[SI-LIST] Re: [EXTERNAL]Re: Step Size for s-parameter extraction

  • From: Charles Grasso <charles.grasso@xxxxxxxx>
  • To: istvan.novak@xxxxxxxxxxx
  • Date: Thu, 10 Feb 2022 09:14:07 -0700

You guys are awesome!! Thank you so much!!
On Thu, Feb 10, 2022 at 9:04 AM Istvan Novak <dmarc-noreply@xxxxxxxxxxxxx>
wrote:

 This message originated outside of DISH and was sent by:
si-list-bounce@xxxxxxxxxxxxx

Hi Charles,

Scott did this gEEk spEEk on 2/18/2021 and you can find the recording at
https://blog.samtec.com/geekspeek/#ondemand
<https://blog.samtec.com/geekspeek/#ondemand>

Regards,

Istvan Novak

Samtec



On 2/10/2022 10:51 AM, Charles Grasso wrote:
Hello Scott - I am interested in viewing your "Bend EM tools" webinar!
Where can I find it?
Chas

On Thu, Feb 10, 2022 at 8:39 AM Hakim <hakim@xxxxxxxxxxx> wrote:

This message originated outside of DISH and was sent by:
si-list-bounce@xxxxxxxxxxxxx

Hi Scott,
Thanks for the reply.

I currently have just the P/N of the diff pair and adjacent ground
balls.
But my trace lengths on the package and pcb could be smaller and the
plane
cutout is big which I'll reduce.

Have been using rad boundary and waveports, also learnt this from your
bend
EM tools webinar ;-). Thanks again!

I've got a HPC license, but I'm limited to 12-cores by that license,
whereas I have access to a +100 core server.

When you mentioned use one fixed frequency at nyquist, this is the
adaptive
mesh settings you're referring to? My Nyquist is about 13GHz, so I've
been
using 15GHz. Haven't tried the auto mode, as I've just recently jumped
from V19.1 for 2021 R2. I'll give that a go.

How about GPU acceleration? Do you use that at all? We were considering
in getting a NVIDIA tesla or something similar, but I read that HFSS is
only useful with a GPU once the problem is of a certain size? I think
it's
more beneficial for the other solvers, correct if I'm wrong.

Thanks,

Hakim

On Wed, Feb 9, 2022 at 6:02 PM Scott McMorrow <scottmcmorrow@xxxxxxxxx>
wrote:

Hakim

I'd forgotten that I wrote this. It holds up pretty well.

The biggest thing that you can do to reduce solve time is to reduce the
domain size of the model. Model only what you need for what you are
doing. For a package to PCB model where you are trying to optimize
return
loss , limit the model to only the signal and ground balls needed.
Limit
the planes to just a bit larger than your ring of ground vias and make
sure
that you place a radiation boundary adjacent to the x/y planes. This
terminates the planes and keeps leakage energy from reflecting back and
forth.

The other thing that you can do is to use only wave ports. They are
better matched than lump or circuit ports and as a result will mesh and
solve faster most of the time.

Otherwise, more cores are your friends. If you don't have many cores
available for your solves, use one fixed frequency at Nyquist for your
Adpative mesh. Then simulation in Auto mode and the solver will apply
all
cores to the adaptation.

Happy Modeling

Scott



On Wed, Feb 9, 2022 at 10:17 AM Hakim <hakim@xxxxxxxxxxx> wrote:

Hi Scott,

I read this post about a a year ago and have used the settings
extensively, which has been great, thanks!

Just wanted to ask if you would use the same settings when running a
parametric setup?

I'm currently simulating a package to PCB model, which inculudes vias
spacing, anti-pad, ground vias, solderballs etc... Any tips you can
share
on how to do this efficiently?

At moment, it's taking me about 45mins per sweep, with a mesh of 15GHz
and Delta @ 0.02. The rest of the settings are the same as you've
mentioned in the post. Max frequency sweep is to 45GHz with 10MHz
steps
and not saving fields.

