MAGIC Magic Mailing List
 
 

From: cfk (cfk AT pacbell DOT net)
Date: Fri Apr 18 2003 - 15:40:44 EDT

  • Next message: cfk: "Re: lambda & SCN6M_SUBM.10.tech27"

    I guess my naivete is showing a bit. Today, I am trying to understand
    geometry, techfiles and transistors a bit more. So, I took current.tar.gz
    from MOSIS and got the SCN6M_SUBM.10.tech27 for the TSMC18 process (which I
    believe to be 0.18um) into my /home/cad/lib/magic/sys directory. Now, I
    invoke "magic -w -TSCN6M_SUBM.10". I draw a pfet whose pdiff is 9 grid
    squares wide by 6 grid squares between the metal1 contacts of the drain and
    source. In the middle, I paint a 2 grid square wide poly (overhanging the
    pdiff by 2 grid squares). When I extract this transistor, the .spice file
    from magic's extractor says the transistor is 2.2u wide by 1.8u long.
    
    I also see in the SCN6M_SUBM.10.tech27 file that statements:
    
    style lambda=0.10(p)
     scalefactor 10 5
    
    So, I am thinking that each grid square is .05u (0.10 * 5 / 10), which would
    make a pfet 9 squares wide as .45u (.05 * 9 = .45), but the .spice file says
    the transistor is 2.2u wide which is almost, but not exactly, 5 times
    bigger.
    
    Similarly, the poly is 2 grid squares long over the pdiff (in the direction
    of current flow), so I would think that it is 0.10u long, yet extract says
    it is 1.8u long.
    
    
    OK, maybe this really means a grid square is one half of the smallest TSMC
    dimension, which for the TSMC18 process I think is 0.18u. So, a pfet 9 *
    0.09 wide is 0.45u and 2 * 0.09 long is 0.18u, and that seems to fit with my
    limited understanding of magic, tech files and the like, yet the extracted
    spice file says the transistor is 2.2u wide and 1.8u long.
    
    Obviously, I dont get lambda, scalefactors & tech files yet and would
    appreciate someone helping me understand a bit more.
    
    Charles
    


  •  
     
    Questions? Contact Rajit Manohar
    cornell logo