kicad-developers team mailing list archive

["Lorenzo" <lomarcan@...>]

> File : /MO220VGGD8.emp 
> Uploaded by : lomarcan77 <lomarcan@...> 
> Description : Somewhat esoteric QFN package with 'special' pads 

OK this maybe would be useful on the users list, since it's a somewhat FAQ,but there's a problem too in it to discuss about.

To begin with this is not your everyday packages... first of all it's a 4x4mm QFN with 24 pins (i.e. a PITA so solder manually. Above these I only putBGAs); also it's a 'thermally enhanced' one: it has a big heat slug under it's body. AND it's a 0.15mm clearance part (maybe 0.20...), so even the thermal pad has to be shaped (it's also for a capacitive touch sensor so it'sreally sensitive to stray capacitances). Since it's from an ADI part, fortunately the datasheet contains a detailed drawing of the module.

The interesting stuff is in fact the termal pad itself. Since it has to be shaped it's built from 6 overlapping pads. Also, since it's so big it isn'tin the paste stencil (too much solder otherwise). The solder paste is defined by a bunch of pads, too. The FAQ is usually for solder mask, but the procedure is the same :D

The problem is: this big composite slug should be grounded, so it will havea netlist assigned. BUT only one of the pads has a designated pin name (the 'T' pin). During PCB ERC of course it gives a lot of 'pad near pad' error(since no netlist is not the same of having the 'T' netlist).

OTOH if I had declared all the compositing pads as 'T' pins (like it's usually done for the TO220 slug) I would need to do a crisscross of tracks since otherwise it would say 'pad not connected'. Not elegant at all, I would say.

What about a special treatment for no-pin-name pads in ERC? they could be excluded from 'some' checks but it's hard to define which ones... I can't think about a simple solution there.

Otherwise, the really cool and correct thing would be detecting that overlapping pads are actually connected (but it's not exactly trivial since we have round/oval/trapez pads). In this way I could assign to all these pin 'T'and it would infer that they're actually connected. More or less like padsautomatically attaches to zone fills.

Comments on how to handle these (not-so-rare) beasts?