|
|
Magic Mailing List |
|
From: Jeff W. Sondeen (sondeen AT rcf-fs DOT usc.edu) Date: Fri Nov 30 2001 - 16:43:33 EST
Carl Grace writes: > > Hi Jeff, > > Attached are: > > mimtest.mag : magic file with m5 bottom plate, mim (CAP_TOP_METAL) top > plate and m6 connecting to mim through a 12 lambda * 12 lambda micontact. > This file causes no design rule errors, since the minimum micontact is > 10*10 lambda. > > mimtest.cif : cif output from mimtest.mag, generated by typing :cif at the > magic command line. When read back into magic using :cif read mimtest, > the micontact is only 5*5 lambda, which is a design rule error. > > Carl thanks, Carl, yes there's an error in the techfile for the case of magic cif'ing in CV5's over CAP_TOP_METAL (so-called mic's) -- will fix it asap -- should make release 2001b this monday. however, there's a tougher problem (described next) that makes me think you may want try the DEEP rules instead of SUBM (and that's why i'm including the magicdev email list in this reply -- they may know of a new way to handle off-grid points that i'm not familiar with yet -- i've included a test case below in case they have time to try it -- or can explain the trick to me). anyway, the problem is that to line up the stacked contacts in SUBM rules, i had to let people draw the gv5's 4x4 lambda, so they can stack nicely over 2x2 gv4's, but they get output in the cif as 3x3 lambda (which the need to be for Mosis). however, centering these 3x3 gv5's over the 2x2 gv2's means that they're really off the lambda grid (which is no problem for Mosis) but is a problem for magic on cif read in, because magic must snap to grid and often gv5's can get spaced too close by 1 lambda. (notice that gvi's are really just gv5's (both become CV5 in the cif) so the same snapping applies to gvi's and mic's). in the meantime, here's a slight change to the SUBM techfile that can help reduce this snapping (just basically adding extra "spacing" in the 'squares' statement to overspace CV5's to help magic out a bit), however, rather than use this, can you try using SCN6M_DEEP.09 and see if it helps ? (CV5's are 4x4 in DEEP rules) it may help your other stacked contact issues (then again, it may be worse so don't delete all you've done). thanks, /jeff mimtest.mag: (revised) for use with SCN6M_SUBM.10.light.tech27 (version 2001a) magic tech scmos timestamp 1007154441 << metal4 >> rect -8 -52 6 -38 << m5contact >> rect 11 -52 25 -38 rect 29 -52 42 -39 rect 46 -52 58 -40 rect 62 -52 73 -41 rect 77 -52 87 -42 rect 90 -52 99 -43 rect 102 -52 110 -44 << metal5 >> rect -39 -18 55 39 rect -8 -35 6 -21 rect -8 -52 6 -38 << gv4 >> rect -7 -41 -5 -39 rect -2 -41 0 -39 rect 3 -41 5 -39 rect -7 -46 -5 -44 rect -2 -46 0 -44 rect 3 -46 5 -44 rect -7 -51 -5 -49 rect -2 -51 0 -49 rect 3 -51 5 -49 << m6contact >> rect 11 -35 25 -21 rect 30 -35 43 -22 rect 48 -35 60 -23 rect 65 -35 76 -24 << metal6 >> rect -42 -2 21 28 rect -8 -35 6 -21 << gv5 >> rect -7 -26 -3 -22 rect 1 -26 5 -22 rect -7 -34 -3 -30 rect 1 -34 5 -30 << micontact >> rect 21 1 51 30 << metali >> rect -27 1 21 30 rect -27 -10 51 1 << gvi >> rect -24 23 -20 27 rect 0 23 4 27 rect -24 -1 -20 3 rect 0 -1 4 3 << labels >> rlabel metal5 -37 -17 -37 -17 3 m5 rlabel metali -25 -8 -25 -8 1 mi << end >>
|
|
|
|