Creating circle in Virtuoso

 Hi,

Unless I am mistaken, it is not mathematically possible to snap a circle so that it is all on grid, but this might just be the terminology that you are using.

After creating a circle you can use Edit -> Advanced-> Convert to Polygon to manually control the conversion of ellipses, donuts and circles to polygonal shapes.  There is a setting "Conic Sides" on the Layout Editor Options form which defaults to 20, but you can try other values. This is also available as a .cdsenv variable setting, for example:

 I hope that this information helps you.

Regards,

Lawrence.

 You will have to calculate the coordinates to get them to snap to the manufacturing grid. Calculate your angle based on the number of segments and that will give you the angle to calcuate each segment. Digging out your old trig book, sine of the angle is used to calculate the X value, cosine the Y value (the radius is the hypotenuse). Round the values to the manufacturing grid as you calculate them and you will get a slightly jaggy circle with everything on grid. You can do some smoothing after you calculate the points, adjusting the length of each segment independently.

Ted

Hi,

I encountered the same problem and wrote a pcell to do it for me: http://atchoo.org/tools/cadence/circlePcell.il

Let me know if you have any problems with it!

 Cheers,

Roger

See also Sourcelink Solution 1800103:

http://sourcelink.cadence.com/docs/db/kdb/1994/Sept/1800103.html

Thanks. This one seem to work. However, I do not understand why a compensate factor need to be added. This factor is get from        

compFactor = 1 / rep~>DBUPerUU

 and used as follows:

      procedure(  CCSroundCoord(  point xgrid ygrid compFactor )
          prog(  ( x y )
               x=round( ( xCoord( point ) + compFactor ) / xgrid ) * xgrid
               y=round( ( yCoord( point ) + compFactor ) / ygrid ) * ygrid
               return( list(  x y  ) )
           )
       )

DBUPerUU represents number of database units in one user unit. In my system it is 1000. If I add this one, the whole calculation will be screwed up.

This is to compensate for small rounding errors that can occur whenever doing floating point arithmetic.

For example, if you do:

floor(1.005/0.005) you end up with 200, not 201, due to the fact that the result of the division is a tiny amount less than 201.

Adding a small "epsilon" compensates for this - in this case, it's adding 1 DBU. 

However, since the calculation in this function is using round(), there should really be no need to do this, because it will round to the nearest, rather than round down or up. So I think it's redundant. It shouldn't really break anything (assuming the grid is not 1 DBU!), but it doesn't do anything useful.

Regards,

Andrew.

 Hi Roger - This is a very cool pCell, nicely done!!  

 Thanks

 Rick