[SI-LIST] Re: time domain simulation w/ PDN planes
- From: Bradley Brim <bradb@xxxxxxxxxxx>
- To: "cristian.gozzi@xxxxxxxxx" <cristian.gozzi@xxxxxxxxx>, "shlepnev@xxxxxxxxxxxxx" <shlepnev@xxxxxxxxxxxxx>
- Date: Fri, 24 Oct 2014 18:32:11 -0700
Hello all,
There are circuit, hybrid and EM solvers. They progress from fast and
approximate to slow and accurate. Each provides unique speed/accuracy and
design capacity tradeoffs. Circuit solvers can include many coupling effects
inside each component and EM solvers can ignore many effects in their modeling
of a design (e.g. losses, thicknesses, geometry details). Hybrid solvers couple
circuit components (e.g. TLs, vias, wirebonds, pads) with EM-solved portions of
the design (e.g. plane pairs). As Yuri describes this does not include
spatially distributed coupling of what he called a TM waves to a TL (i.e.
stripline). However, this distributed coupling will be orders of magnitude
lower than discrete types of TL-to-plane coupling that is considered by hybrid
solvers. How many orders of magnitude? Multiple. Enough that it would be
extremely difficult to quantify with 3D full-wave solvers and likely requires a
more theoretical/analysis-based quantification. TL-to-plane coupling occurs at
locations such as: vias through planes, TL crossing plane edge, TL parallel (or
nearly so) to proximate plane edge, etc. Spatially distributed TL to TM wave
coupling is negligible and will be swamped by any imprecision of modeling the
discrete couplings. Even in a 3D full-wave solver the adaptive mesh will
continue to focus on these discrete locations and probably never focus on
spatially distributed coupling areas for other than self-effects of the TLs
themselves. Would say this distributed coupling effect is "in the noise", but
more precisely "it's noise w.r.t. the noise" :-)
In most high speed circuits the dominant sources of PDN noise are: (1)
switching circuits, such as digital cores, TX and RX buffers and voltage
regulators, (2) signal vias through multiple planes. First order effects are:
power-aware IBIS models as Romi guided, considering core power noise sources in
your digital devices and modeling PDN-to-signal via coupling throughout the
circuit
SPEED2000 performs hybrid simulation in the time domain and is uniquely able to
address some PDN and signal applications. However, the originally stated goal
can probably be accomplished with a time domain circuit simulation engine as
long as it applies models for the full PDN extracted to include the dominant
self and coupling effects, the relevant sources of PDN noise (devices, VRMs)
and power-aware IO models.
Best regards,
-Brad Brim, Cadence
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Cristian Gozzi
Sent: Wednesday, October 22, 2014 7:20 PM
To: shlepnev@xxxxxxxxxxxxx
Cc: si-list@xxxxxxxxxxxxx; sherman.chen@xxxxxxx; nilesh_kamdar2@xxxxxxxxxxxx
Subject: [SI-LIST] Re: time domain simulation w/ PDN planes
Hi Yuri
Both Ansys SIwave and Cadence PowerSI have great modeling of interaction
between PDN and trace... They do not simulate transmission line and plane
separately as you described.
For instance in SIwave a 2D MoM is used to capture trace-trace, trace-plane and
plane-plane coupling all along the signal path... The MoM matrix is then
assembled together with transmission line model with the hybrid EM circuit
(spice like) sim...
Of course it's always a matter of the frequency vs simulation time trade-off
you are looking for...
For instance if you have an area of high Swiss cheese / perforation and you
want to well capture noise between trace and plane in these conditions, in that
case the 3D solvers are better, but you may need to reduce the problem size
only around the plane you want to simulate, otherwise the sim could be too
huge...
Fortunately the great job you and other EDA companies are doing on HPC can
drastically reduce the sim time of 3D for a PI analysis and makes things
possible today ;-)
Regards
Cristian
Il 22/ott/2014 13:14 "Yuriy Shlepnev" <shlepnev@xxxxxxxxxxxxx> ha scritto:
> Hi Sherman,
>
> Now you explained the problem :-)
> All hybrid solvers (both frequency or time domain) that use separate
> transmission line and transmission plane (parallel plane) models will
> not take this type of coupling into account.
>
> The reason is simple. Transmission plane model of the parallel planes
> includes electric field with only one component perpendicular to the
> planes and magnetic field components parallel to the planes (may be
> formalized as TM waves). These fields couple or excited by only to the
> currents flowing in the direction orthogonal to the planes - currents
> on vias for instance. On the other hand, T-line strips have currents
> predominantly parallel to the planes and are simulated in these hybrid
> solvers separately with the coupling only at the vertical transitions.
