kicad-developers team mailing list archive

[Lorenzo Marcantonio <l.marcantonio@xxxxxxxxxxxx>]

On Wed, Aug 01, 2012 at 12:01:59AM -0700, Cirilo Bernardo wrote:
>   IDF3 will suffice for what you describe - providing rough sizes for major component so that the mechanical engineers can package the board properly, as well as hole positions and sizes. However, IDF 3 is too primitive for other jobs since components are represented as a vertical extrusion from a footprint shape on the board - I can't use it, for example, to verify that a right-angle surface-mount LED will align correctly with a hole in the panel (this is actually an issue in a current project of mine - I basically create a SolidWorks model from an IDF3 file then replace critical components with realistic solid models).  IDF 4 addresses many of the shortcomings of IDF 3, but I get the impression that industry has abandoned IDF 4 in favor of PSI-5 (which unfortunately is not a freely available specification).   Anyway, if anyone is really keen on at least exporting IDF3 and IDF4 from the KiCAD *.brd, I can help out to make sure the files are

The idea on IDF is that you actually want the model you cad importer can substitute it using the partnumber in the IDF file. Otherwise just take the 'path' extrusion. Of course that requires origin aligned models and IDF records.

The origin is an issue even for STEP and other stuff (pick and place too!) since there is no standard on where the origin is (centroid, pin 1? a DPAK centroid is on the body only or pins included?) or what the zero orientation is (well, there *is* the IPC standard for origin and orientation but afaik nobody uses it)

> I like to imagine that the software will be so sophisticated one day that there will be thermal models for components and tools to convert a board file to the input needed by the FEM tools (but maybe not in my lifetime ...)

Tools for generating meshes from board already are there. Thermal models for component are usually reduced to thermal resistance with air and with board. Without forced airflow that's a pretty good approximation. For some reason big transformers *can't* be approximated so easily (tried it, didn't work XD). Also I suppose that really knowing the math you could convince OpenFOAM or something similar to do thermal FEM analysis. We only made simulations with another firm (they used a 15000eur-and-something package).

After not-so-much satisfying results now we're doing 'simulations' glueing thermocouples everywhere (even in transformer pottings:P btw potting resin is a better thermal conductor than you can imagine...)

> Yes, a GUI for board thicknesses between copper layers and copper layer thicknesses would be useful in the future.  For now if you want to do something like simulate a stripline, you can only extract part of the data from a board file and the rest you have to enter manually.  I suspect most people using KiCAD won't be working with high-speed data lines and RF applications on most projects, but it's still nice to slowly build up such features.  Even in my plans for the MCAD demonstration I have notes telling me that I need to enter at least the overall board thickness and for flexibility I may need to consider copper thickness even though I can't think of any real use for it at the moment.

Actually board thickness is in the file! look in the vrml exporter, it uses it. Layer stackup is *not* stored (iirc mentor does a $4000 package *only* for stackup given impedances required). Sadly I don't fit in your 'mostly' category... even an USB2 line is a differential microstrip and must be done with a controlled 45/90 impedance and meanders to make it work. The 'must' is not a joke, it simply doesn't work otherwise:( With *really* high frequencies I fear Agilent is the only way to go...

-- 
Lorenzo Marcantonio
Logos Srl