[SI-LIST] FWD: Re: Re: HFSS solution issue

Manually forwarded due to list malfunction

------------- Begin Forwarded Message -------------

Date: Fri, 24 Jan 2003 09:55:10 -0800
From: bpanos <bill.panos@xxxxxxxxxxxxxxx>
To: achandra@xxxxxxx
Cc: Dan_Swanson@xxxxxxxxxxxx, mike_bihan@xxxxxxxxxxxx,
        si-list@xxxxxxxxxxxxx
Subject: Re: [SI-LIST] Re: HFSS solution issue


Arun-
We have the same issue regarding disk swapping and we have 4 GB of RAM, on a Sun
Blade system. We've set the appropriate setting in the config file, but still
there is a lot swapping across the hard drive.



Chandrasekhar Arun wrote:

> Hello Dan and all HFSS users,
>         Excellent comments from Dan and it really summarises what every HFSS 
> new user needs to know reg. solving his/her model. A few points from my 
> experience
> 1. My primary application with HFSS is solving chip-package 
> interconnects and I use a delta-S of 0.005. This number is not 
> sacrosanct, but when coverted to dB is -0.043 dB for insertion loss and 
> -46 dB for return loss
> IL=3D20log10(1-0.005) and RL=3D20log10(0.005)
> The key is finding an optimum between solution accuracy and time. I 
> expect my structures to have an insertion loss of about 1 dB at 5 GHz 
> and it is Ok for me if my HFSS results are 1.043 or 0.957 dB. However if 
> you are trying to extract the S-parameters of a low-loss short line, 
> maybe this error is too high. However, already to achieve 0.005 eroor I 
> have to live with about 25000 tetrahedra and this takes a few hours to 
> simulate for 20 points and there is no way I can think of delta-S lower 
> than 0.005!
> 2. If you are usinh HFSS at low frequencies, remembet to Solve Inside 
> Metals !
> 3. To compare your results with VNA measurements always de-embed and 
> re-normalise with 50 ohms
>
> one question
> does anyone know how you can manipulate the amount of memory used by 
> HFSS while solving in the RAM and the disk? I have 1GB of RAM, still 
> HFSS always swaps some data with the disk. I remember that the deafult 
> is 70% RAm and 30% dis, but I do not see this always!
> Regards
>  
> Arun Chandrasekhar
> PhD Student: RF Packaging     
> IMEC | MCP/MaRS Group =09
> Phone:  +32 16 28 82 28 =09
> Fax:    +32 16 28 10 97 =09
> Email:  achandra@xxxxxxx <mailto:achandra@xxxxxxx>  
> Seeds for Tomorrow's World <http://www.imec.be/>  
>  
> IMEC, Kapeldreef 75, 3001 Leuven, Belgium   |  Phone:+32 16 28 12 11  =
> Fax:+32 16 22 94 00  |   Website: www.imec.be <http://www.imec.be/>=20
>
> This e-mail and/or its attachments may contain confidential information. =
>  It is intended solely for the intended addressee(s).
> Any use of the information contained herein by other persons is =
> prohibited.  IMEC vzw does not accept any liability for the contents of =
> this e-mail and/or its attachments.=20
>
> -----Original Message-----
> From: Swanson, Dan [mailto:Dan_Swanson@xxxxxxxxxxxx]
> Sent: Thursday, January 23, 2003 17:23
> To: 'mike_bihan@xxxxxxxxxxxx'; si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: HFSS solution issue
>
> Bi, Han
>
> I think you are a little confused on how the solution setup goes.
>
> Adaptive Single Frequency can be run by itself or it
> is the starting point for a Frequency Sweep.
>
> So you adapt the mesh at some frequency, typically the highest
> freq you are interested in. Then you start the frequency sweep
> using that adapted mesh. So the mesh is "overkill" for the low
> freqs and just right for the high freqs. Knowing this, if you
> are trying to cover many octaves in frequency, then maybe
> several runs using different adapt freqs is more efficient.
>
> So in the setup menu you select both the Single Freq Adapt
> and the Freq Sweep you want at the same time. The software then
> does both.
>
> If you run a Discreet Sweep, results should be fine. A Fast Sweep
> tries to derive the result from the minimum number of discreet=3D20
> freq solutions. Watch out for the both ends of the sweep range
> and always check the results at least once with a discreet sweep.
> If you ask for a very wide Fast Sweep range, the software often
> suggests that you break it up into smaller segments.
>
> As to number of adapts, that is very tricky. Basically you need
> to identify the key parameter in your analysis and figure out
> what it takes to get there. For example, I do a lot of resonators.
> So I look at convergence of resonant frequency as function of meshing
> parameters. Remember this is very general purpose software, you always
> know more about you particular problem than the developer or the =3D
> software.
> You have to use that knowledge to guide the software to the right
> solution.
>
> Using the default settings, the starting mesh in HFSS is very coarse.
> I typically set LAMBDA_REFINE_TARGET (system variable) to .02 or .033
> Refinement percentage to 40% (default is 20%).  And maybe run 5 to 10
> adapts. I set Delta-S limit to 1e-6, Delta-S achieved is very problem
> dependent and does not always imply absolute convergence. Of course
> you are looking for a convergence curve that is a nice smooth =3D
> parabolla.
>
> The geometrical resolution in your model will also affect convergence.
> I tend to draw cylinders, etc kind of coarse, 8 or 12 sides to a =3D
> cylinder.
> I do this to speed things up. But sometimes if you make things too =3D
> coarse
> the convergence curve will oscillate, or you may converge to the wrong
> final value. For example, look at impedance of a coax line as a =3D
> function
> of how many segments you use to define the cylinders.
>
> All this is in my new book, coming out in May 2003:
>
> http://www.artechhouse.com/default.asp?Frame=3D3DBook.asp&Book=3D3D1-5805=
> 3-3=3D
> 08-6&Cou
> ntry=3D3DUS&Continent=3D3DNO&State=3D3DMA
>
> Best regards,
>
> Dan
>
> Dan Swanson                     EMAIL:  d.swanson@xxxxxxxx
> Forem USA                               PHONE:          978-834-4085
> 37 South Hunt Road              FAX:    978-388-7077
> Amesbury, MA  01913
>
> > -----Original Message-----
> > From: Bi Han [mailto:mike_bihan@xxxxxxxxxxxx]
> > Sent: Thursday, January 23, 2003 10:52 AM
> > To: List` Si
> > Subject: [SI-LIST] HFSS solution issue

> > hi, ansoft guies:
> >
> > There is one simple question about HFSS.
> >
> > Since Adaptive 'Single frequency' analysis will allow users
> > to setup 'Requested Passes', what number will be suitable? I used 7.
> >
> > However, issues come up!
> > The results,obtained by 'Single Frequency', differs greatly
> > from 'Sweep' analysis
> >
> > results? Because of the higher accuracy of 'Single frequency'
> > analysis, I prefer it than sweep one. (Actually,sweep results
> > is the Pass 1 results on Single Frequency analysis.)
> >
> > Then, what value does 'sweep' analysis have, with so poor
> > accuracy, whose mesh will not be refined. In another way, in
> > which way should we treat the data from 'sweep analysis'?
> >
> > Sincerely,
> > Bi, Han
> > Sjtu, EMLAB
>
------------- End Forwarded Message -------------

------------------------------------------------------------------
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: