Intelligent data transfer: are we smart enough to make the change from Gerber?

MY MARCH 2004 column on intelligent data transfer generated some interesting comments about the slow transition away from Gerber. Many users indicated that they were still using Gerber because it was provided "free" in their CAD system, while intelligent data interfaces are not. But I couldn't find anyone on the CAD side of the market who admitted charging for the intelligent data interfaces. It troubles me that we would hold back progress just for the sake of getting something at no charge.

Free Gerber viewers have promoted a false sense of security by letting designers think that they can verify Gerber data with a quick glance at the screen. For example, it is virtually impossible to see a 0.003" solder mask sliver problem when scanning a job in a viewer with the naked eye. Is a flee viewer worth risking losses in the thousands because we didn't fully validate DFM? It will come back to bite us later at 10 times the cost of a decent DFM tool.

But everyone has developed their processes around Gerber data, and we have become familiar with all the idiosyncrasies that come with it. We have become so set in our ways that we fail to see the true benefits that intelligent data transfer brings. First, we must understand that a few CAD systems still can't produce an IPC-D-356 file natively; some that appear to do so actually utilize a third-party script or program (not maintained or warranted by the CAD vendor) to generate it for them. If they don't provide any means for producing the IPC-D356 file, we are forced to use a CAM or DFM tool to accomplish the task. Remember, if the Gerber data wasn't good coming in, the netlist that comes out isn't going to be good either.

Next, consider that outputting IPC-D-356 is one post-process, while outputting Gerber is yet another. This leaves us with two "translated" views of our board data, each with its own issues. Toss in drill data and we now have a third post-process translation. Plus, we have three separate files, each in their own format.

Now compare the steps above to that of using, say, the IPC-2581 Offspring format. There is still a translation process, but now we are talking about just one translation, and the file format is much better suited to handling the complexities of today's board layouts. An intelligent data transfer format is designed to streamline the flow of information from design to manufacturing and support the ever-increasing complexities of our products. It is geared towards eliminating as much of the reverse engineering and double processing as possible by allowing each individual segment of the manufacturing process to obtain what they need to know about your design in order to supply you with an accurate and quality product.

For example: A designer uses rounded rectangles for the lands on his surface mount devices. Invariably, when Gerber data is produced, these rounded rectangles often end up drawn, even if 274X was chosen during output. The data gets read to a viewer and the picture on the screen looks pretty, so off it goes to the board shop. The board shop is, at some point, going to want to run a netlist generation from the Gerber data, either for comparison purposes later or DRC/MRC purposes now. Either way, in order to get an accurate netlist generation, the CAM operator is typically forced to run a draw-to-flash conversion on all those drawn rounded rectangles because that is the only surefire way for many of the CAM tools to ensure that they recognize those pad locations as being the end of a net.

No matter how automatic that draw-to-flash conversion process might be, it still takes time to perform and there are still opportunities for error during that conversion process. Had that same board shop received something like Offspring, the netlist would have already been there to begin with and the rounded rectangles would have been placed using the native rectangular pad shape with a rounded corner option enabled.

Technology itself may force us to change. Many of you are designing products that use increasingly higher frequencies and/or are being even more miniaturized. This has led to increased interest in embedded passive technology. How do you plan on sending your embedded passive designs out for production? Are you just going to continue to provide Gerber? It's doubtful. EP technology is nothing new, but many experience difficulty building boards with it.

Passive materials can be quite reliable and can solve problems related to high-speed and miniaturization, given that the proper amount of information is passed from design to manufacturing and the right techniques are then used to produce the board. Gerber can show us a picture that represents our passives, but not an intelligent list of the passives' reference designators or target values, or any other material characteristics that might help, such as additive, subtractive, preferred vendor material, etc. Some of this can be provided in a text file, but the most important parts, like device values, are missing. Intelligent data transfer is going to be one of the true means for solving this problem, and could be what ultimately forces you away from Gerber.

So, do you start working towards a transition to intelligent data transfer, or do you just keep using Gerber and wait for necessity to cause change? Personally I'm not waiting. Join me on a journey towards the future. It can happen if we all do it together.

JEFF MILLER is vice president, sales and marketing at Wise Software Solutions (www.wssi.com). He can be reached at [e-mail]. His column appears quarterly.