[SI-LIST] Re: Decoupling capacitors - mesh density

Raj-

Good comments. In our experience we've found that most
boards/planes we run into are rectangular is shape. As
such, we find that the # of meshes in the plane model
are primarily a bandwidth determining factor.

If you do have complex shaped planes or planes with
voids in them, then the # of meshes directly effect
how accurately the non-rectangular features are represented.
Round holes are approximated as closely as possible with
orthogonal meshes. If the mesh size relative to the hole
is small, then a fairly good approximation takes place,
otherwise round features will be "rectangularized". This
isn't necessarily bad, but is something that users should
be aware of.

On a related note, I wonder if anyone is aware of any studies
done on the implementation of non-constant meshes in plane models.
It can be computationally advantageous to only finely mesh the areas
of a plane where complex shapes occur and to let the majority
of the plane where nothing "interesting" is happening use a less
dense mesh. Problems occur in the circuit model where the dense
and not-so-dense meshes interface which lead to inaccuracies.
FEM solvers and the like routinely use adaptive meshing, but 
in those cases the meshing isn't related to actual circuit elements,
but of discretization of the problem space.

Of course with the excellent performance on large meshes reported
by Raj on his companies tools, perhaps the discussion of more
efficient meshing techniques is rather moot....

-Ray Anderson
Sun Microsystems

>
>
>Sogo, Ray,
>
>The limiting constraint for the number of meshes is usually the spatial
>resolution instead of the one-tenth wavelength criterion.  The planes are
>typically of complex shapes with multiple patches and voids. Moreover, there
>can be a large number of via connections between the planes.  In order to
>adequately represent the geometry, the vias connected between planes, and
>components (such as decaps) mounted on the board, it is often necessary to
>use a few hundred meshes to the side even for small boards or packages.
>
>The simulation times for HSPICE are unusually high because HSPICE was not
>designed to solve such circuits containing transmission line meshes.
>Therefore, special solvers are needed for such applications.  By way of
>comparison, using our special solver (PowerSI) we were able to solve a
>208x208 transmission line mesh plane model in about 14 seconds.  In 23
>minutes, we were  able to compute 100 frequency points. Loading time is
>negligible. The data was obtained on a Pentium III 866 PC with Windows 2000
>OS.
>
>Best Regards,
>
>
>Raj Raghuram
>Sigrity, Inc.
>"Achieve what others can't"
>raghu@xxxxxxxxxxx
>http://www.sigrity.com
>4675 Stevens Creek Blvd. , Ste 130
>Santa Clara, CA-95051
>PH: 408-260-9344 x116
>CELL: 408-390-7614
>FAX: 408-260-9342
>
>
>> -----Original Message-----
>> From: si-list-bounce@xxxxxxxxxxxxx
>> [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Ray Anderson
>> Sent: Tuesday, May 28, 2002 3:25 PM
>> To: si-list@xxxxxxxxxxxxx
>> Subject: [SI-LIST] Re: Decoupling capacitors
>>
>>
>>
>>
>> In an earlier message I wrote:
>>
>>   >I have written some simple C code that can generate  plane models
>>   >equivalent to the ones in sqpi for use with hspice with any arbritrary
>>   >meshing. Models with very dense meshing can be very expensive in
>>   >terms of compute time.
>>
>>
>> As an experiment I generated a 208x208 mesh plane model. Just
>> to read it into Hspice (2001.2) without even doing any simulation
>> at all took 1398 seconds ( ~23 minutes) on a 440 MHz Ultra-10
>> workstation. Had this model been used for a simulation it would
>> have been able to simulate a board 10ft 3in  square accurately to
>> 1 GHz (or a much smaller board to a much higher frequency).
>>
>> What's it prove? Not much except to point out that very large
>> models consume lots of computer resources, and that users need
>> to make intelligent choices on how detailed of a model makes
>> sense for any particular application......
>>
>>
>>
>> -Ray Anderson
>> Sun Microsystems Inc.
>>
>>
>> ------------------------------------------------------------------
>> To unsubscribe from si-list:
>> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>>
>> or to administer your membership from a web page, go to:
>> http://www.freelists.org/webpage/si-list
>>
>> For help:
>> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>
>> List archives are viewable at:
>>              http://www.freelists.org/archives/si-list
>> or at our remote archives:
>>              http://groups.yahoo.com/group/si-list/messages
>> Old (prior to June 6, 2001) list archives are viewable at:
>>              http://www.qsl.net/wb6tpu
>>
>>
>>
>
>------------------------------------------------------------------
>To unsubscribe from si-list:
>si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>
>or to administer your membership from a web page, go to:
>http://www.freelists.org/webpage/si-list
>
>For help:
>si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
>List archives are viewable at:     
>               http://www.freelists.org/archives/si-list
>or at our remote archives:
>               http://groups.yahoo.com/group/si-list/messages 
>Old (prior to June 6, 2001) list archives are viewable at:
>               http://www.qsl.net/wb6tpu
>  
>

------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
http://www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List archives are viewable at:     
                http://www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages 
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu

Other related posts: