At 09:36 AM 6/25/01 +0100, Andy Gulliver wrote:
>My experience of metric-specified parts on an imperial grid has led me to
>create the PCB symbols on a 1-mil grid, rounding the *absolute* co-ordinates
>of each pad to the nearest mil. (The use of absolute co-ordinates is
>critical to avoiding cumulative rounding errors, which can be significant on
>even medium-sized parts). The positional variation from this method is
>small, and the time taken to create the symbol is more than saved when it
>comes to tidying up the autorouter's efforts!
I'm not sure I understand why the use of absolute coordinates improves the
situation. Perhaps I should simply say I don't understand. I'm not even
sure what Mr. Gulliver means by "absolute coordinates" as it relates to
footprints.
The practice of making parts in a mil grid is one which I have followed.
The parts are made in the footprint library. The rounding must not be
truncation but must round to the nearest mil. If this is done, the error in
pad position will never be more than one-half mil, and therefore the pad
spacing error will not be more than one mil. I don't think it matters what
type of coordinates are being used. All coordinates in footprints are
relative anyway.
Perhaps this is what he means: if the gerber coordinates are not absolute,
they may be subject to additional rounding or truncation when the gerber is
generated, if the relative grid being used is not on a mil grid or if parts
have been rotated non-orthagonally. So one should either use absolute
coordinates in the photoplot setup (not a bad idea anyway, and avoid
centering on film which will make for a colossal nuisance if you or anyone
else ever wants to import the gerber), or plot to a tenth mil even if the
pads are ostensibly all on one-mil grid. A slight increase in plot size,
practically none if trailing zero suppression is used.
>I've used this technique on parts down to 0.4mm lead pitch on a 6/6 rule and
>it hasn't caused any problems.
Well, I'd expect it to lower yield a tad at that level. For 0.4mm pitch the
pads are only spaced 16 mils to start with, perhaps they are 10 mil pads
with 6 mils between. Since 0.4 mm is hard on 15.75 mils, one gets a full
mil shift in four spaces; either the space is reduced to 5 mils for that
pad pair, or the pads are 9 mils wide. Best case, spacing is reduced by 14%
when it is already pushing the process.
Yes, keeping pads on grid seems to help the present autorouter, and this
may be worthwhile. But hopefully we are saying bye-bye to this autorouter
in short order. For the long run, parts should be designed to their
specifications and not distorted into an English grid straightjacket.
Facing this, I think I'd rather make the parts accurately now and tolerate
needing to clean up the autorouter results.
(Perhaps we might be seeing an improvement in Protel's grid system next
release. The databases already accommodate mm dimensions so an improved
system, as has been suggested many times, would not create back-translation
problems. I have no special information, it is merely that the idea popped
into my head).
(I'll also mention the use of loop removal, which is an easily-overlooked
but invaluable tool for cleaning up router messes. Just draw the correctly
placed track and the garbage is automatically ripped up. *Much* faster than
deleting and replacing.)
(One more note: It is nearly unforgivable that the autorouter leaves messes
as it often does with off-grid pads, when the detection and removal of same
does not require anywhere near as much intelligence as does a good
autorouter, any any-angle routing when within a certain distance of a pad
would have been even better, but I assume that the effort that might have
been used to fix that problem has instead been poured into the new router.
[... tapping fingers on table....])
[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/subscrib.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]
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
