A recent post reported an experience of substantial gaps (estimated at 4
mils) appearing in short pattern produced using the virtual short
technique. It was theorized that this might be related to round-off error,
either with Protel or at photoplot. Another possible explanation was that
the fabricator had altered the data.
experimental conditions:
(dimensions in mils)
a footprint was created consisting of three pads
(1) square pad 70 by 70, location 0,0, designator 1, multilayer
(2) identical pad to above at location 100,0, designator 2, multilayer
(3) shorting pad 29.996 x 30 at location 50,0, designator null string,
bottom layer
This footprint was placed on a PCB in three rotations: 0, 45, 90 degrees.
Pins 1 were assigned net VCC, pins 2 were assigned nets TEST1, TEST2, and
TEST3 RESPECTIVELY. DRC errors appeared from on-line DRC. A footprint-scope
design rule allowing 0.001 mil clearance for that footprint was created,
and the DRC errors disappeared.
The board bottom layer was photoplotted using RS-274X automatically
generated apertures, 0.005 match criterion, and absolute origin, and with
three different precisions (2:3, 2:4, and 2:5) and the plot inspected with
CAMtastic and by import back to Protel.
Results:
The orthagonally rotated parts (0, 90 degrees) on all plots showed perfect
abutment of pads. The 29.996 x 30 pad was plotted with a 30x30 aperture. It
appears that the match criterion resulted in the assignment of a
rounded-off aperture, which is the desired result.
However, the 45 degree part was drawn with a 7.5 mil aperture, it was not
flashed. There were differences in the appearance of the plot as viewed
with CAMtastic and with Protel. Substantial "triangular" gaps (approx 1 mil
hypotenuse, the hypotenuse is actually an arc, the gaps are the kind that
show when space is filled with track and the tracks turn 90 degrees and
there is insufficient track overlap) appeared within in the drawn pads that
were not visible in the Protel imports. When the Protel imports were viewed
in draft mode, it appeared that the track overlap was sufficient to keep
gaps from appearing.
The algorithm used to draw the pads is inefficient, resulting in more than
twice as many tracks as necessary.
The clearance between the drawn pads, as expected, varied with precision:
from CAMtastic:
2:3 71 microinch
2:4 66 microinch
2:5 2 microinch.
The Protel imports produced the same measured clearances (using
Report/MeasurePrimitives), except that the 2:5 result measured as zero gap,
though a thin separating line was visible on screen.
The CAMtastic-viewed central pad appeared to be substantially distorted in
shape.
COMMENTS AND CONCLUSIONS:
Not all conditions were examined. It is possible, for example, that the use
of film centering, which is unfortunately the Protel plot default, would
result in additional roundoff error. This could cause a gap of up to 1 mil
to appear in even the orthagonal shorts.
This experiment used a gap of 0.002 mil, which is the recommended gap. The
design rule allowed a gap of 0.001 mil or greater. However, the problem
report mentioned a gap of 0.00005 inch, which is 0.05 mil or 50
microinches. If the match criterion was left at 0.005, this would create,
it is expected, a true 50 microinch gap on the idealized gerber. If that
was a typo and the gap was 5 microinches, there would still be a gap of
0.005 on the idealized gerber. It is expected that this gap would show in
industry standard gerber viewers.
(Idealized gerber means gerber as it would plot on an infinite resolution
photoplotter.)
A fabricator seeing such a gap may believe that it represents an
unintentionally small clearance; fabricators routinely enlarge such
clearances to make them fabricatable. It is thus a reasonable hypothesis
that the fabricator created the large gaps reported; since they appeared on
the final board, obviously they were fabricatable, yet they were reported
as being at the low end of fabricatability, which is consistent with this
hypothesis. The fabricator would not make them larger than necessary.
SUGGESTIONS:
Using virtual shorts, do not rotate them non-orthagonally. Keep the gap as
small as possible but not so small as to create a DRC violation with
allowed clearance set at 1 microinch. A 2 microinch gap satisfies this. Set
the match criterion no lower than 0.005 mil. There is no strong reason to
make it lower than the actual resolution of the plots, which is no better
than 0.1 mil for standard photoplotters, so one could set this criterion at
0.1 mil. Keep the virtual short pads on-grid in the native database units
(which are Imperial), so that no round-off will occur in photoplot.
If one must photoplot in metric, use the finest resolution available; but
even this is not guaranteed to completely remove the opening in the
idealized gerber.
If there is concern that a gap may appear to the fabricator, it would be
prudent to provide a note that the pads are to be shorted to each other, to
prevent a zealous fabricator from "improving" the film.
The use of aperture tables instead of RS-274X is a possibility, but this is
not recommended. Protel does not allow fixed aperture assignment, a
shortcoming that would bear review.
Protel should review the photoplot routines for accuracy and efficiency,
and the use of rotated flashes should be considered; CAMtastic's sub-mil
accuracy should likewise be reviewed. The subject PCB file with CAM setups
will be supplied on request.
With RS-274X, pads should not be drawn unless the 999 D-Code limit is reached.
An alternative to virtual shorts is also possible, and it is now more
practical than it was when the virtual short concept was developed, because
of the existence of plentiful mechanical layers and multiple customized
photoplot setups which can be simultaneously generated.
A two-pin component can be made with a shorting track on a mechanical layer
dedicated to this purpose, as part of the footprintg. The layer on which
the short is intended to exist should be removed from the main photoplot
setup file, and an individual layer setup created for it. In that setup the
mech layer mentioned should be set to be included in "all plots." Once the
setups are done, they need not be touched for future revisions and the
special additional layer will be routinely plotted.
However, this is not as simple a process as the original virtual short
component, which requires only a design rule and which will call attention
to itself if the design rule is omitted.
[EMAIL PROTECTED]
Abdulrahman Lomax
P.O. Box 690
El Verano, CA 95433
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To leave this list visit:
* http://www.techservinc.com/protelusers/leave.html
* - or email -
* mailto:[EMAIL PROTECTED]?body=leave%20proteledaforum
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
*
* Browse or Search previous postings:
* http://www.mail-archive.com/[email protected]
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
