver. Both are aimed at manufacturing where the Dk/Df info isn’t needed. I believe both bodies were considering adding this info in order to support EM tools. I’m not sure how far the request got.
    – Colin

Leave a Comment