Geoff, Thank you for the comments.
Yes, I did suggest rather radical surgery in the quest for PERFECT GERBERS. PERFECT GERBERS is asking for a lot but that is what we need. The problem with using Aperture Macros is that it implies a tremendous software task. How can KiCAD verify design rules on planes if the aperture info in KiCAD is not EXACTLY THE SAME as the patterns in the photo plotter? I don't think this can be done. :-( Radical? Maybe but do we intend to detect and flag "Cutoff" or "Missing Spokes" in planes? If so they we already did all the hard work with thermals so drawing them with simple shapes should be easy. If we don't completely error check plane rendering then KiCAD plane generation is very weak. To draw a thermal all we need is a center CIRCLE for the inside PAD, the region outline drawn with arcs and lines, and spokes. Verify of spoke to region outlines, other spokes, drills or other nets is easy. Now we have perfect circular plane voids and full error checking without Aperture Macros. Maybe I missed something? Also, the circular void as line segments issue needs a resolution. Do we chop an arc into segments until we get some max error? Maybe .01mm? A big pad is going to give a lot of segments. I would suggest simply using arcs. The ARC commands are very solid, they are pretty well defined and have been there since -D. I can accept the idea of using ARCs to draw the outline of a region, this will give perfect arcs limited only by the photo plotter resolution. One could then use -X polygons to fill in the main body of the region just as scan lines can be used. I found -X Poly operation rather consistent though scan lines are less complex. Summary: Just Say No to Aperture Macros Outline copper regions with Lines and Arcs. I am open to ideas. Thanks, Bob Kondner -----Original Message----- From: kicad-users@yahoogroups.com [mailto:[EMAIL PROTECTED] On Behalf Of gharlandau Sent: Friday, September 21, 2007 1:02 PM To: kicad-users@yahoogroups.com Subject: [kicad-users] Re: composite layers and negative plots I fully agree that there are many aspects of the RS274X specification that leave a lot to be desired, but I'm still not convinced of the merits of totally forsaking the use of Aperture Macros and Polygon commands. The copy of the RS274X standard which I have (an Acrobat file which I downloaded from a website some years ago) makes no specific reference to being able to use Arc commands in conjunction with Polygon commands, so I concur that it would be desirable to avoid using that combination. And I also think that it would be highly desirable to avoid incorporating different layers within the same Gerber file (re using both "dark" and "clear" polarities). But I think that totally avoiding the use of Aperture Macros could be taking things too far. As long as the value of each parameter (for each primitive that is used within an Aperture Macro) is explicitly defined, and no "algebraic substitutions" are required in that regard, I would regard the usage of Aperture Macros as acceptable (assuming that no attempt is being made to "push the limits" of the RS274X specification at the time). A really "extreme" view of Gerber files is that they should *never* incorporate *any* Arc commands either, least some PCB manufacturers not be able to cope with them. Would you advocate that as well? I have not really looked at the source code for generating Gerber files so far, but when I get a chance, I will take a look and see if there are any aspects where improvements could be made to at least reduce (and preferably totally eliminate) the possibility of Gerber files' contents ever being misinterpreted. In that regard, I would regard it as preferable to use an Aperture Macro (incorporating a type 21 / "centered line" primitive) to define an aperture for a pad having a rectangular shape on a non-orthogonal angle, rather than using a Polygon command, and there could well be various other aspects which could also be improved. And maybe something else which should be considered at some stage is providing the feature of generating (sets of) ODB++ files from PCB files. While I am not currently an expert on such files, it is still my understanding that they don't suffer from the shortcomings that afflict Gerber files (due to the shortcomings of the RS274X standard). However it is also my understanding that many PCB manufacturers are not yet equipped to handle such files, and I gather that some of them haven't even heard of such files. But doubtless any PCB manufacturers who actually can cope with such files would still prefer to receive them whenever that is possible. Regards, Geoff Harland. "Robert Kondner" wrote: > > Hi, > > That RS-274-X spec has more holes than Swiss Cheese. > > For example: If you start a Poly Fill and use a ARC I guess the > ARC becomes part of the the poly edge, correct? > > Doing so results in a polygon with circular voids. This is a huge > area for "Interpretation Errors" in gerber processors. Now add to > the fact that these polygons could be in clear or dark areas of a > single layer and the multiple layers COULD be combined. Good luck > finding 2 gerber processors that render this to the same film image. > > Chance are about 5% that you get a call from the PCB vendor. > Chances are 95% that you get back a bad PCB. Worse yet the location > for finding errors in circular voids is in an internal plane. > Shorts in a plane are like an unforgivable sin, nothing can save > you. > > After many years of doing PCB designs and the writing of a gerber > reader (MUCH more difficult than writing gerbers) my suggestion is > Keep It Simple (KISS Solution). > > The generation of gerbers and their resulting rendering to film > must be ABSOLUTLY PERFECT. We live and die by this level of > perfection. To achieve this level of perfection think of a gerber > as a vector driven bitmap. No Aperture Macros and no Polygons need > to apply. Even if 99% of the code in the industry reads and > processes PERFECTLY that 1% is a reason NOT to use them. > > We need ROCK SOLID routines for flattening a higher level > structure of shapes. The output is a set of scan lines that paints > the required image. > > I write in Delphi, I can easily generate .dlls, and the resulting > scan lines could convert directly to KISS generation of gerbers. I > have a simple gerber reader that is used to read from LOTS of > different apps. It took years learning what liberties every app > programmer could take with gerber generation. Lets NOT to the same > thing in DRIVING film plotters. > > Summary: > > We live or die with gerber generation. If we even need to talk to > a fab house about gerber processing we are dead meat. > > I would love to see a solution got this gerber generation issue. > > Thanks, > > Bob Kondner Please read the Kicad FAQ in the group files section before posting your question. Please post your bug reports here. They will be picked up by the creator of Kicad. Please visit http://www.kicadlib.org for details of how to contribute your symbols/modules to the kicad library. For building Kicad from source and other development questions visit the kicad-devel group at http://groups.yahoo.com/group/kicad-devel Yahoo! Groups Links
