[gta02-core] PCF50633 footprint and footprints in general

Werner Almesberger werner at openmoko.org
Sun Jan 10 07:06:05 CET 2010

Dave Ball wrote:
> I've had another go, with the solder paste layer and using the PDF
> you suggested.

Much better !

> Is there a proper method for working out how much
> paste should go on the central gnd pad?

NXP's AN10365 [1] has a few useful advices on pages 14 and 15:

a) "The solder paste should cover approximately 20 % of the total land

b) "the solder paste pattern, including the spacing between the
   deposits, should have a coverage of 35 % of the land area"

c) "A HVQFN48 with an exposed pad of 5.1 mm × 5.1 mm, for example,
   should have nine solder paste deposits that are arranged in a
   three-by-three array."

We can also use analogies. E.g., the HVQFN48 package has a central
pad of 5.1 +/- 0.15 mm squared [2] and the recommended footprint
has a central pad of 5.1 mm squared [3]. The PCF50633's HVQFN68
has a non-square central pad of 4.8 +/- 0.15 mm times 4.3 +/- 0.15
mm, so the footprint's SLx/SLy should be 4.8 and 4.3 mm as well.
(You used the upper bounds 4.95 and 4.45 mm.)

Comparing [1] with [2] and [3], we get: b) 3^2/5.1^2 = 34.6%
(recommended 35%) and a) 9*0.75^2/5.1^2 = 19.5% (recommended 20%),
so NXP do indeed as they say.

Using c), since 4.8 and 4.3 mm are a bit smaller than 5.1 mm, we
should probably also have a 3 x 3 array - like HVQFN48 - instead
of a 4 x 4 array.

Your pattern has an overall size SPxTot/SPyTot of 3 mm squared.
9 mm^2 are 43.6% of 20.64 mm^2 (40% of 22.0275 mm^2), which is
a bit larger than the 35% recommended in b). 2.65 mm (34.0%) or
2.7 mm (35.3%) would be closer.

Your area for a) came up out perfectly, with 16*0.5^2/4.8/4.3 =
19.4%. Taking into account the change to 9 pads, we could use
0.65 mm (18.4 %) or 0.7 mm (21.4 %).

Hmm, I think I have to teach fped to "print" calculation results,
so that we can check such percentages without using "bc".

[1] http://www.nxp.com/documents/mounting_and_soldering/AN10365.pdf
[2] http://www.nxp.com/acrobat/packages/sot619-1_po.pdf
[3] http://www.nxp.com/documents/reflow_soldering/sot619-1_fr.pdf

> The figures I've used so far
> are just what seemed reasonable!

Not too shabby for dead reckoning :-)

Ah, and please keep the silk on copper issue in mind.

By the way, in the last few days, I made several usability
improvements in fped's GUI. E.g., there are now tooltips on
almost all items, and quirks such as false drags when clicking
(but not dragging) a draggable point are gone.

- Werner

More information about the gta02-core mailing list