Re: [PEDA] bug plotting rotated fill
At 11:24 AM 12/18/01 -0800, Brad Velander wrote: Abd ul-Rahman, have you noticed any of the rotated fills where the drawn gerber generation leaves gaps in your otherwise solid fill? Like a database that I am looking at right now which has approx. 0.5 mil gaps in the gerber draws of a Protel fill. Not pretty! Drawing rotated fills was basically a mistake. Further, it is not easy. Protel draws tight fills, i.e., the draws exactly abut their neighbors. When you rotate a fill, however, the line endpoints typically become irrational numbers, so there is going to be roundoff error, plus there is the photoplot rounding. To compensate for this, it would be necessary to plot oversize by the appropriate amount, and to correct the fill outline accordingly. *Much* easer to define the macro and specify the rotation, clean and simple. At least I think it is, I haven't tried it. I will. [EMAIL PROTECTED] Abdulrahman Lomax Easthampton, Massachusetts USA * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * To post a message: mailto:[EMAIL PROTECTED] * * To leave this list visit: * http://www.techservinc.com/protelusers/leave.html * * Contact the list manager: * mailto:[EMAIL PROTECTED] * * Forum Guidelines Rules: * http://www.techservinc.com/protelusers/forumrules.html * * Browse or Search previous postings: * http://www.mail-archive.com/proteledaforum@techservinc.com * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Re: [PEDA] bug plotting rotated fill
Leonard I had a case a few months ago that seems might be related to this unit conversion problem. I had created a donut shaped fine pitch shield made up of concentric 4mil arcs, each 8 mil radius larger than the previous arc. The two sides of the PCB had traces lined up with the spaces on the opposite side. Each side had about 80 concentric arcs. About 30 arcs into the structure an error occurred. An arc was skewed. It crossed other arcs. The matter was not noticed until the board house examined the files with an independent gerber viewer. Protel 99SE and Camastic both showed this arc in the proper position while Gcprevue showed it crossing other arcs. To fix it I drew the arcs in metric units and the arcs appeared properly in both Camtastic and GCPrevue. It would seem that Protel 9SE and Camtastic must share some unit conversion code that has a bug. Dave Eloranta [EMAIL PROTECTED] -Original Message- From: Leonard Fischer [mailto:[EMAIL PROTECTED]] Sent: Monday, December 17, 2001 5:13 PM To: 'Protel EDA Forum' Subject:Re: [PEDA] bug plotting rotated fill Wow. That is interesting and kind of amazing that it is workable. And the first paragraph where you just said it was a board with submicron precision really threw me for a loop. Question/Suggestion/Possibility: It seems to me that part of the issue here is that Protel and Camtastic are both designed for larger physical boards. Ideally of course Altium would fix the precision issues. If they don't, or don't quickly enough to do what you need, could you work at some multiplication/magnification of the final scale, say 10x, then just photographically reduce the Gerbers? I'm assuming there is some way to do that (Camtastic?), not that I know how. This could help with the precision issue in Protel and the magnification in Camtastic. Just like working with 4x decals and tape! I'm also assuming that the ceramic module is relatively small compared to a PC Board - I'd be interested in hearing more about the ceramic module, like how big it is and what kind of components you put on it, if that's not proprietary. Len Fischer Trax Softworks -Original Message- From: Abd ul-Rahman Lomax [mailto:[EMAIL PROTECTED]] Sent: Monday, December 17, 2001 11:10 AM To: Protel EDA Forum Subject: [PEDA] bug plotting rotated fill I'm working on a board which requires high precision, in the submicron range. Theoretically, Protel supports photoplotting with four-place decimal metric, i.e., one-tenth micron resolution; however, I ran into some problems with PCB and with CAMtastic. Rotated fills, which are not flashed but which are drawn, were seen to have some stray draws. I've sent files to [EMAIL PROTECTED], but for our bug list, I am also reporting the matter here. That rotated fills (and pads, I expect) are drawn instead of being flashed is a bit distressing; this board already generates a lot of data and drawing those fills will make the gerber files huge. Rectangles (and other shapes) can be created with