Issue link: https://iconnect007.uberflip.com/i/378511
14 The PCB Design Magazine • September 2014 the simplest to the most complex, are still ma- nufactured using Gerber, which tells me that this is an image format that the industry trusts. And the industry is right to trust it—it's the best there is. Used properly, it delivers on its promi- ses, without fail, every time. So let's have a look at some of Coates' ar- guments. He quotes Viasystems as stating that, "about 25% of the data packages they receive have issues relating to: • Missing layers, fabrication drawings, drill files, etc… • Netlist format violations • Netlist exception violations." If true, this is indeed a sorry state of affairs, and needs rectifying. But if this is the extent of the problems, then there is nothing wrong with the format. Viasystems' issues are in fact due to some rather trivial bugs in the CAD ven- dors' implementations, so the solution is to fix the implementations rather than to adopt completely new software by switching to a new format. The article is not clear about whether these omissions and violations relate to ODB++ or Gerber files, or a mix of the two. However they arise, I can only recommend that Viasys- tems report these issues to their customers with a request to contact their CAD software suppli- ers. If the CAD software vendors fix these simple bugs, the issues will be resolved once and for all. If they are unable or unwilling to do so, there is no solution: neither in Gerber, nor in ODB++, nor anywhere else for that matter. Coates also mentions that Gerber files so- metimes contain syntax errors, low numerical accuracy and other errors. This is no doubt true, but again these are simply bugs in the Gerber output. Do we need a new format to fix syn- tax errors in the current one? Surely the solu- tion is to fix the bugs in the Gerber output. And ODB++ itself is not immune to syntax errors; if anyone would like some invalid ODB++ files, I can provide a few. The reality is that Gerber files very rarely generate the wrong image. This is because while only a few applications read ODB++ reliably, there are countless more that read Gerber with near-perfect reliability. This is because: • The Gerber format is simple • Its specification is well-written, easy to read, detailed and precise • Most of its implementations are mature • As it is so widely used, the implementations are thoroughly field tested, so most bugs have been ironed out • The format is supported by excellent free viewers such as GC-Prevue Advocating the adoption of a new and much more complex format to eliminate simple bugs is a very curious solution indeed. Consider only that a CAD software developer struggling to produce a simple Gerber file correctly is not mi- raculously going to write a bug-free implemen- tation for the more challenging ODB++ format. If one wants bug-free software it is best to stick with Gerber, as Gerber is a simpler and more mature format than ODB++, it is far less pro- ne to bugs, and its bugs are far easier to find and resolve. Switching to a new imaging format introduces a whole raft of new issues and bugs that would take many years to sort out. Imaging software is complex and takes a long time to get right. Adopting ODB++ to solve bugs in Gerber output is like using a sledgehammer to swat a fly: The solution is far more damaging than the issue ever will be. Table 1 summarizes Coates' claims regarding the benefits of ODB++ vs. Gerber. Here is my take on the aforementioned be- nefits: 1. False. A simpler, more reliable format in fact needs less diagnostics 2. False. An error can be more easily identified in a simpler format. 3. False. ODB++ is not miraculously error-free. 4. False. IPC-356 supports the actual customer net name. It may be that the software Coates uses does not display it, but this then is a problem in that software. 5. False. Gerber has negative apertures and so can handle planes perfectly. (I should add that this is the first time I see the claim that ODB++ is more compact than Gerber!) THE GREAT GERBER VS. ODB++ DEBATE continues feature