kicad-developers team mailing list archive

[Lorenzo Marcantonio <l.marcantonio@xxxxxxxxxxxx>]

On Mon, 30 Aug 2010, Alex G wrote:

> Would it be possible to post the specs (and maybe the LP calculator) in
> a public place without putting yourself at a legal risk?

The calculator was free to download, so I can't see any problem in this.
Also it's an 'anonymous' file, no license file or stuff needed. Usually
for free retired downloads there are no big problems in keeping them
available.

The spec were paid and licensed (there's my name on each page :D)
so I can't supply them (you can find the old revision on the net, if you
look carefully:D). The new revision has unified the tables so it works
with both reflow and wave processes with both lead and lead-free solder.

Also nothing prevents me to do a 'reference' implementation containing
the formulas (an the constants) to pick up to build a complete app.

In short there aren't a lot of things of interest (for land patterns!)
in that standard... all the info needed for the calculations is
comprised in pages 6--18!

There are four important steps to do:
- How to 'sum' tolerances (they call it 'profile tolerancing method'),
  because there are manufacturing tolerances (body tolerance+pin
  tolerance), a placement tolerances and land tolerances for
  the PCB.
- How much the fillets (toe, heel and side) are to be for a given
  manufaturing level (minimum, nominal, maximum) to achieve a good
  production yield. That's the most important stuff since there is
  a table for each component (chip, soic, melf, and so on) with
  *empirically tested* values (i.e. these people soldered *a lot* of
  boards with different fillets to determine the best value). Courtyard
  is also determined in this step.
- The determined land size is then checked for clearance (it's a process
  parameter, usually between 4 an 8 mils) and eventually trimmed (also
  if the body is low on the board the pad is trimmed against it to avoid
  solder creep under the package).
- At the end everything is rounded to a 0,05mm grid (that's easy but
  look out for clearance :D)

So it's an almost-trivial arithmetic routine (the most difficult thing
in there is a square root) with *a lot* of cases in the second step. It
could even be done in FORTRAN without troubles :P

The LP calculator would then be useful for cross checking the results.

If you have some public place to put up the msi I'll just upload it. Ah,
it requires .NET2, some guy even made it work under wine, it seems.

--
Lorenzo Marcantonio
Logos Srl