At the moment I'm just sweeping one variable at a time and manually
setting some variables to what I think will improve the response. I'm
hoping there's some way where I can cut-down the settings to get a
good
feel of where the optimum point is and what variables are key, then
home in
with tighter settings to sign-off.

Any other input would be great too

Thanks

Hakim

On Tue, Feb 2, 2021 at 12:11 AM Scott McMorrow <
Scott.McMorrow@xxxxxxxxxx>
wrote:

Ryan
Here's some of the "secret sauce". Out of 900 jobs the Samtec SIG has
run in the past month

Min Runtime 1 minute
Max Runtime 21 hours
Avg Runtime 1 hour 15 minutes
Median Runtime 22 minutes

We run small jobs on 16-core machines and large jobs on 48-core
machines
Connector models run in HFSS
Breakout models run in HFSS 3D Layout with HFSS Components used for
the
connector (or package ball) section near the launch.
Package models run in SIwave and HFSS
Board models with traces and connector launches run in SIwave with
HFSS
Regions.

Large connectors with say 64 ports, take around 4-8 hours depending,
using 192 cores on four 48-core machines.
Boards with multiple launches and traces that we use for cable
soldered
to boards can take anywhere up to a day in HFSS alone, and minutes in
SIwave with HFSS Regions.


* Box stock Ansys uses 30% mesh growth per adaptation cycle, which
is much too high and creates much larger meshes than necessary. 10%
mesh
growth per cycle was determined by many experiments to minimize solve
time.
* First Order Basis Functions are generally optimal. Mixed order
can cause non-linearity in the Mesh Adaptation profile.
* Never ever, ever, ever, ever save fields. This slows down
Frequency Sweeps significantly. If field saves are turned off, HFSS
will
solve terminal mode models without saving fields and optimize
S-parameter
generation. If you want to look at fields, run an additional solve
at a
specific frequency point.

New features in HFSS that help significantly:


* Tau Flex Initial Meshing with Dynamic Surface Roughness (this is a
huge performance increase)
* Auto solve mode with Maximum of 2, 3 or 4 Broadband Adaptation
Frequencies (depends on number of cores and nodes available)
* Do not save Adaptive Mesh Control files. (saves on memory and time
and is a design setting option.

Always remember that the Impedance that you resolve in HFSS is
directly
related to Adaptive Maximum Delta S, this is documented by Ansys as
an
asymptotic error bound. The relationship is not quite exact but
structural
impedance error is approximately the same as the Delta S. For a 50
ohm
line, Delta S Å¡.0075 resolves to approximately 50 +/- 0.4 ohm
(actually it
always resolves low, so 50 ohm TDRs at 49.6 ohm. Box stock DeltaS of
0.02
guarantees at least a 2 ohm error.

All of the above settings will optimize performance running on HPC
cloud
providers, also, including Nimbix and Ansys Cloud. Failure to use
some
of
these settings will totally mess up solve performance, wasting time
and
money.

Scott


From: si-list-bounce@xxxxxxxxxxxxx <si-list-bounce@xxxxxxxxxxxxx> On
Behalf Of Ryan Lott
Sent: Monday, February 1, 2021 5:47 PM
To: Scott McMorrow / Teraspeed <scottmcmorrow@xxxxxxxxx>
Cc: all.si.list@xxxxxxxxx; SI-LIST <si-list@xxxxxxxxxxxxx>
Subject: [EXTERNAL][SI-LIST] Re: Step Size for s-parameter extraction

Scott,
I know you run HPC but how many adaptive passes on average to reach
.002
for your average model at 30GHz? Model size/ram? Nodes? Just curious
if
you
know off the top of your head. I know it has to be HUGE but would be
informative and entertaining to hear your numbers. Imagine youre
doing
full
64+ port solves too direct solve mixed-order too. No monkey
business...
See
you're also doing 10% reduction- you think 10% setting for adaptive
pass
does better on average with convergences efficiency-wise on
connectors
than
default (20 or 30% - cant fully remember)? Makes sense according to
my
gut
but I've never ventured down that path myself. Never had issues
before
but
I'm always looking for spoons to make things faster while as or more
accurate.

On Mon, Feb 1, 2021, 4:10 PM Scott McMorrow <scottmcmorrow@xxxxxxxxx
<mailto:scottmcmorrow@xxxxxxxxx>> wrote:

Rajan,
First, overall accuracy is set by adaptive meshing, not by sweep.

Here is what I would suggest:

Adaptive Mesh Frequency 30 Ghz
Maximum DeltaS 0.0075 (if you want higher accuracy change to 0.002)
Maximum Refinement Per Pass 10%
Order of Basis Functions First order
POrt Maximum Delta Z0 - 0.01%
Number of triangles for waveport Min 500 Max 5000

Interpolating Sweep
Linear Step 0 GHz to 60GHz StepSIze 10 MHz
Max solutions 500
Tolerance 0.01% (eliminates interpolation glitches)
Advanced Options Enforce Passivity Passivity Error Tolerance - 1e-6










On Mon, Feb 1, 2021 at 3:57 PM Rajan Hansa <all.si.list@xxxxxxxxx
<mailto:all.si.list@xxxxxxxxx>> wrote:
Hello Experts,
I am extracting PCIe Gen 5 package traces in Ansys HFSS to check
Xtalk
specs for PCIe lanes. Extraction is done from 0 Hz up to 60 GHz.

I am told that for such extractions discrete sweep type should be
used
owing to its high accuracy compare to interpolating sweep type. do
we really need to use discrete sweep type at a high data rate?

Could anyone please help me understand how we should select step
size
when
discrete sweep type is used? I understand step size is proportional
to
trace length but I am not sure which equation should be used to
find
step
size from the trace length.

Thanks in Advance,
Rajan


------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx<mailto: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
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

<
//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

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

List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

<
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

List archives are viewable at:
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

<
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>




------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx<mailto: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
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

<
//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

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

List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

<
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

List archives are viewable at:
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

<
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>




------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx<mailto: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
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

<
//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

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


List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

<
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

List archives are viewable at:
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

<
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>


Scott McMorrow,

Office: 401-284-1827
www.samtec.com
<http://www.samtec.com>
<http://www.samtec.com
<http://www.samtec.com>





This email and any appended documents are only for the intended
person/entity and may contain information of Samtec, Inc., that is
PRIVILEGED, PROPRIETARY, CONFIDENTIAL, AND/OR PROTECTED BY LAW. If
you
are
not the intended recipient you are hereby notified that any
dissemination,
disclosure, use or copying of this email or its contents is
prohibited. If
you this message in error, please notify Samtec immediately and
delete
the
email, attachments and all copies. The intended recipient should not
disclose the content to third parties or reproduce the content
without
Samtec's written consent.







delete the email, attachments and all copies. The intended recipient
should not disclose the content to third parties or reproduce the
content
without Samtec's written consent.

------------------------------------------------------------------
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
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>

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


List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

List archives are viewable at:
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<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
<//www.freelists.org/webpage/si-list>
<//www.freelists.org/webpage/si-list
<//www.freelists.org/webpage/si-list>


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


List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>
<http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>


List archives are viewable at:
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>
<//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>


Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>
<http://www.qsl.net/wb6tpu
<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
<//www.freelists.org/webpage/si-list>

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


List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
<http://tech.groups.yahoo.com/group/si-list>

List archives are viewable at:
//www.freelists.org/archives/si-list
<//www.freelists.org/archives/si-list>

Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
<http://www.qsl.net/wb6tpu>



-- 

Charles Grasso

Dish Technologies

 (c) 303-204-2974

(h) 303-317-5530

(e ) charles.grasso@xxxxxxxx

(e2) chasgrasso@xxxxxxxxx

------------------------------------------------------------------
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 forum  is accessible at:
               http://tech.groups.yahoo.com/group/si-list

List archives are viewable at:     
                //www.freelists.org/archives/si-list
 
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu
  

Other related posts:

  1. Home
  2. si-list
  3. Archive
  4. 02-2022