A lively panel discussion Sept. 29 revealed that PCB designers have some strong opinions about the data formats that convey design intent to manufacturing. Several audience members expressed support for the Gerber data format that has been around for over 30 years. But other audience members and panelists agreed that a more intelligent and up-to-date format is needed, and that an open industry standard called IPC-2581 appears to be the best way forward.
The panel, titled "Data Transfer in the 21st Century," was held at the PCB West conference in Santa Clara, California. The moderator was Mike Buetow, editorial director of the UP Media Group. Panelists were as follows, shown left to right in the photo below:
The panel's purpose was to introduce the new IPC-2581 Consortium, a group of OEMs, EDA companies, and electronics manufacturers whose collective aim is to accelerate the adoption of IPC-2581. Cadence played a leadership role in establishing this consortium and has been joined in it by Zuken, another major PCB CAD provider. Other members are shown in the graphic below.
The following background will help set some context for the panel:
Perhaps the most important point to remember is that while the Gerber format can transfer image data, it does not transfer stackup data, materials, design intent, or a netlist. IPC-2581 provides a complete transfer of design data and promises to save millions of dollars that are wasted today by managing multiple files in different formats.
Time to Move Up from Gerber?
Shah opened the panel discussion by noting that while Gerber is a proven format, "it really does not have the intelligence customers need today to communicate design data to manufacturing." One thing it lacks is netlist information - manufacturers have to do a lot of reverse engineering to get that, he noted. An audience member noted that an IPC standard netlist file can be sent with Gerber data. Shah responded that "the beauty with IPC-2581 is that it's all in one file."
A few audience members expressed concern that IPC-2581 or ODB++ might send out information design teams don't want to convey, but Shah noted that designers can control what data goes out, and Clark gave a similar assurance for ODB++. When an audience member said he saw no reason to use ODB++, given that fabricators still want Gerber files, Clark responded that "there's a tremendous risk when you provide low-level machine driven formats. A lot of reverse engineering is done on those files, more than you'd probably like."
Later in the discussion, an audience member said that there will still be a need to support Gerber as a "least common denominator" format, and declared that "when we output Gerber every fab house in the world can read that data." Clark's response: "Yeah, but using that logic we should go back to Morse Code."
"The push for IPC-2581 is not that Gerber is a bad thing," Almeida said. "Using IPC-2581 or ODB++ lets us consolidate on a common format to bring data back and forth, as opposed to multiple file formats from multiple EDA vendors." Korf said that Sanmina-SCI manufactures several thousand designs per week, and that there's a lot of extra work involved in comparing and checking all the diverse files that come in with Gerber data. With ODB++, a lot of that checking goes away.
Offering a user perspective, Carter noted that "the vision I have for IPC-2581 is beyond Gerber replacement, it's intelligence. I want to control my IP. I don't want information that's not necessary going to an overseas fabricator somewhere." He also cited the need to work with his "counterpart" on the manufacturing side so that both have the same information. "I've got it in Cadence [Allegro] and he's got it in his tools," Carter said.
IPC-2581 or ODB++?
Given that both IPC-2581 and ODB++ are presumably "intelligent" data exchange mechanisms, why IPC-2581? Clark said that ODB++ has thousands of users, and is available to other PCB CAD vendors (including Cadence and Zuken) through a partnership program. But it's still a proprietary format owned by Mentor, and several audience members challenged that status.
"As an Allegro user, I have to go to the Mentor web site to get an ODB++ translator," said one. "That always bothered me. The people writing translators are not the best suited to do that." Said another: "You have to sign a non-disclosure agreement with Mentor. That's not fair."
Clark responded that "our concern is that if it just proliferated, there would be variances and it would become a non-standard. Our concern is that if we make it publicly available, the format could run amok." Shah replied that single-entity control is a good idea, but it should be through a standards body such as the IPC.
"You have to license it [ODB++], it's not open," Almeida said. "That's why we're looking at 2581. We need to rely on something that's neutral, where nobody has a vested interest in changing the format."
One audience member had a practical concern. ODB++ is the database of a commercial tool, he said, and when the tool changes, the format may change. He cited an example in which a vendor in China had an older version of ODB++ and couldn't read the current one. "I'm thinking 2581 will be more stable," he said. "I'm hoping 2581 catches on."
Will Mentor support IPC-2581? "Right now our strategy is all focused around ODB++," Clark responded.
Why a Consortium?
Given that IPC has already ratified IPC-2581, why is a consortium needed now? Dieter Bergman, technical director of the IPC, spoke from the audience about the formation of IPC-2581. He noted that it "doesn't work as a standard unless there's an industry group that wants to do it. It's an opportunity to really take a giant step. Thank God some of the CAD companies were ready to do this and found support for it. We will support it to the nth degree and if something is not right we'll fix it."
Shah said that Cadence helped launch the IPC-2581 Consortium based on customer demand. "There's a huge demand from customers to have something that's open and neutral," he said. Thanks to the consortium, he noted, "we can demo a design data exchange between Cadence and Zuken."
"We think [IPC-2581] is going to be the format of choice going forward," said an audience member. "Not only does it include everything needed for the industry, but it's very dynamic and easy to upgrade the format to what we need."
First let me express great admiration for the venerable Gerber format, which has been around longer than almost anything in EDA except maybe SPICE simulation. But PCB designs are much more complex than they used to be, there's far more data going into manufacturing, and it's time to move on to a format for the 21st century. A number of companies in the PCB design and manufacturing flow, including Cadence and several large OEMs, are betting on IPC-2581 and are working hard to move it forward.
More information about the consortium and about IPC-2581 is available here.
I'm really glad to see IPC-2581 move forward. I hope it gets adopted universally and quickly.