RS-274X Aperture Macros, which include a rotation parameter. They would then also be importable as single primitives (with appropriately modified gerber import routines.) Fortunately the fills only occur on one layer which will not be fabricated with the PCB (it is a multilayer co-fired ceramic module) but will be added later as a separate metalization on the top of the module. Then, looking at the files in CAMtastic, the aperture tables did not properly import, there were errors in aperture creation on the order of 1.5 microns. This is a bit hard to fathom; a metric 150 micron draw embedded in RS-274X should become, one would think, a 150 micron aperture in the CAMtastic aperture table, but it doesn't, instead it is 149.9 microns. (Some other apertures had higher deviations.) I'd guess that metric apertures are being translated to inches on creation, then displayed back as metric. But the aperture tables were editable to the correct values, so it is not that the database does not support sufficient precisions. However, the maximum zoom level in CAMtastic does not provide as high a magnification as does PCB, it was barely adequate, if cramped, for working in the micron range. PCB provides plenty of magnification before it refuses to zoom in further, it looks to me like roughly 0.1 micron per pixel. [EMAIL PROTECTED] Abdulrahman Lomax Easthampton, Massachusetts USA * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * To post a message: mailto:[EMAIL PROTECTED] * * To leave this list visit: * http://www.techservinc.com/protelusers/leave.html * * Contact the list manager: * mailto:[EMAIL PROTECTED] * * Forum Guidelines Rules: * http://www.techservinc.com/protelusers/forumrules.html * * Browse or Search previous postings: * http://www.mail-archive.com/proteledaforum@techservinc.com * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Re: [PEDA] bug plotting rotated fill
At 07:52 AM 12/18/01 -0600, Dave Eloranta wrote: It would seem that Protel 9SE and Camtastic must share some unit conversion code that has a bug. I doubt that they share code. I'd be interested to look at the actual gerber generated with the arc problem that Mr. Eloranta reported. I looked on the Knowledge Base and was unable to find any articles relating to these problems. But it is not necessarily easy to find such To my mind *any* plot errors are mission-critical. I did receive an immediate response from Michelle Covington at Protel US support that she verified the problem. I did make it easier for her by providing not only a cut-down edition of the problem board file, but also the CAM definitions. In any case it was nice to get such a prompt response. There should be only one round-off in the process: plotting to less than full precision or different units (or export to database formats of lower precision), and round-off should never be truncation but nearest value, even preferred. Conversion of high-precision units to plot at lower resolution could cause various kinds of issues. For example, if a fill draw is rounded down in sizeand position, gaps could appear in fills unless the fills were overlapped; this should receive special attention. For some reason, fill routines tend to have had, in my experience, problems. Tango never did get it completely right, as I recall. And even when the fills were right, they tended to have about twice as many primitives as necessary, plus they had problems with intersecting outlines. Protel does not support, as I noted, true rotation of pads through the plot process, they are not flashed, but they are drawn, with an aperture which we do not control in RS-274X (we might be able to control it if we use an aperture table, which has other problems). So we might think we can make a sharp corner with a fill, but it actually turns out rounded if the fill is other than orthagonal. [EMAIL PROTECTED] Abdulrahman Lomax Easthampton, Massachusetts USA * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * To post a message: mailto:[EMAIL PROTECTED] * * To leave this list visit: * http://www.techservinc.com/protelusers/leave.html * * Contact the list manager: * mailto:[EMAIL PROTECTED] * * Forum Guidelines Rules: * http://www.techservinc.com/protelusers/forumrules.html * * Browse or Search previous postings: * http://www.mail-archive.com/proteledaforum@techservinc.com * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Re: [PEDA] bug plotting rotated fill
