Helloooo, TechNet!
My questions today are about bare-board netlist test. I have three
such questions, about the use and specifics of such test.
1. How many fabricators actually use supplied test data? We've
been supplying IPC netlist data as a standard part of our fab-
rication dataset for quite some time. Speaking from historic
results, many fabricators seem to not use it, preferring instead
to test to gerber-derived netlist data. Thankfully, designed-in
problems in our post-process are rare, but we thought we were buying
(and recommending that our customers should buy) locked-in comparison
to design intent. We've gone to great lengths to make an audit trail
from the engineering schematic through the CAD database to the IPC
file, and have had to do elaborate things to ensure that all nets
are represented correctly in said database.
Is this a case of it being "so much easier" to simply check that
the fabricator's shipped result matches the design "as received"?
(the GIGO theory) Why would this be so?
I'll answer this one myself, guessing that we must specify com-
parison to the design data in our purchase order or other such
controlling document. More on this in question #3.
2. We have been using a popular translator, from Allegro to IPC,
from CircuiTest. Over the past several months, we have noticed
repeated troubles with description of which side of a board the
element appears on. Perhaps this is a bellweather of actual util-
ization of the IPC data? Apparently, our output file describes
external layers as "0" and "1", and multiple vendors have been
interpreting in reverse fashion. Is this software-vendor specific,
ie a weakness in the spec? We've also had shops that are suspected
of actually using the data for a long time, without that confusion.
Is this subject to the specific pieces of software in play (ie fine
if using CircuiTest on both sides)? (I confess I have not read the
specification myself, shame on me!)
3. For obvious reasons we would like to extend our umbrella of
checking to the content of the IPC dataset. Therefore, we are looking
at doing gerber comparison in-house. We have looked at several
CAM type solutions, but find that our workload would cause us to
need at least 2 staffers to be full-time CAM operators. This CAM
function still appears to belong in the fabrication arena. Instead
we are deciding to upgrade some batch-type internal tools to handle
the netlist comparison. To this end, we would like to know of any
vendor that may have the gerber-to-IPC part of the translation avail-
able as an unbundled, batch type routine, whether it be an executable
on DOS/UNIX/NT or simply source code. Any takers or clues?
Thanks in advance, and in total for such a fine resource as TechNet.
Best Regards to all who got this far!
Jeff Seeger Applied CAD Knowledge Inc
Chief Technical Officer Tyngsboro, MA 01879
[log in to unmask] 508 649 9800
|