We are currently evaluating some of our Smt assy processes. This includes
a review of land pattern dimensioning assumptions.     
 
  When applying the formulas for land dimensions on smaller chip style comp- 
onents (ie.RM1206's, CC0805's, etc.), I am having trouble with the "Gmin"
assumptions ('gaps'). Does anyone use these formulas actively? The premise 
seems elegant and sensible enough. When I calculate for the "Z" dimensions
(across the outer extents of the land pattern) and "X" dims (the pad width)
it seems to work out quite nicely. However; because of the large tolerances
on the terminations, the numbers at the gap come out seemingly too small.
Even by RMS'ing all the tolerances, the numbers don't 'move' enough to make
a useable pattern. The way I understand it is this: (to make this short,
I'll have to assume the abbreviations for dimension features are familiar)

  Gmin = Smax/rms (this is minimum comp length 'minus' 2x maximum 
         termination, then 'adding' the rss'd comp/Tol. and term/Tol. )

         minus 2x min. heel fillet supplied by IPC's empirical data.

         minus the "rss'd"  comp/Tol, fab/Tol, place/Tol.

  Is this correct?

  In the third step above, should the comp/Tol be 'Stol(rms)' per 3.3.1
"Component Tolerancing", pg.13?   
  
  Should the max/min heel fillet tolerance found in Section 8.2, "6.0 
Tolerance and Solder Joint Analysis", pg. 4 of 4., also be included in 
the above or (as I suspect), is this a resultant MMC/LMC condition in the 
fig. 4 table just showing what the fillets could be at their limits?

  I hope this isn't an overly confused question. I am willing to clarify
specifically. 
 
         James Patten, Product Design
         TRW, Space & Electronics Group
         Spacecraft Technology Division
         (310) 813-8914