> In a simplified case with parallel planes with strips between them and
> TM waves propagating along the strips will be exactly orthogonal (do
> not
> interact) to the TEM waves propagating along the strips (even in
> presence of dielectric and conductor losses). Such modes do not
> interact!
>
> In reality, TM waves are not exactly TM and do not propagate along the
> strips. Small coupling takes place mostly due to discontinuities,
> inhomogeneities and losses. To construct a model with the coupling,
> you will need 3D analysis that accounts for all major factors
> contributing to the coupling. You should see that the coupling at the
> transition to the strip will be dominant. Though, I have not seen such
> investigation - may be someone on this list can provide further
> references.
>
> Best regards,
> Yuriy
>
> Yuriy Shlepnev, Ph.D.
> President, Simberian Inc.
> 3030 S Torrey Pines Dr. Las Vegas, NV 89146, USA Office
> +1-702-876-2882; Fax +1-702-482-7903 Cell +1-206-409-2368; Virtual
> +1-408-627-7706
> Skype: shlepnev
>
> www.simberian.com
> Simbeor - Accurate, Fast, Easy and Affordable Electromagnetic Signal
> Integrity Software
> 2010 and 2011 DesignVision Award Winner
>
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx
> [mailto:si-list-bounce@xxxxxxxxxxxxx]
> On
> Behalf Of Chen, Sherman
> Sent: Wednesday, October 22, 2014 11:47 AM
> To: nilesh_kamdar2@xxxxxxxxxxxx; si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: time domain simulation w/ PDN planes
>
> Nilesh,
>
> Thanks for the paper. My question is regarding how to model the
> coupling from PDN to the transmission line, not the transmitter or the
> receiver chip.
> The coupling effect across the whole span of the tline need to be
> calculated. To do that in circuit simulator what I can think of is to
> use a multi-port sparam which's ports are extracted btw the locations
> along the tline and the VRM, at the interval of say lambda/20. At each
> location the port need to be set at four points: DP+, DP-, Vcc on
> power plane, and GND on the ground plane.
> Any comments on this thought?
>
>
> -----Original Message-----
> From: nilesh_kamdar2@xxxxxxxxxxxx [mailto:nilesh_kamdar2@xxxxxxxxxxxx]
> Sent: Wednesday, October 22, 2014 11:38 PM
> To: si-list@xxxxxxxxxxxxx; Chen, Sherman
> Subject: RE: [SI-LIST] time domain simulation w/ PDN planes
>
> Sherman
>
> If I understood your question correctly, I believe that this
> simulation can be handled in ADS. To do this you need to create an
> SI/PI model of the physical interconnects that includes all the
> power/ground planes and also signal lines. Then you can apply specific
> time domain stimulus in a Transient simulation.
>
> Here is a DesignCon 2014 paper that explains how we can do this:
>
> Paper:
> http://cp.literature.agilent.com/litweb/pdf/5991-4083EN.pdf
>
> Slides:
>
> http://www.xilinx.com/events/designcon2014/11_WE5Slides_Touchstonev2SI
> PISPar
> ameterModels.pdf
>
> If you have more questions about this technique, you can contact me
> offline.
>
> Thanks
> Nilesh Kamdar
> Applications Engineer, Keysight Technologies
>
>
>
> -----Original Message-----
> ----------------------------------------------------------------------
>
> Msg: #1 in digest
> From: "Chen, Sherman" <sherman.chen@xxxxxxx>
> Subject: [SI-LIST] time domain simulation w/ PDN planes
> Date: Tue, 21 Oct 2014 14:07:19 +0000
>
> Hello experts,
> I'm trying to evaluate the impact of power plane noise on the HS signals.
> The sim setup would require to apply a noise source on the PDN
> (somewhere btw power plane and ground plane), also a PRBS source at
> the initial end of the t-line. To my knowledge it looks only Cadence
> Sigrity Speed2000 can run such SI/PI combined sim since the feedback I
> got from Keysight and ANSYS both said that neither ADS nor SIWave can
> handle such sim case where time domain sim needs to be run w/ the noise
> coupled from power/ground plane.
> Note the coupling needs to be from the whole plane rather than
> injected at one point for the latter we can simply use a sparam model
> of the PDN as the means of noise injection but this would not be the
> equivalent of the plane coupling, right?
> Any suggestion will be appreciated.
>
> Best Regards,
>
> Sherman Chen
> Signal Integrity
> EMC Global Hardware Engineering
> Tel: +86 21 60951100-3329
>
------------------------------------------------------------------
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:
