Mail Archives: geda-user/2014/11/19/22:06:46
I had some free time last night and took a look at the export filters
on LP699502. Here are some observations :
- My usual workflow for obtaining a 3D model of a PCB is :
(a) Export ps from pcb in multifile.
(b) Convert ps to dxf using "pstoedit -f dxf <fname>.top.ps <fname>.dxf
(c) Import dxf into a sketch in solidworks
(d) Extrude PCB from a sketch overlapping the outline
(c) Cut out the necessary mounting holes and just enough pins to be
able to assemble the mecahnically significant components (connectors,
etc)
- Using export to "OpenSCAD" (Bert's exporter),
(a) I get a syntax error. "Parser error in line 1514: syntax
error", with the following line highlighted: "include
<PACKAGES.scad>". I did a git checkout on files that changed while I
was looking though the models, but that doesn't seem to have helped.
(b) From glancing through the comments in the code, it seems to me
like this only generates a model with a rectangular PCB outline. This
is a showstopper.
(c) I have so far only been trying to get the unpopulated PCB in 3D
so far, and combined with the syntax error I haven't really paid much
attention to the ability to import models. I like the approach used by
this exporter for handling models, though there are various quirks
which probably will need me to make code-level changes to make it work
with my setup. These have to do with footprint naming conventions and
matching the footprint and model origins. It would be nice to have a
offset that can be written into to the model to specify the offset
from the footprint origin.
- Using export to "scad" (Milan's exporter).
(a) Export seems to work fine at a basic level (again, I only know
about the bare PCB. I have not yet looked at component models and how
they are picked up. It isn't obvious to me at this point how the
exporter finds the models, but that's likely because I haven't really
looked for any substantial documentation besides just briefly skimming
through the code.
(b) I export with Silkscreen (green), all pins and pads with gold
plating, no copper otherwise, and with outline as outline layer.
(c) Opening the resulting .scad file on openscad produces a very
nice image at first glance for small, simple PCBs.
- SCAD workflow (Milan's exporter, possibly even Bert's, though I
didn't get as far with Bert's so I don't know)
(a) The SCAD that is generated opens on openSCAD for small, simple
PCBs. This render is pretty, nice, and correct, though not
particularly useful in that format.
(b) Complicated PCBs don't seem to be rendered immediately on
opening openSCAD for some reason. As an aside, complicated PCBs won't
even compile without increasing the maximum number to elements to
render from the default 2000 (I increased it to 100000. The PCB I'm
using seems to have about 18000 or so).
(c) Compiling the file to CGAL (F6) (which is necessary to be able
to export to STL, DXF) takes forever. It took 41 minutes on a small,
rectangular 4 cm2 PCB with less than 10 components on it on my machine
(i7, 4G RAM). The limiting resource seems to be RAM. The larger PCB is
still compiling, many hours now, and has been stuck on 92/100 for
atleast 2 (though there is still the occasional scroll in the log, so
it isn't actually stuck).
(d) Once compiled, for some reason, all the colors seem to be
messed up. The majority of the surface of the model shows up as gold.
The interior surfaces of the plated drills seem to be green, or
atleast there's a few slivers of green visible.
(e) FreeCAD always takes an indefinite amount of time to open the
openSCAD directly. Even the small PCB. I gave it about an hour or so
and gave up. This is after selecting maximum number of edges in
polylines to 6.
(f) FreeCAD imports the STL in a reasonable length of time -
relatively instantaneously. However, the STL imported is devoid of all
color information. It shows up just as a block of grey. When I change
the display to wireframe I see all the necessary elements are there -
even the text on the silk. They're all the same color, though, so
aren't immediately visible. I should note that the exported STL was
from a compile which resulted in a render with already messed up
colors. It's possible that the problem lies there and not in the
import.
(g) Export to CSG from openSCAD is quick. Import of CSG to freeCAD
is again infinitely long. I have not seen it complete.
- Given these observations, I think SCAD may not be the correct format
to be attempting to do this in. The only real advantage I see with
SCAD is interoperability with the limited SCAD environment. From the
technical perspective, having a full mesh _may_ make vibration and/or
thermal simulations nicer, but that's going to need a whole lot of
work to model the solder joints and so forth. IDF may indeed be a
better bet, with IDF-SCAD conversions scripted outside of PCB. We're
not talking about real-time 3D renders within PCB anyway - which is
what openSCAD and similar native 3D formats would afford us. Since
we're only really taking component placement information from PCB and
providing everything else externally, it's probably easier (and
cleaner) to formalize that relationship in IDF.
On Tue, Nov 18, 2014 at 11:29 PM, Britton Kerin <britton DOT kerin AT gmail DOT com> wrote:
> On Mon, Nov 17, 2014 at 1:33 PM, Levente Kovacs <leventelist AT gmail DOT com> wrote:
>> On Mon, 17 Nov 2014 22:35:59 +0100
>> Bert Timmerman <bert DOT timmerman AT xs4all DOT nl> wrote:
>>
>>> Short-comings, weak points, strong points: are all very valuable, so
>>> we can iron out them flaws and improve pcb.
>>
>> I checked out your branch, and I think it is super cool!
>>
>> I have the following questions:
>>
>> * Is there any support for 3D models of a component?
>> * If yes, how can make your exporter find the models?
>> * It renders plated holes as copper areas. Is this intentional?
>> * Unplated holes rendered as copper covered by something.
>> * the 0 point of the layout is the 0 point of the PCB file which is IMHO
>> not the best way. I would make it the center of the outline.
>
> I would think it would be better to have them agree, at least as far as
> displayed coordinates are concerned. If you need to pick a different
> point as the default center for rotation or something that's ok.
>
> Britton
--
Chintalagiri Shashank
Indian Institute of Technology, Kanpur
http://blog.chintal.in
- Raw text -