Need a way of removing shape on a layer NOT of a specific dimension

Discussion in 'Cadence' started by som.pathak, Jul 25, 2006.

  1. som.pathak

    som.pathak Guest

    I am trying to find a way in DIVA/SKILL to remove a specific layer that
    is NOT of a given dimension. Can anyone help?


    thanks
     
    som.pathak, Jul 25, 2006
    #1
  2. som.pathak

    vdvalk Guest

    I assume that your data is essetially flat!

    One way of doing this is to split the task into two separate tasks.

    1) delete all data that is smaller
    2) delete all data that is larger.

    Here is one method I like to employ.

    ( it is not foolproof, but works on most cases!)

    Size the data by -1/2 of thet target data amount per edge
    Then size this resuly by +1/2 of the target data. This data represents
    areas where the data was too big.

    carefully process these areas to delete the areas that are too big.
    ( Note that several choices need to be made here )...

    With the resulting data, undersize by 1/2 of the target (less the
    smallest resolution the tool works with) and ovesize the resulting
    data. All of the sub-minimum data will be gone.

    This works for generic data, but has a few issues on boundary cases.
    Data that is
    almost right ( 1 dbu off of target) may get left behind. Also data that
    has spacings smaller than root2 of width can get merged in degenerate
    cases ...

    But this works for most data.

    i.e diva code could looks something like:
    bigdata = geomSize( geomSize( data, (-0.5*target) , (0.5*target )
    notbigdata = geomAndNot( data bigdata )
    notsmallnotbigdata = geomSize( geomSize( notbigdata,
    ((-0.5.target)+.01), ((0.5*target )-.01 ) ) )



    YMMV
    -- Gerry
     
    vdvalk, Jul 25, 2006
    #2
  3. som.pathak

    som.pathak Guest

    Gerry,

    Thnks... worked like a charm.... althought you were right that it would
    not process the boundary cases...

    Som
     
    som.pathak, Jul 26, 2006
    #3
Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.