At 02:47 PM 10/1/2010, you wrote:
Rick Collins wrote:
Where I want to get us, is being a consistent customer, for whom they
no longer need to think about step b).
From what I can tell, they don't bother with the two steps. The
machine picks the part from the feeder and before placing it, the
operator verifies it is oriented correctly. Done once for a given
feeder and a given side of your board, the rest of the parts from
that feeder should be good.
If for whatever reason the designer used 2 different footprints for
the same part occuring
several times on a board, if the footprints are position/rotation
inconsisten...
I have no idea why anyone would do that. Sure the designer can
totally screw up a design. What is your point?
This is 100% reliable and not really a lot of time on their part.
Even if they do the steps you are talking about, they will do the
step I have outlined. They aren't engineers and they don't think
like engineers. They don't want to figure out what things don't
work, they just want to make them work. Their way is much easier
in the long run I am sure.
The guy talking to me has 'MSc.' before his name ;-) No idea how
much he's involved with the
actual operation of the machines.
Even austrian farmers try to figure out and avoid reasons for
problems - no cowboy mentality?
I don't care much about Austrian farmers. I am talking about what
board assembly houses do. I am sure they are capable of working with
specs and heading problems off at the pass. But they seem to know
what works best for them and assuming that everyone from the three
different camps are on the same page is not something that works for
them I gather.
See above / please check yourself.
I don't have PCB, so I can't check.
in that case you have to believe me or others, that with the
internal coordinate system
of 'pcb' X+ is to the right and Y+ is down.
Why do I care what any PCB package uses internally? If they don't
have positive Y up and positive X down at the user interface, how the
heck am I supposed to coordinate that with Gerber files and the rest
of the world? Why not use polar coordinates? Are you talking about
something truly internal or at the user interface?
The assembly house I'm talking to, offers to provide standard parts.
I imagine,
they use a combination of machine vision and having resolved step a) from
above "once and forever" with their part suppliers.
When you say "parts", do you mean footprint data for the CAD packages?
No, I mean this assembly house holds some standard parts in store
(0603, 0805 resitors, caps,...
of common values) and will sell them to you if you like - can save
you and them
some hassle.
I use full turnkey. They buy all the parts I need. So far it has
been TONS easier and I am still making good profits.
The bottom line is, ask your assembler what they want. Don't
assume anything.
I will.
[snip]
About the .xy-file I'll have to read, how the footprint coordinates
and placement in the board influence the actual values. I think it
will be a bit tricky to check the footprints, since pcb doesn't show
the true coordinates but computes an offset on the fly to make all
screen coordinates positive - this is a bad idea for working on .fp-files.
That doesn't make a lot of sense to me, but I'm not sure why it is bad.
To check, whether all the footprints I use conform to IPC-7351(B), esp.
if the centroid is at (0, 0) of footprint it would be easiest, to
just load them
into the design program. But pcb is cheating on you: the footprint-definition
describes say a 2-pad part with pad-centers at (-2.0mm, 0mm), (2.0mm, 0mm)
and centroid at (0mm, 0mm). When loading the footprint definition
(that's the .fp-file)
on it's own, pcb will do some guesswork to squeeze everything in it's
positive coordinate quadrant and compute an offset (failing occasionally btw.
leaving parts of text and lines in nirvana).
Then it will tell you that above pad-centers are at (0.7mm, 1.5mm),
(4.7mm, 1.5mm)
and center mark at (2.7mm, 1.5mm). The same applies if the definition had been
(2000, -100), (2004, -100) and (2002, -100) - there's no way to tell
the numbers
in the definition by looking at the GUI.
I don't care much how a CAD program works internally. But it has to
use the centroid of the part for indicating the coordinates of the
part location or it is not compatible with the assembly system. My
assembly house didn't care much about the rotations since that is
easily verifiable and fixed. But if the coordinates are offset from
the centroid they claim to have a hard time finding and fixing
that. The only coordinates that matter are what the designer sees
and what shows up in the XYRS file. I don't care at all how the CAD
system arrives at the values.
And what I'm trying to figure out atm, to verify the data to be sent to the
assembly house is, how the footprint definition, the guess work and the
actual placement get munched into the XYRS file.
If you are looking at anything other than the Gerber files and the
XYRS file, you are digging too deep. Just pull up the top and bottom
Gerber files in a viewer and measure to the center of a few parts, or
all the parts. This is not something I would ever expect to need to
check. If a CAD program can't generate a proper XYRS file, then why
use it? I almost abandoned FreePCB because the XYRS file is
generated by an external program that is cumbersome to use. I plan
to spend some time trying to fix this shortcoming and if that doesn't
work, I will be working with something else on the next board I design.
I use 0,0 as the lower left corner of the board and my fab drawing
gives coordinates of the fiducial marks on the board along with
major drill holes (like mounting points). So all coordinates on
the board are positive. Why would you want it different? I don't
know what a .fp file is.
I don't want it different for the board (which would require
reversing Y-orientation in pcb),
but for a loaded footprint definition.
You've lost me on this one. At the user interface, if the
coordinates indicated don't match what is output to the Gerber file
and the XYRS file it would be very hard to work with.
BTW, all part coordinates should be wrt the centroid of the part,
not pin 1. Some CAD packages used to use pin 1, but it is standard
practice to use the package centroid now.
Centroid always appeared natural to me - it's the best position for
physical rotation axis as well .
What sort of checking of the footprints do you want to do? You
should use a Gerber viewer to verify the Gerber files. Nothing
inside the CAD system matters if the Gerber files aren't
right. What would be great is a viewer that understands the part
shapes and positions the parts according to the XYRS file on top of
the Gerber file images so you can verify alignment and orientation.
Since the assembly house won't have/use my footprint definitions and
I don't want
to make a drawing of each and every part, if a standard clearly
states, how it looks,
I have to check, whether the CAD-internal definitions conform to the standard.
If no standard existed at all and I really have to make drawings
with whatever tool,
I still have to confirm the CAD-internal definition is identical to
the drawings.
I don't get what you are doing. Why would you provide a footprint
definition to the assembly house? They need the Gerber files for the
boards with fabrication drawing (or the boards themselves) and the
XYRS file along with an assembly drawing and BOM. They have no use
for a footprint spec. It can be useful to specify the package in the
BOM/XYRS file, but the footprint spec gives them no useful info. At
most they could use the data sheets, but I've never been asked for one.
Rick
_______________________________________________
geda-user mailing list
geda-user@moria.seul.org
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user