W8632EP Datasheet, PDF (3/8 Pages) Keysight Technologies – Model Quality Assurance

03 | Keysight | Model Quality Assurance (MQA) - Brochure
MQA
MQA is a unique software product developed to solve the following problems:
1 SPICE model validation
is becoming increasingly
important and significantly
more difficult. This is
because:
2 Model validation encom-
passes much more than just
overlaying the measurement
results to simulation results
of the model.
3 Model validation should
be automatic and
customizable.
–– As the channel scales down,
second-order physical effects make
device modeling more complex.
–– Macro models and binning models
have been used extensively.
Validating these models is much
trickier than global models.
–– A natural consequence of
foundry business requires a
better way of communicating
between modeling engineers
and designers. Designers
often need to check whether
the models satisfy their
requirements for some specific
circuit design needs.
–– What appears to be a good model
for certain application can turn
out to be a terrible one for other
applications.
–– After all, measurement is limited to
the number of physical devices in the
test structure and the resolution of
instruments.
–– Model validation should include the
following checks:
–– Accuracy of the model (compare
with measurement).
–– Completeness of the model (have
all the major physical effects
important to the design been
modeled?).
–– Mathematical robustness of the
model (no kink in first and second
derivative).
–– Capability of the model to predict
physical trends (very important in
design optimization).
–– Model simulation results using
benchmark circuits.
–– The quality can only be guaranteed
after fixed QA procedures are in
place.
–– Manually validating a model is
nearly impossible considering the
large number of checks for different
device sizes, temperatures, and bias
conditions.
–– Model reporting is often
time-consuming and should be
expedited.
–– Model QA routines often change
with model modifications; a
customizable QA platform is
needed.
–– QA tools should help users debug
model issues and point out
potential problems.
Model QA
Check model completeness
–– Are Isub, Igate, 1/f noise, etc. covered
in the model?
Model parameter range check
Check the trend of model characteristics
–– Verify that the trends of Idsat, Vth,
Gm, Gds, vs W/L/T are correct
Compare model performance of different
models, such as:
–– HSPICE model with Spectre model
–– BSIM3v3 model with BSIM4 model
–– Macro model with its core model
––
Check and compare different process
corners
Check analog/RF design targets using
special routines
–– Smith and polar chart plotting
–– Check the trend of network
parameters
–– Load-pull and harmonic balance
simulation and plotting
–– Thermal noise characterization
Check numerical robustness of the model
–– Any kinks in Gm or Gds curves?
Check bin continuity
Check benchmark circuit performance
–– Ring Oscillators for example
–– Users can input their own circuits
Calculate point simulation value
according to user’s specification
Check model accuracy with the
measurements