[SI-LIST] Re: SI Simulation of GHz signals
- From: "databits" <databits@xxxxxxxxxxxxx>
- To: <clifford@xxxxxxxxxxxxx>, <si-list@xxxxxxxxxxxxx>
- Date: Mon, 26 Jul 2004 19:50:16 -0500
Clifford...
The best solution that I have found:
** Use S-Parameter models for all the passive devices...=20
** Use a SPICE model for the driver/receiver as received from your =
vendor
** Get a mixed mode simulator that has a native harmonic balance / =
frequency
domain system (Ansoft Designer or Agilent ADS, other??). The harmonic
balance is the frequency domain portion, while the time domain is =
typically
handle by some method of convolution.
Sparameter simulations by themselves are lightning fast! The time =
domain
sim (i.e. convolution) does slow things down a little... but it is a =
heck
of a lot faster than a stack of hspice models for the passives.
In my experience, I have been able to get a...
"transmitter, 0.125 meter daughter card, connector, 1 meter backplane,
connector, 0.125 meter daughter card, receiver"=20
... to solve in less than 10 minutes for 3, 6, 10, 10+ GB/s
Of course, the generation of the sparam models took significantly =
longer.
But then again that is where the passive vendor comes in to play. =20
(Being involved in the passive interconnect market, the company I work =
for
does this kind of work all the time to support our products. Get me =
offline
for further details.)
With regards to passives... s-parameter models... and the sparameter =
models
that have signal / return tailored to a specific application.... Well, =
it
is definitely easier to create specific pattern sparam models... But =
sparam
modes that do not have a specific return path predefined can also be =
made in
many cases and have been done that way for years. =20
I did a bunch of work in the IBIS interconnect standard... and got a =
number
of hooks built in such that the standard allows for multiple types of
modeling methods... but also provides a format to map nodes to pins of
externally referenced s-param models. S-Parameters were specifically
include to handle multiple frequency and lossy effects!
I my (not so) humble opinion, the sparameter will be the defacto model
standard for speedy passive (cable, connector, pcb) interconnect =
modeling.
If needed, the IBIS standard can help with creating a standard node =
mapping.
But you can always use the standard touchstone file.
If needed feel free to contact me directly.
_gus
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] =
On
Behalf Of Clifford van Dyk
Sent: Monday, July 26, 2004 12:32 PM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] SI Simulation of GHz signals
Hello
I would like to perform reliable simulation of GHz signals (up to=20
3.125GHz), specifically Xilinx RocketIO. I would like to include in the=20
simulation the effects of the following:
Driver->PCB trace (incl. vias)->connector->cable->connector->PCB
trace(incl. vias)->Receiver
The models that I have obtained from the various vendors are HSPICE=20
models. I have evaluated two of the most recommended S.I. toolchains:=20
Mentor G. HyperLynx and Cadence Spectraquest. Both tools use HSPICE as=20
the simulation engine, and essentially act as a front-end gui to HSPICE, =
as well as extracting the PCB trace/via models. My experiences thusfar=20
with both tools have not been good. Anything but the most simplistic of=20
traces causes the tools to either crash or take rediculous time to=20
process (of the order of hours for even a simple net). A further issue=20
is that both of these tool vendors claim that the HSPICE simulator is=20
not necessary, and that the simulation can be performed without it, but=20
practically this is not the case, due to a lack of availability of=20
reliable models in anything other than HSPICE format. The conversion=20
from HSPICE to any of the custom modelling types is also, in my opinion, =
non-trivial and potentially an extremely tedious manual process.
I believe that S-Parameter based simulation provides much faster=20
simulation, but again there is a lack of availability of S-Parameter =
models.
Can anyone recommend a method for simulating the above signals that is=20
simple, robust and reliable, or is the simulation of such signals still=20
premature? Coming from a HW design background, I am fairly new to S.I.,=20
but it seems surprising that there is no industry-standard modelling=20
type (equivalent of IBIS) that cable/connector vendors will provide, but =
maybe this level of simulation is in its infancy, and S-Parameter models =
will emerge as the standard?
Is the simulation of such signals entirely necessary? I am dubious about =
the reliability of the results of such simulations, and I am wondering=20
whether it is not more practical to just take all the precautions=20
possible and hope for the best!
Please let me know if you have any advise, or a good solution to my =
dilemma!
Kind regards,
Clifford
------------------------------------------------------------------
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 FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
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:
