Mail Archives: geda-user/2015/07/28/20:52:39
Erich Heinzle wrote:
> One of the problems with your Gerbers is that PADS generated them with a 27
> inch or so translation in the upwards direction (-ve y direction in PCB),
> which seems to confuse gerbv.
Thank you for your perseverance with those gerbers! And they aren't
mine... They were made 7 years ago by a company that no longer exists
- a once-noble effort called Openmoko. The only connection I have to
these gerbers is that I was the one who convinced Sean Moss-Pultz, the
founder of that bygone company, to release them publicly - they were
proprietary prior to April of this year.
> I have had some success converting your SMD top and bottom layer gerbers to
> a PCB footprint format,
How do you determine where one footprint (element) ends and another
fp/element begins? I mean, even if one could recover the complete set
of pads and nothing but pads (e.g., if the pads are flashed like they
are supposed to be while everything else is drawn differently), how do
you divide and group this total set of pads from one side of the board
into elements, as in these belong to U201, those belong to U202 and so
forth?
As for this specific GTA02 board layout, the only part of value to my
project is the 40x30 mm block at the bottom of the bottom layer
(excuse the pun), the one enclosed in the metal shieldcan footprint
with reference designator SH1. This section is the only part I seek
to reuse from Openmoko's design; the rest of their board design is to
be chucked and replaced with a few connectors providing access to the
extracted modem section.
Put another way, none of the components on the side of the board which
Openmoko called the top are applicable to my work, only the modem
block on the side they called the bottom.
> and am currently refining square pad detection for
> through hole pads.
There are no through holes on Openmoko's board except a few mounting
holes, and certainly no TH components.
> I have not played with the other layers yet as your
> priority seemed to be workable footprint geometries - and I think I am just
> about there.
I am torn between two possible approaches: triband or quadband. I
would like to build a quadband modem, but I don't have a reference PCB
layout for the quadband version, only schematics and a PDF component
placement chart. For the slightly less capable triband version, we
have these PADS and gerber files from Openmoko.
As far as I could tell, simply plopping the quadband RF front end
(Epcos M034F chip module) in the place of Openmoko's triband RFFE
without redoing the whole layout in a different way won't work - the
pin locations for critical GHz RF signal paths are too different
between the two RFFE designs. So I got this dilemma:
Option 1: Openmoko's triband PCB layout is known-working, so it's very
tempting to copy it verbatim for my first prototype. But because that
verbatim copy would have to involve a translation of the complete
layout (of the section of interest) from PADS to PCB, it looks like it
will take a gargantuan effort. All that effort would subsequently be
wasted when we redo the whole layout to gain quadband capability.
Option 2: We can choose to not bother with Openmoko's layout at all
except for the footprints, and go straight to quadband. But that
quadband design is totally unvetted (just because I found a schematic
on an obscure Chinese site that was drawn >10 y ago by a now-disbanded
division of TI doesn't mean that the circuit depicted therein actually
worked), and the prospect of having to redo the layout completely from
scratch, with nothing but a PDF component placement chart to guide me,
is very scary - I am not exactly a GHz RF PCB design expert...
So that is my dilemma.
SF
- Raw text -