[SI-LIST] Re: How accurate is HSPICE's field solver?
- From: "Michael Kurten" <michael.kurten@xxxxxxxxxxxxxxx>
- To: scott@xxxxxxxxxxxxx
- Date: Fri, 7 Feb 2003 18:07:47 -0800
Hi Scott,
what do you think about one aggressor and 4 victims (inverting your
approach).
If we add up the max. crosstalk amplitudes coupled to the 4 victim lines,
can this value be seen as a max worst case crosstalk estimation which would
happen to the aggressor due to the superposition theorem or is this approach
to simplistic?
Can this approach be used to quantify the max crosstalk between pins of a
connector, traces of a TL, ... in a reasonable manner?
Thanks
Michael Kurten
Fujitsu Siemens Computers@Serverworks
2451 Mission College Blvd.
Santa Clara, CA-95054
Tel.: (408) 922 1038
Fax.: (408) 922 3200
mailto:michael.kurten@xxxxxxxxxxxxxxx
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Scott McMorrow
Sent: Thursday, February 06, 2003 3:24 PM
To: Pat_Diao@xxxxxxxx; silist
Subject: [SI-LIST] Re: How accurate is HSPICE's field solver?
Pat,
No, I did mean 1 victim and 4 aggressors. With this configuration with
most simulators you can perform two types of simulations, crosstalk and ISI.
A crosstalk simualtion is performed with all 4 aggressors driven with a
pattern and the victim in a quiescent logic state, either active high,
active low, or both.
An ISI simulation is performed with the 4 aggressors and the victim
driven with patterns. Choosing different patterns to stimulate all
possible crosstalk and signalling modes is always an interesting exercise.
best regards,
scott
--
Scott McMorrow
Teraspeed Consulting Group LLC
2926 SE Yamhill St.
Portland, OR 97214
(503) 239-5536
http://www.teraspeed.com
Pat Diao wrote:
>Scott,
>
>Did you mean 1 aggressor and 4 victims below? It won't change the coupling
>coefficient either way, but physically in most simulators the aggressor is
>the one that carries the active voltage.
>
>Just want to clarify...
>
>Pat
>
>
>
>Pat Diao
>ASAT Inc.
>Fremont, CA
>phone: (510) 249-1227
>
>
>-----Original Message-----
>From: Scott McMorrow [mailto:scott@xxxxxxxxxxxxx]
>Sent: Thursday, February 06, 2003 12:48 PM
>To: mherndon@xxxxxxxxx
>Cc: si-list@xxxxxxxxxxxxx
>Subject: [SI-LIST] Re: How accurate is HSPICE's field solver?
>
>
>Matt,
>
>For single ended simulations I most always use 5 coupled lines, 1 victim
>and 4 aggressors. And for differential simulations I use 6 or 10
>coupled lines (3 or 5 differential pairs) There are several reasons why
>I do this:
>
>1) By extending the simulation trace geometry out to 2 aggressors on
>either side, I guarantee that the adjacent aggressors are operating in
>their normal electromagnetic field configuration, with their impedance
>altered correctly by the adjacent aggressors traces. This reduces some
>errors in the overall crosstalk and eye pattern simulations. Although
>neighbors that are 2 conductors away from the victim have very little
>direct influence upon the victim, they do have an indirect influence due
>to their coupling to the nearest crosstalk neighbor, having a tendency
>to alter the dynamic impedance of the nearest neighbors, and therefore
>the amount of energy available for crosstalk on the victim.
>
>2) Signals always travel through packages and often travel through
>connectors. In both cases, coupling is almost always much stronger than
>it is on the PCB. (Unless the PCB stackup is poorly designed.) It is
>often necessary to simulate many neighbors in packages and connectors.
> In order to keep the simulations symmetric and not induce artifacts due
>to different driven phases, I find it useful to extract as many
>conductors in the PCB as will be simulated through the package and
>connectors. For packages that are designed with poor power/ground
>structures, I find that there are additional modes of propagation
>between the signal conductors and package power conductors that cannot
>be accounted for if all of the signal conductors are not driven. This
>will sometimes require a large number of conductors to be extracted from
>the PCB, in order to include all of the system effects. Oftentimes I
>find that for non perfectly terminated busses the crosstalk will
>saturate in the package before ever reaching the PCB.
>
>3) At high frequencies fine pitch BGA via breakouts can introduce a
>large amount of crosstalk, similar to that caused by connectors. Since
>these patterns are square arrays, I will use a 3x3 array of vias,
>extract them using a full wave field solver, convert the s-parameters
>into a spice circuit with BroadBand Spice, and then assign the signals
>and grounds based upon the ball-out pattern of the BGA. This will often
>cause the simulation to need quite a few parallel lines to be extracted
>from the PCB.
>
>I hope this answers your question, Matt.
>
>Best regards,
>
>scott
>
>
>
>
------------------------------------------------------------------
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:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//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:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//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:
