Magic Mailing List |
|
From: Jeff W. Sondeen (sondeen AT rcf-fs DOT usc.edu) Date: Tue Dec 04 2001 - 10:20:38 EST
Prakash writes: > Hi Tim, Jeff, > I was working on the extract section of a tech file > for a process that we were working on, this tech file is different > from other MOSIS type tech files in the sense that it does not have > different kinds of contacts for different diffusions like pdc, ndc, > psc, nsc etc but has a single tile type "contact" which is used for > all kinds of contacts, something similar to "generic contact" used > in MOSIS tech files. I found that with this type of contacts, > resistance extraction fails and extresis reports the error "MISSING > SD CONNECTION" and "MISSING GATE CONNECTION". Same is true if we > use a MOSIS Tech file and put gc instead of the specific contacts > and do the extraction. Subsequent ext2spice generates a SPICE > netlist with a lot of transistors missing. I examined the > file.res.ext file generated by the "extresis all" command and found > that it had some "killnode" statements for some nodes ( with the > intention of defining resistance at those nodes, I assume) but > resistances were not defined at those nodes in subsequent lines and > hence in the transistor definitions that followed, the old node > (i.e., the killed node) was used, hence ext2spice did not generate > transistors at the killed nodes. Is there a fix for this ? Any > suggestions ? Hi, here's what ftp://ftp.isi.edu/pub/sondeen/magic/new/beta/CHANGELOG says about the problem: ... *.tech27: Usage note on using generic contacts: Problem with the 'extres' command, as reported by a sharp user: > The problem as I observed is that, during 'extres', MAGIC > produces the following warnings [with generic contacts] > > :extres all > > Unreach transistor in ExtractSingleNet at 3 -4 > Unreach transistor in ExtractSingleNet at -5 -4 > Missing SD connection > Missing SD connection > Unreached transistor in O at 3 -4 > Unreached transistor in O at -5 -4 The reason for this problem is that the 'extres' command is implemented in magic code by using the magic symbolic contacts (those defined in the 'contact' section of the tech file). However, since generic contacts ('gc', as well as generic vias 'gv1', etc.) can't be in that section (since they "connect" types on non-adjacent planes), they confuse the 'extres' algorithm and generate the above errors. The only procedural workaround is to write then read cif/cal (letting the generics turn into magic symbolic contacts) and extract/extres with that. hope this helps, /jeff > > > Best Regards, > Prakash. > > > __________________________________________________ > Do You Yahoo!? > Buy the perfect holiday gifts at Yahoo! Shopping. > http://shopping.yahoo.com
|
|