Hi Karthik,
here is my suggestion: if you want to validate the PDN performance you can use
an S-parameter or Z-parameter model, but if you want to "visually" see what
happens inside the PDN and come up with improvements you can use a schematic
based RLC model. ChipQuake software standard version generates both s-parameter
and schematic RLC for a chip+package+PCB system (used for PDN architecture
exploration).
Regards,
Cosmin
http://www.piscanner.com
On Thursday, August 6, 2020, 10:08:51 PM PDT, Zachariah Peterson
<zmp@xxxxxxxxxxxxxxxxxxxx> wrote:
RLC models are partially phenomenological; they are a nice way to think
about what happens in a real PDN, transmission line, or any other system
that is electrically distributed in space and may include discrete
components. There are certainly parasitics that live in the PDN just due
to its geometry, and the behavior of signals in the presence of these
parasitics can be described in the language of RLC models.
About the only parasitics you can accurately include as equivalent
circuit elements in an RLC model is the plane capacitance for a uniform
plane pair with simple geometry, input capacitances on integrated
circuits, bond-wire inductance, and any other parasitics that might be
described in component datasheets. If you have direct measurements of
all these that certainly helps. Note that this might only be true for a
SISO system.
Wideband S-parameter models are fine, but I personally hate using
S-parameters for everything because they get so overgeneralized. If you
can extract an S-parameter model for your PDN you can create a transfer
function for a SISO PDN. For a real multiport PDN (SIMO or MIMO system),
you need a broadband transfer function matrix, and then you basically
know everything about the PDN. If you're in the Laplace domain, you can
look at poles to see the transient response and the resonances.
What you really want is the PDN's Z-parameters, which will show you the
self-impedance and transfer impedance for different ports/groups of
ports in the PDN. However, no one talks about Z parameters unless you're
using some field solver like Ansys. My opinion is use this to calculate
the power bus ripple from the transient current draw into
single/multiple ports in the frequency domain and then convert back to
the time-domain, then you can get the peak-to-peak or RMS ripple and
compare to your specs.
Take a look at Istvan Novak's 2019 DesignCon presentation where he talks
about some of these issues for PDN filtering and impedance optimization:
http://www.electrical-integrity.com/Paper_download_files/DC19_Tutorial_SLIDES_HowToDesignGoodPDNFilter.pdf
I know this is a long-winded answer to a simple question, but hopefully
this helps.
--
Zachariah Peterson
Owner, Northwest Engineering Solutions LLC [1]
On 2020-08-06 18:52, Karthik Chandrasekar wrote:
Experts
I wanted to know for Package/PCB PDN models for Core PDN's is the typical
recommendation to use RLC models (lumped/distributed ) or S-parameter
models. Any tradeoffs or factors to consider and which is more accurate ?.
The context for the question is sometime I see S-parameter PDN models to be
inaccurate depending on the tools used
Regards
Karthik
------------------------------------------------------------------
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:Â Â
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