At 03:12 PM 12/17/01 -0800, Leonard Fischer wrote: If they don't, or don't quickly enough to do what you need, could you work at some multiplication/magnification of the final scale, say 10x, then just photographically reduce the Gerbers? I'm assuming there is some way to do that (Camtastic?), not that I know how. The gerbers will be provided to a fabricator who could certainly scale them down. It's really only a matter of manipulating an RS-274X parameter. This could help with the precision issue in Protel and the magnification in Camtastic. Just like working with 4x decals and tape! Yes, though I always worked at 2X. I did do some magnified designs with Tango years ago. I wrote a little routine that would magnify a board, it is not a difficult problem. It gets a bit more complicated with Protel, what with design rules, etc., etc. But still doable. I think it could be done in the spreadsheets. I'm also assuming that the ceramic module is relatively small compared to a PC Board - I'd be interested in hearing more about the ceramic module, like how big it is and what kind of components you put on it, if that's not proprietary. Some of it is quite proprietary, NDA and all that. But the process is not proprietary. There are a number of fabs, among them National Semiconductor: http://www.national.com/appinfo/ltcc/0,2583,364,00.html This particular module has flip-chip ASICs on one side and the other side will have micro-machinery added, yes, moving parts The design requires using vias at closest spacing, so no tracks between vias in substantial areas of the module. Because the module must be designed for more than one fab, the design must meet worst-case design rules, which in this case are 2 mil track and space and 4 mil vias, with 12 mil space between vias (the via sizes do not include catchpad, otherwise it would seem that you could put track between them. No.) Because the ASICS are basically BGAs at spacings inadequate for track between vias, routing is severely constrained and reassignment of gates and drivers was necessary to create a routable design, even with 21 layers. The module is 34 mm wide by 127 mm long maximum, i.e., 5 inches. So I could work, if necessary, at 10X scale. There were certain aspects of this job which show the power of the Protel spreadsheet for creating complex patterns, I've thought of writing an article about it. However, the design also made very blatant a major bug with blind and buried vias, that they display on layers where they do not exist. Fortunately, this design turns out to be conceptually easy to understand; otherwise it would have been a nightmare and I would have had to move it to another CAD system that properly handles blind/buried vias. Right now I consider the blind/buried via display problem the worst outstanding issue A workaround would be to create a pad stack using pads co-incident with the via, then suppress the via display. But in this case the database would become truly huge, it is already a bit cumbersome to work with, one reason I am now assembling a faster computer. [EMAIL PROTECTED] Abdulrahman Lomax Easthampton, Massachusetts USA * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * To post a message: mailto:[EMAIL PROTECTED] * * To leave this list visit: * http://www.techservinc.com/protelusers/leave.html * * Contact the list manager: * mailto:[EMAIL PROTECTED] * * Forum Guidelines Rules: * http://www.techservinc.com/protelusers/forumrules.html * * Browse or Search previous postings: * http://www.mail-archive.com/proteledaforum@techservinc.com * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Re: [PEDA] bug plotting rotated fill
Abd ul-Rahman, have you noticed any of the rotated fills where the drawn gerber generation leaves gaps in your otherwise solid fill? Like a database that I am looking at right now which has approx. 0.5 mil gaps in the gerber draws of a Protel fill. Not pretty! Sincerely, Brad Velander. Lead PCB Designer Norsat International Inc. #300 - 4401 Still Creek Drive, Burnaby, B.C., Canada, V5C 6G9. Tel (604) 292-9089 (direct line) Fax (604) 292-9010 Website: www.norsat.com -Original Message- From: Abd ul-Rahman Lomax [mailto:[EMAIL PROTECTED]] Sent: Tuesday, December 18, 2001 10:32 AM To: Protel EDA Forum Subject: Re: [PEDA] bug plotting rotated fill SNIP To my mind *any* plot errors are mission-critical. I did receive an immediate response from Michelle Covington at Protel US support that she verified the problem. I did make it easier for her by providing not only a cut-down