[SI-LIST] Re: Long simulation diverges

Hassan -- I'm not a SigXplorer user, but I do have some experience in 
simulation algorithms and modeling.  These types of problems typically 
arise from bad data passed into a model, or from a bad algorithm (or 
algorithm implementation) in the simulator.  In either case, error 
builds up throughout the course of a transient simulation, and at some 
point takes over the results.  Fortunately, since the simulator is not 
bound by any energy constraints, the numbers usually get so huge that 
the error is very noticeable.

By running the same simulation with a lossless line, you have for the 
most part ruled out the possibility of a bug or errant parameter in one 
of your driver models.

If the transmission line model is parameterized (analytical equations 
for RLGC parameters), and your parameters are within acceptable bounds, 
then I would suspect the simulator and go to tech support.  Transmission 
line modeling and fast simulation in the presence of nonlinear devices 
is a subject of continuing research and development, and these types of 
problems are not terribly uncommon.

If the model is one in which you are supplying RLGC data points to the 
simulator, I would start by ensuring that the RLGC values you are 
supplying correspond to a physically realizable transmission line 
system.  The ability to supply RLGC data points to a transmission line 
model is useful, but can come back to bite if used inappropriately.  The 
associated math is fairly involved, so again, I would run them past tech 
support.  If the data is good, then it points to a bug in the simulator.

As far as workarounds, if there were a bug in the simulator, I would 
start by adding some loss -- either inline resistance or conductance to 
ground -- which can often damp nonphysical numerical oscillations.  If 
you're worried about introducing error with the loss, design a filter 
which only applies the loss at the frequency at which the explosive 
oscillations are occurring.  (You may, however, only end up finding 
another oscillating mode at another frequency which manifests itself 
later in the simulation record.)  These are just ad hoc 
grad-student-type tricks which certainly don't guarantee success in one 
case much less all.  Since, unlike graduate labs with donated software, 
you have actually paid for it, I would work through tech support.

Good luck,

   -- Steve

-------------------------------------------
Steven D. Corey, Ph.D.
Time Domain Analysis Systems, Inc.
"The Interconnect Modeling Company."
http://www.tdasystems.com

email: steve@xxxxxxxxxxxxxx
phone: (503) 246-2272
fax:   (503) 246-2282
-------------------------------------------





Hassan O. Ali wrote:

> 
> I'm using SigXplorer 14.1 to perform a very long simulation: 25,000 cycles 
> with a 2.5GHz 
> periodic (101010...) stimulus (period = 400ps). I have observed that the 
> solution 
> diverges badly (voltages in the order of kilovolts!) after the first 200ns 
> (500 cycles). 
> 
> The circuit is very simple: just an ideal 3.3Vpp voltage source with 0-100% 
> rise/fall 
> time of 10ps connected to a 50ohm load through a 2-inch 50-ohm microstrip 
> line. The 
> microstrip line, however, is specified to be frequency-dependent for up to 
> 18GHz. The 
> same simulation runs fine if I use an ideal 50ohm transmission line (TLine).
> 
> Has anyone encountered the same problem and found a solution for it?
> 
> Regards.
> 
> Hassan.
> ------------------------------------------------------------------
> 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: