-
- News
- Books
Featured Books
- design007 Magazine
Latest Issues
Current IssueOpportunities and Challenges
In this issue, our expert contributors discuss the many opportunities and challenges in the PCB design community, and what can be done to grow the numbers of PCB designers—and design instructors.
Embedded Design Techniques
Our expert contributors provide the knowledge this month that designers need to be aware of to make intelligent, educated decisions about embedded design. Many design and manufacturing hurdles can trip up designers who are new to this technology.
Manufacturing Know-how
For this issue, we asked our expert contributors to share their thoughts on the absolute “must-know” aspects of fab, assembly and test that all designers should understand. In the end, we’re all in this together.
- Articles
- Columns
Search Console
- Links
- Events
||| MENU - design007 Magazine
Karel Tavernier: Gerber Definitely Not "Dumb" Data
June 13, 2012 |Estimated reading time: 9 minutes
Dear Mr. Shaughnessy,
I write in response to a recent letter to the editor (click here, and here) published on your site, in which Julian Coates of Mentor Graphics describes the Gerber format and data as “dumb,” taking his cue from a comment about Gerber data in your column on data transfer formats. Mr. Coates is well known for his allegiance to the ODB++ format, and it is natural for him to support it. However, using this term for a format that is often described as “the backbone of the electronics industry” and is used daily by professionals all over the world, is both counterproductive and erroneous.
As a veteran of the printed circuit board industry, and as the managing director of Ucamco, owner and caretaker of the Gerber format, I would like to respond to Mr. Coates' letter, with the professionality and candour for which my company is widely known and respected. As Gerber is dedicated to the bare board industry, all of my comments refer to this part of the electronics manufacturing cycle.
First, what is Gerber? it is an image description language used specifically by the bare board industry that enables PCB designers to transfer accurately and reliably even the most complex PCB design data, including drill data. The de facto standard for the bare board industry, Gerber is the easiest data transfer format available to CAD designers and CAM engineers. It can be used by smaller, less sophisticated CAD systems as well as top-of-the-range flagship systems. It is simple to understand and use, and its output is easy for designers to check both visually and automatically with any of the myriad free or cheap Gerber viewers available to them, so that they know at a glance exactly what they are sending to their PCB manufacturing partners.
It is a superlative image description format. No more, no less. It does not, for example, provide netlist or stackup information. The first is a non-issue as the IPC-356 netlist format is a simple and perfectly adequate standard, so all the designer needs to do is to add an IPC-356 file to the Gerber archive. Here, if there is an issue, and it's one that really baffles me, it is that designers so often omit the netlist, leaving manufacturers to work it out for themselves. This is not a Gerber issue – leave out the netlist in ODB++ and you have the same situation.
Stackup could be improved, admittedly, but this data is actually pretty simple, so if the stackup is properly described in the documents using clear text files or file naming conventions, even the most complex stackup can in very little time be described informally, either manually or by using software such as that supplied by Ucamco.
Even with these limitations, Gerber is far more intelligent than Mr. Coates gives it credit for. Rather than dumb, it is simple. And in its simplicity it is an incredibly powerful, astoundingly accurate image description language – one that is capable of describing a PCB image to within a tiny 0.1 nm. Not that describing an image is easy. Far from it. In fact it's difficult to get it right – ask any graphics software developer just how frightening geometric software is. Reading an image 100% right 100% of the time is an even bigger challenge.
No, it's just that we've been at it for decades, long enough to have ironed out all the issues and to have made Gerber the tried and tested, easy-to-use format that it is today. In over 30 years of service to our marketplace, we have gathered a wealth of experience and know-how that gives us a deep understanding of our users' needs and their marketplace. We also take their input and issues seriously enough to incorporate their messages into our ongoing improvements programme, clarifying and adapting the Gerber format.
Thus we have developed it from its origins as an input format for plotters – which it still is – to the bare board industry's CAD-to-CAM data transfer standard. For our latest revision we have deliberately engaged an external consultancy that is conversant with imaging software but not with Gerber, so we receive an unbiased, fresh view on what is clear and what is not. All of this costs us a lot of time and money, yet we offer this to the bare board industry as a free service. The Gerber format can be freely accessed and downloaded by anybody who wants it, with no limitations and no need to register with a user group in order to be able to use it. Ucamco is also recognised as being the only company that does not use its control over a format to favour its software.
No wonder Gerber is used so universally.
ODB++, while it too is a perfectly valid data transmission format, could not be more different. To the point that it, and the policies surrounding its use, are considered by many to be somewhat Byzantine. Prior to using it, companies must first sign a non-disclosure agreement, be approved as users, join the ODB++ Solutions Alliance, and abide by the changing policies that govern ODB++ use. Then the format itself, which unlike Gerber, describes the fully assembled board, is anything but simple. On the contrary it is difficult to write and to read, demanding specialist knowledge and training.
Furthermore, the CAD systems used to write ODB++ files must necessarily be sophisticated and powerful enough to handle the task. Reading them is just as difficult – in fact, reliable ODB++ writers and readers, like that provided by Ucamco, are few and far between. This is an issue not only for CAM but also for bare board designers, because whereas they can easily read and check Gerber data for errors prior to sending it out, they cannot do so with ODB++ data. Too complex for direct, visual checking, ODB++ must be read by a reliable reader or system that the designer can trust.
This becomes even more important when you consider the dangers of software bugs. Each new image format will initially be plagued with them – a fact that the PCB industry knows only too well. Unidentified bugs in a CAD output or a CAM input image file often go undetected in CAM and can be an expensive, damaging waste of resources. It is for this reason that so many are reluctant to move their bare board production from Gerber to any other format. Indeed, those datasets that contain the data in both Gerber and ODB++ formats often contain a readme.txt file stating:
BARE BOARDS MUST BE FABRICATED WITH GERBER, DRILL AND IPC-356 NETLIST PROVIDED. BOARDS ARE NOT TO BE FABRICATED FROM ODB++ FILE.
In some cases where ODB++ is accepted, there may be a proviso to the effect that ODB++ be generated by Valor's Enterprise verification tool [rather than anything else]. This does not indicate that there is anything intrinsically wrong with the ODB++ format. It does however indicate a concern about the reliability of the newer ODB++ software.
And then we come to the issue of intellectual property (IP). Mr Coates argues that ODB++'s “one file” structure does not render it more prone to IP leakage than Gerber's “many file” structure. Unlike Mr. Coates, I see no great difference between the two structures: Just as ODB++ is a directory with many files zipped into one file, a Gerber archive consists of many files zipped into one file.
I do, however, agree with him that in principle, IP is as safe in ODB++ as it is in Gerber – it simply depends upon which data is passed on by the designer to the manufacturing partner. PCB manufacturers will need the bare board data whereas assembly partners will need component and other final product data. However, it is worth noting that whereas it is easy to see what is being passed on in Gerber format, the complexity of the ODB++ format makes data far less visible, so accidental data transmission and therefore IP leakage is a legitimate concern when using ODB++.
As Mr. Coates rightly points out, CAM engineers always read the Gerber data in their CAM systems, where it can be converted without hassle to ODB++. He uses this as proof, again correctly, that Gerber files will contain the same amount of IP as an ODB++ file for the same PCB. After all, one can manufacture a PCB starting from the Gerber files. But here, his argument that Gerber data is “dumb” collapses. For if a manufacturer can easily convert Gerber to supposedly intelligent ODB++, the intelligence must already be an integral part of the original Gerber files.
Having said all of this, we can debate data formats until the cows come home: This is not actually the main issue for our users. The principal concern for designers and manufacturers centres around data and design quality, and the ease of their implemention. Choice of data format is in a certain sense secondary, and will depend on the answers to questions such as: “Is this a proper design?” “Does it say what it should say?” “Is the data correct and accurate?” “Can it be used in a hassle-free way?” “Is it safe?” “Is it reliable?” “Can I read it, use it, and put it straight into my system quickly, easily and with minimal costs, whatever my equipment and software?” If the Gerber format is used, the answer to all these questions will always be a resounding “YES!”
More so if the format is used properly. Problems can, of course, arise with Gerber files, but these are always due to poor usage and implementation, and never to the format itself.
First, plain wrong Gerber files exist. So do plain wrong ODB++ files. In fact, being a more complex format, I would expect ODB++ files to contain more issues than the easier-to-read Gerber files.
Second, Gerber files can be poorly written, the main culprit being painted pads and areas. It probably happens in ODB++ files too, although it is less likely because companies taking on the daunting task of writing an ODB++ output must necessarily have the skills to do it properly. The Gerber output of these companies is probably pretty good too.
Third, and worst, is that some companies (unfortunately, some are big OEMs who should know better) continue to use Standard Gerber rather than Extended Gerber. Standard Gerber alone is not an image description language, so it must be supplemented with an informal image-defining “wheel file,” and it is so limited that designers must use painting. Standard Gerber offers not a single advantage over Extended Gerber but brings numerous disadvantages, so it's high time that all those companies still using the now obsolete Standard Gerber moved across to the Extended format.
The Extended format enables clear, unambiguous communication between designer and CAM engineer, and as a consequence, it guarantees higher throughputs, better manufacturing processes and yields, and higher quality products. As one PCB maker said,“If only we received proper RS-274X files, it would be a perfect world.”
On that note I'd like to close, with thanks to Mr. Coates for having encouraged me to clarify some of the real reasons why Gerber remains the de facto data transmission format for PCB design and manufacture. Rather than abuse and poor usage, Gerber deserves immense respect for providing the bare board industry with its clearest, simplest, fastest, most compelling and unambiguous data transmission language. And Ucamco deserves thanks and recognition for ensuring the ongoing evolution of this invaluable tool, and for continuing to provide it freely and free of charge to whomever wants to download it.
With thanks and kindest regards,
Karel Tavernier,Managing Director, Ucamco