Major Major and Standard Standard

We ask for your bill of materials, Gerber and centroid files to assemble your PCBs. All those pieces of information are necessary to properly program our machines to place your parts. That’s pretty standard stuff, but did you know that when the Gerber format reference book was first published, Jimmy Carter was President of the United States, Russia was the “Soviet Union” and Voyager 1 was well inside the Solar System?

Use of the format has been going on even longer. Yeah. It’s been around a while. For some reason, it has been very difficult to get everyone to agree to and use a standard file format. Gerbers really don’t have enough information in them to do the job properly, but it is the standard. Hopefully not for too much longer. How many of you reading this were even born when Gerber was new?

There are a number of formats around that are better than gerber and Screaming Circuits will accept many of them. First, your CAD software probably will export an “ASCII CAD file”. This is a good format. Some export ODB++, which is one of the newer formats, again a good choice. One of the newest standards is the IPC-2581. It’s been around a few years and is now getting a lot of attention. If you happen to use Eagle CAD, you can also send us the Eagle “.brd” file.

IPC-2581 includes the best of ODB++ and GenCAM. It has all of the fab data, assembly data, netlist and BOM. Everything needed in one convenient file. My understanding of the format is that you can exclude portions of the data set that you consider proprietary. You can learn more about the format here. There’s more background information on the subject at PCD&F magazine.

Duane Benson
Where’s Henry?
I need an inductor.

http://blog.screamingcircuits.com/

Data Transfer in the News

A couple new articles are out on the IPC-2581 and ODB++ data transfer formats.

On Oct. 2, longtime EDA journalist Richard Goering provided a well-written writeup on the “lively panel discussion” (“Data Transfer in the 21st Century”) we held during PCB West on Sept. 29. Richard does a nice job capturing the frustration of the designers present and historical give-and-take that has led us to the current situation.

And yesterday, EDN weighed in with interviews of participants from the data transfer panel held at PCB West and other key spokespersons.

Given the new support for IPC-2581 by Cadence and Zuken, among others, this issue isn’t going away.

Talk Isn’t Cheap

Communicating is hard. It took thousands of years just for man to develop a common language. I don’t suppose, then, even in our “enlightened” state, we should expect it to be easy to develop a common, complete method for describing all the myriad features of a printed circuit board.
This week at PCB West, the Silicon Valley annual trade show, a special panel will convene to address just that decades-old issue. (Disclosure: I’m the moderator.) I don’t expect the group to solve all the industry’s data problems in just 90 minutes, but I do think a few key aspects will be noted.

Here’s a question I plan to raise: Would the problem of unintelligent data files be essentially resolved if the initial cost to upgrade were lower?

Upstream, Intel, for example, sends an army of engineers to its suppliers to help them implement new processes. Few companies have the resources of Intel, of course. No fabricator does. And this leaves the fabs in a bind: They know that Gerber is insufficient, and spend countless hours massaging (often without their customer’s knowledge) the bad or incomplete data received from design. But with tooling jobs stacking up on their desks, and margins cut to the bone, they claim no resources to spend on implementing one of the richer data transfer formats like ODB++ or IPC-2581.

So who pays?

Neither IPC nor Valor make any money directly from their respective data transfer formats, so it’s unlikely either would see the value in extending themselves further by underwriting the onsite development and implementation work. (Whether they should anyway is a column for another day.) Designers tend to be risk-averse: They are unlikely to risk their jobs on something upper management is not mandating. Thus, it may be that the fabricators need to start assigning a CAM engineer to its key customers — perhaps one at a time, to keep costs down — to help them get up and running — no matter which rich format they choose.

The argument for switching to a superior format(s) is that manufacturers will save money down the road. I understand, however, that quantifying the cost savings is exceedingly difficult. Moreover, as one CAD developer told me, there’s an unwritten incentive for the status quo (read: Gerber) because manufacturers don’t want to appear inflexible.

I would argue that the industry’s margins can’t afford to keep sending bad data downstream and hoping for a miracle in return. Fabricators over the past decade have lost most of their influence over the printed circuit board development. This is an area where they can truly coach their customers — and add value in the process. They should grab it.

Talk Isn’t Cheap

Communicating is hard. It took thousands of years just for man to develop a common language. I don’t suppose, then, even in our “enlightened” state, we should expect it to be easy to develop a common, complete method for describing all the myriad features of a printed circuit board.

This week at PCB West, the Silicon Valley annual trade show, a special panel will convene to address just that decades-old issue. (Disclosure: I’m the moderator.) I don’t expect the group to solve all the industry’s data problems in just 90 minutes, but I do think a few key aspects will be noted.

Here’s a question I plan to raise: Would the problem of unintelligent data files be essentially resolved if the initial cost to upgrade were lower?

Upstream, Intel, for example, sends an army of engineers to its suppliers to help them implement new processes. Few companies have the resources of Intel, of course. No fabricator does. And this leaves the fabs in a bind: They know that Gerber is insufficient, and spend countless hours massaging (often without their customer’s knowledge) the bad or incomplete data received from design. But with tooling jobs stacking up on their desks, and margins cut to the bone, they claim no resources to spend on implementing one of the richer data transfer formats like ODB++ or IPC-2581.