edition of the problem board file, but also the CAM definitions. In any case it was nice to get such a prompt response. SNIP [EMAIL PROTECTED] Abdulrahman Lomax Easthampton, Massachusetts USA * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * To post a message: mailto:[EMAIL PROTECTED] * * To leave this list visit: * http://www.techservinc.com/protelusers/leave.html * * Contact the list manager: * mailto:[EMAIL PROTECTED] * * Forum Guidelines Rules: * http://www.techservinc.com/protelusers/forumrules.html * * Browse or Search previous postings: * http://www.mail-archive.com/proteledaforum@techservinc.com * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Re: [PEDA] bug plotting rotated fill
Wow. That is interesting and kind of amazing that it is workable. And the first paragraph where you just said it was a board with submicron precision really threw me for a loop. Question/Suggestion/Possibility: It seems to me that part of the issue here is that Protel and Camtastic are both designed for larger physical boards. Ideally of course Altium would fix the precision issues. If they don't, or don't quickly enough to do what you need, could you work at some multiplication/magnification of the final scale, say 10x, then just photographically reduce the Gerbers? I'm assuming there is some way to do that (Camtastic?), not that I know how. This could help with the precision issue in Protel and the magnification in Camtastic. Just like working with 4x decals and tape! I'm also assuming that the ceramic module is relatively small compared to a PC Board - I'd be interested in hearing more about the ceramic module, like how big it is and what kind of components you put on it, if that's not proprietary. Len Fischer Trax Softworks -Original Message- From: Abd ul-Rahman Lomax [mailto:[EMAIL PROTECTED]] Sent: Monday, December 17, 2001 11:10 AM To: Protel EDA Forum Subject: [PEDA] bug plotting rotated fill I'm working on a board which requires high precision, in the submicron range. Theoretically, Protel supports photoplotting with four-place decimal metric, i.e., one-tenth micron resolution; however, I ran into some problems with PCB and with CAMtastic. Rotated fills, which are not flashed but which are drawn, were seen to have some stray draws. I've sent files to [EMAIL PROTECTED], but for our bug list, I am also reporting the matter here. That rotated fills (and pads, I expect) are drawn instead of being flashed is a bit distressing; this board already generates a lot of data and drawing those fills will make the gerber files huge. Rectangles (and other shapes) can be created with RS-274X Aperture Macros, which include a rotation parameter. They would then also be importable as single primitives (with appropriately modified gerber import routines.) Fortunately the fills only occur on one layer which will not be fabricated with the PCB (it is a multilayer co-fired ceramic module) but will be added later as a separate metalization on the top of the module. Then, looking at the files in CAMtastic, the aperture tables did not properly import, there were errors in aperture creation on the order of 1.5 microns. This is a bit hard to fathom; a metric 150 micron draw embedded in RS-274X should become, one would think, a 150 micron aperture in the CAMtastic aperture table, but it doesn't, instead it is 149.9 microns. (Some other apertures had higher deviations.) I'd guess that metric apertures are being translated to inches on creation, then displayed back as metric. But the aperture tables were editable to the correct values, so it is not that the database does not support sufficient precisions. However, the maximum zoom level in CAMtastic does not provide as high a magnification as does PCB, it was barely adequate, if cramped, for working in the micron range. PCB provides plenty of magnification before it refuses to zoom in further, it looks to me like roughly 0.1 micron per pixel. [EMAIL PROTECTED] Abdulrahman Lomax Easthampton, Massachusetts USA * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * To post a message: mailto:[EMAIL PROTECTED] * * To leave this list visit: * http://www.techservinc.com/protelusers/leave.html * * Contact the list manager: * mailto:[EMAIL PROTECTED] * * Forum Guidelines Rules: * http://www.techservinc.com/protelusers/forumrules.html * * Browse or Search previous postings: * http://www.mail-archive.com/proteledaforum@techservinc.com * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
