Mailing list archives : pcb-rnd

ID:5906
From:rn...@igor2.repo.hu
Date:Sun, 15 Jan 2023 19:02:25 +0100 (CET)
Subject:Re: [pcb-rnd] wishlist - Automatic Nudge of Refdes
in-reply-to:5905 from Gabriel Paubert <pa...@iram.es>
  This message is in MIME format.  The first part should be readable text,
  while the remaining parts are likely unreadable without MIME-aware tools.
 
--722076672-1857507435-1673805509=:25839
Content-Type: text/plain; CHARSET=UTF-8
Content-Transfer-Encoding: 8BIT
Content-ID: <alpine.DEB.2.20.2301151858321.25839@igor2priv>
 
Hello Gabriel,
 
On Sun, 15 Jan 2023, Gabriel Paubert wrote:
 
>	Hi Igor2 and Evan,
>
>Actually in my own library, for most passives (1206 and smaller), I have
>the refdes on the assembly layer, since this is what I prefer and need
>most of the time. Refdeses on the silkscreen take up far too much space
>on the PCB when you have lots and lots of decoupling capacitors, which
>is frequent for RF and for high speed logic (FPGAs, etc).
 
I absolutely support that!  I was considering writing about this in my 
original reply, but I thought it was long enough already... But now that 
you brought it up...
 
Our stock lib is low quality, when measured with today's requirements. 
It's mostly inherited footprints from geda/pcb's stock lib (heavily cut 
down), plus the parametrics I designed before the data model switchover, 
to look similar in style to those old static file footprints. It was nice 
some 15 years ago, but it's really just 90s stuff.
 
All these static file footprints were drawn back when geda/pcb didn't 
support drawing on anything else but silk and copper (the outline layer 
was really a copper layer) and when elements restricted you to have only 
lines and arcs on silk and one hardwired refdes, always on silk.
 
pcb-rnd has evolved a lot, especially on the data model side. New, modern 
footprints are very different:
 
- there should be less silk, like the huge frame around so() and dip() are 
a big waste for dense boards
 
- in parallel there should be much more mechanical drawings on the assy 
layers, many lines and arcs that really shows how the part would look 
like, helps a lot in overseeing the placement/orientation, doesn't pollute 
the silk layer
 
- you can have multiple text objects in a subc, including multiple refdes 
floaters, and you can easily move them between layers, and on screen we 
have the red subcircuit virtual layer which shows, on-screen-only the 
refdes, so you don't even need to put it on any layer if you only want to 
see it while editing
 
- plus a nice polygon (normally a rectangle) on the keepout layer
 
So yes, another solution, in case you routinely want your silk layer free 
of refdes text, is to use a better, custom footprint lib that has the 
refdes text floater on an assy layer. Then when you need refdes on silk, 
you can move it.
 
(I personally like to have it on silk whenever there's enough room, it 
helps debugging boards)
 
>
>Case in point, look at the layout on an LTC6950 (unfortunately just
>obsoleted by Analog): you need 8 decoupling capacitors on one side of
>the chip along with 4 differential pairs. Don't try with 0402, only 0201
>fit, and even then I had to use a 50µm grid!
 
Nice example, this kind of dense designs really benefits from modern 
footprints.
 
>Igor2, I have found a few bugs in pcb-rnd, but I am now very short on
>time to distill test cases.
 
No hurry. If you have any of them blocking or extremely annoying, report 
those, else keep them for later when you have the time.
 
Best regards,
 
Igor2
--722076672-1857507435-1673805509=:25839--
 

Reply subtree:
5906 Re: [pcb-rnd] wishlist - Automatic Nudge of Refdes from rn...@igor2.repo.hu