Diva - Tolerance for DRC checks

Discussion in 'Cadence' started by Tim, Oct 5, 2006.

  1. Tim

    Tim Guest

    Hi,

    Is there a way to set a global tolerance for DRC checks in Diva or
    Assura?

    I have circular devices which are getting error flags because the
    spacing is say 9.9999um due to the circles instead of the 10um in the
    sep check.

    It would be nice if I didn't have to change the number for the spacing
    in every rule.

    Thanks in Advance,
    Tim
     
    Tim, Oct 5, 2006
    #1
  2. Tim

    vdvalk Guest

    Tim:

    Diva and Assura do what you ask them to. (no more, no less)
    So if you want a rule of 9.9999um spacing, then write it.


    On working with "circular" geometry, I do not understand how you work
    with circles.

    Typically, geometry gets converted to polygons. I think that Cadence's
    idea of this conversion
    is both wrong and broken. They like to use the mathematical approach to
    the ideal converson and
    then inscribe a polygon that is (at best) an internal poligonal tangent
    to a circle. (i.e. they take points
    from the ideal circle and they connect them to form a polygon. )
    This tends to drive the engineers nuts ... A 10.00000u circle converted
    to a polygon is always going to fail
    a 10u width check using this method.
    My proposed method ( which I have posted to this group and sent to
    Cadence >5 years ago ) is to create a
    polygon which is tangent to the circle (or larger) and then you do not
    get this problem.

    Since when is DRC ever nice? We all hope for DRC to be exact. We cannot
    afford it to be nice.

    -- Gerry
     
    vdvalk, Oct 5, 2006
    #2
  3. Diva has an automatic tolerance of one half database unit when either
    edge being considered is non-manhattan. This tolerance is disabled when
    the check involves equality, as in sep <= 10. There is no global
    tolerance setting to change available.

    It is possible that the rendering of the curve is the cause of your
    problem, not the tolerance used. Please read the discussion of this in
    the Diva Reference manual, "Layer Processing Concepts/Shapes/Conics and
    Paths". The jist of it is that curves are approximated as polygons and
    the vertices of those polygons are off the curve. Usually, the midpoint
    of each edge is on the curve.
     
    Edward Kalenda, Oct 6, 2006
    #3
  4. Tim

    Tim Guest

    Thanks for the responses!

    I solved my problem, it was the way VLE was creating the "circles".
    Once they are converted to polygons I saw the problem. I used 36 edges
    for the circles so it would have 1 flat edge every 10 degrees. The
    middle of a flat edge wasn't on the center grid line a vertex was which
    was slightly off grid. I rotated the "circle" 5 degrees which then
    lined up the flat edges on the 0, 90, 180, 270 grid lines. When I
    oversized 1 layer to create the next DRC no longer flagged the spacing
    as an error. The builtin tolerance then worked.

    Thanks Again,
    Tim
     
    Tim, Oct 6, 2006
    #4
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.