Reading time ( words)
Editor’s Note: This response refers to an article by Mentor Graphics’ Julian Coates, which ran in the February issue of The PCB Magazine, and was received in early June. Coates was given the courtesy of a rebuttal so that they could be published side-by-side in the same issue. The rebuttal follows.
In a recent an article by Julian Coates of Mentor Graphics, Smart Data Formats Automate CAD/CAM (The PCB Magazine, February 2014), in which Coates promotes more widespread adoption of the ODB++ format, the arguments he uses indeed make it seem like OBD++ is the great panacea for our industry, one that promises to eliminate all problems for CAD-to-CAM data transfer without any downsides.
In order to promote ODB++, Coates unfortunately reverts to Gerber-bashing rather than explaining the strengths of ODB++. And his arguments are highly misleading, as they are based on some tired old fallacies that I would like to address here. Before starting, though, it's important to clarify that when referring to Gerber, I mean RS-274X Extended Gerber, the current Gerber format. This supersedes the earlier RS-274-D Standard Gerber format, which is obsolete. Bashing RS-274-D Standard Gerber is like railing against Windows because MS-DOS only allowed eight-character file names. If Coates wants to bash RS-274-D, I'll gladly join him. Having said this, less than 2% of all jobs are transferred using the old format, so it's practically a non-issue.
Julian Coates’ Rebuttal:
With respect to Karel, I think he may be missing the main point. Consider this:
- No doubt Gerber is a very fine format for defining the graphical layers of a PCB
- IPC-D-356 is perfectly fine for defining a netlist
- Excellon needs no improvement; it defines the location and diameter of drilled holes quite well.
- Component placement lists can define component positions and rotations quite well also
- PDF is a good format for rendering drawings
- GenCAD and FATF are good for defining the parts of a PCB assembly for testing purposes
- Word is good for capturing text, especially “Readme” documents that explain to a CAM engineer how all of the above file-types should relate to each other, and how to reintegrate all that data back together so as to enable an efficient software-driven new product introduction (NPI) process.
Certainly, if all you want is accurate graphical data, then I am sure Gerber meets the requirement, and Karel is to be congratulated on his perseverance in improving that particular 50-year-old NC format. At a recent industry debate on this topic, he suggested that the best way forward is to use Gerber for the graphical data and another format for all the other information that Gerber cannot carry. Thus, he promotes the idea of intelligent, all-encompassing formats for carrying data, but excluding the graphical part. Why reject the advantage of having all of that other information linked to the graphical objects as well, and vice-versa? The problem that needs solving is taking all of that fragmented data into a single coherent model comprising both the PCB bareboard and the assembled PCB. Keeping parts of the product-model separate for simplicity is fine if you are only interested in a narrow subset of the PCB product-model, but it is a big problem if you need a complete definition of the product, as do all DFM and NPI engineers! There is no escape from the fact that, sooner rather than later, the data must be integrated. Read the full article here.
Editor's Note: This article originally appeared in the July 2014 issue of The PCB Magazine.