[gta02-core] PCF50633 footprint and footprints in general

Werner Almesberger werner at openmoko.org
Sun Jan 10 15:01:56 CET 2010

Dave Ball wrote:
> fixed

Looks much better now, thanks ! The measurements on the pad area
are a bit hard to read. Maybe move the two 7.4 outside the chip
and put the 2.7 and the 0.7 into the now available white space ?

> >Ah, and please keep the silk on copper issue in mind.
> What's the best way to deal with this?  The silk rectangle I added
> is the chips actual outline, but the pads extend outside this.  I'm
> assuming the silk gets used for placement so needs accurate corners?

This is a problem that still needs solving.  Silk is currently used
for both package and keep-out area and it's often not quite clear
which is which. One problem is that keep-out should really be half
the distance or the resulting silk screen will look very confusing.

Here are some more thoughts on this issue:

The package outline is also useful even if we can't draw it on the
silk screen. KiCad had a nice 3D preview mode that uses package
drawings. With some effort, fped could generate suitable 3D models
from an outline and some height information.

> Also, this symbol doesn't have any orientation information - is
> there a prefered way of adding that?

We mainly have some sort of silk screen squiggle that maintains a
somewhat reasonably respectful distance to all pads. Examples:

- traditional: small circle as index marker, SC70-5
- tradition out-of-the box: small circle outside the outline,
  DFBM-CS320 (I'm not so happy with this one, since this may end up
  on neighbour components) or FH23-39S-0.3SHW (better since it's
  deep on keep-out land anyway)
- tradition reinterpreted: circle around a pad, FBGA96-P0.8
- getting creative: dent in the outline, WM8753FL
- getting more creative: bend in the outline, WLCSP12-P0.5
- do as the cavemen did and just draw what it is: SOD-523
- index side instead of pad: TDFN34-16

I don't know if there's any convention that covers all cases,
particularly the ones where there isn't much room for extra silk

One work-around may be to introduce a "special" mode to draw
this kind of technical data (similar to the package drawing) that
doesn't go to the silk screen layer. We could select this mode by
having different silk types (like we have different pad types) or
just by having "magic" frame names.

Invisible things could be translated to one of KiCad's technical
layers or not get exported to KiCad's footprint at all.

> Ahh. Cool.  The one that keeps throwing me are all the landing
> points for measurements; many of them when you click to drag a
> measurement all of the other points disappear.  It takes a bit of
> trial and error to find points that I can use to get the measurement
> I want.  Am I doing something wrong?

It's a bit like the electronic voting parody, isn't it ? :)

You're doing nothing wrong. There are three types of measurement
points: minimum of a kind, maximum of a kind, and smallest of a
kind but greater than the minimum of the same or another kind.

Before you select the first point, points that could be one of
these are shown. Once you've chosen the first point, only the
points that yield a possible combination with the first point are

E.g., if you first pick a minimum point, only all the maximum
points and all the points "following" the first point are shown.

If the point you pick is a maximum point, all the minimum points
are shown.

If the point you pick is a potential follower, all the minimum
points it follows are shown.

These rules aren't easy to apply consciously, but trial and error
tend to yield a valid combination quickly.

The reason for having such a complicated mechanism is that a point
may appear multiple times and I think it would be awkward if one
had to fully qualify which instance one wants to use.

E.g., for a maximum point, one would have to specify which value
from a table or loop the variables affecting its position assume.

This feature is somewhat experimental. Ideas for improvements are
always welcome.

- Werner

More information about the gta02-core mailing list