So who pays?

Neither IPC nor Valor make any money directly from their respective data transfer formats, so it’s unlikely either would see the value in extending themselves further by underwriting the onsite development and implementation work. (Whether they should anyway is a column for another day.) Designers tend to be risk-averse: They are unlikely to risk their jobs on something upper management is not mandating. Thus, it may be that the fabricators need to start assigning a CAM engineer to its key customers — perhaps one at a time, to keep costs down — to help them get up and running — no matter which rich format they choose.

The argument for switching to a superior format(s) is that manufacturers will save money down the road. I understand, however, that quantifying the cost savings is exceedingly difficult. Moreover, as one CAD developer told me, there’s an unwritten incentive for the status quo (read: Gerber) because manufacturers don’t want to appear inflexible.

I would argue that the industry’s margins can’t afford to keep sending bad data downstream and hoping for a miracle in return. Fabricators over the past decade have lost most of their influence over the printed circuit board development. This is an area where they can truly coach their customers — and add value in the process. They should grab it.

Intelligent Design

In my monthly column for PCD&F last month, I was ostensibly discussing standards and how they come to be. The first standard I worked on was IPC-D-350, one of the first of the would-be slayers of Gerber, the so-called unintelligent data format. Indeed, I’ve spent a good part of my life watching electronic data transfer formats come and go, and at the end of the day, Gerber, warts and all, has remained the one to beat. So I’m not prepared to rise up and shout to the heavens that IPC-2581, the latest iteration in 40 years’ worth of attempts at an “industry” standard, is at long last the answer.

But as we noted in “Around the World
,” there are enough notable differences in the process this time around to make it newsworthy. First and foremost, there are real live CAD tool vendors not just showing up at the meetings, but actively participating (!).

To understand why this is significant, we must go back to my IPC-D-350 days. Digital Equipment and the late, great Harry Parkinson were instrumental in trying to revive interest, and we at IPC also had support from several smaller software folks like Dino Ditta at Router Solutions and Steve Klare at Intercept Technology. But we never managed to break through, and a big part of the problem was the major CAD vendors’ collective refusal to offer IPC-D-350 as an output (or input). The response always was, “We’ll do it if our customers ask us.” But what they were really saying was, “We don’t want to make it easy for our customers to migrate their designs to a competitor’s tools.”

In the meantime, AT&T offered up RS-274X (aka extended Gerber), which UCamco continues to support, and Valor developed ODB++, and (like Gerber) while it was originally conceived as much a machine language as a format for electronic design data, it was accepted by fabricators desperate for something, anything, more intelligent than Gerber.

Under the leadership of Dieter Bergman, IPC also continued the fight, enlisting the help of the National Institute of Standard and Technology (NIST) through not one but two (GenCAM, Offspring) successors to IPC-D-350. (For a short history of the standards, click here.) Yet even now, after decades of trying, no group has been able to dismount Gerber from its perch, and it’s long past time we did. Data transfer formats are not something anyone ever will make money from, but every day we go without a better one, everyone will lose some.

Curiously, just a few weeks ago, I was contacted by David Gerber, son of H. Joseph Gerber, who invented the photoplotter and the eponymously named de facto standard that ran it. Gerber’s genius cut across many industries, from electronics to apparel, and he was awarded the 1994 National Medal of Technology for his life’s work.

For such an esteemed inventor, Gerber’s backstory is even more interesting than his career. As a teenager in 1940, he fled Nazi Germany for America. As an aeronautical engineering student at Rensselaer Polytechnic Institute, he discovered a way to reduce the time-consuming nature of graphing calculus problems using (seriously) an “expandable ruler” created from the elastic waistband of his pajamas. And of course, he formed The Gerber Scientific Instrument Co. in 1948, which is still going strong today.

The younger Gerber is writing a book about his father’s exploits. I look forward to learning more about the life of one of our industry’s true unsung heroes. But at the same time, I’m going to do everything I can to help retire one of his legacies.

In our cover story this month, Hemant Shah and Keith Felton of Cadence explain a new consortium taking root. The consortium is backed by a Who’s Who of OEMs and EDA vendors, including Harris, Ericsson, Fujitsu, nVidia, Sanmina-SCI, Cadence, Zuken, Adiva and Downstream Technologies. Its goal is to accelerate the adoption of IPC-2581 as an open, neutrally maintained global standard to encourage innovation, improve efficiency and reduce costs. The members are committed to adopting IPC-2581, which as I noted gives this latest effort a big leg up on all previous attempts.

Where does UP Media Group stand on this? For 20 years, we have supported the development of an intelligent, robust format for electronics data transfer. As such, we fully support the consortium’s effort to ensure a viable, supported and independent data transfer format that is driven by user needs.

That new task group attempting to update IPC-2581 recognizes that design needs will at some point “break” Gerber. Many of the players are new to the game, and a lot of the old rivalries appear to have died off due to retirements and, well, death. That’s good, because the industry needs a better standard than Gerber. Thanks in part to his son, Joseph Gerber’s name and many contributions will hopefully never be forgotten. But it’s time his namesake data format is.