Jazz Technology Conerefence

1

Jazz Customer PSP feedbackCMC Q4 2006
MeetingSan Francisco12/15/2006
2
Company 1

• Here are our comments1) Continuity for high
  order derivatives at vds0 is a necessity for
  high performance programmable analog circuit
  design. FET devices were used as switches in the
  signal path biased at vds0. PSP models showed
  the expected trend of distortion behavior (ie IP3
  curve had a slope of 3), while the bsim3 models
  were completely incorrect. PSP models have
  therefore significantly increased our confidence
  in our simulation results. 2) Our experience is
  that although simulation time is longer, by up to
  a factor of two, it is still acceptable. Our
  design is on a bicmos process and so the
  simulation time is not solely dependent upon the
  PSP models alone. With the increasing affordable
  computational power offered these days by new
  computers, this becomes less of an issue. The
  length of simulation time tends to only become an
  issue when it starts to exceed a certain pain
  threshold. Beside, it is still quicker to run a
  longer simulation than waiting for three months
  to find that we need another design spin. 3) No
  one has reported any convergence problems using
  the PSP models with spectre. This includes the
  original veriologA code and the spectre mmsim 6.1
  binary models.

3
Company 2

• For most RF and Analog circuits, linearity
  parameters, such as IP3 and IP2 for RF circuits
  and THD for analog circuits, are key parameters
  for products. Meanwhile, in todays differential
  circuits world, MOSFETs are very often needed as
  switches or variable resistors with Vds biased at
  0V, offering superior solutions in many
  applications. With BSIM model, due to its
  fundamental limitation for VDS0 scenario, its
  impossible to simulate the linearity of the
  circuits with MOSFETs used in the way as
  described above. PSP model takes away such
  limitation it enables us to seek better
  solutions. Its not a matter about simulation
  time its a matter about something possible or
  impossible. Its fundamental. We had successful
  designs with SP model, and we are using PSP model
  for our next generation products.

4
Company 3

• We evaluated of PSP mostly on a single transistor
  basis. On circuit level we did simulation for
  only a cascaded switch and a passive mixer with
  PSP. For the circuits under study we had the
  following observations
• PSP converges faster in HB analysis for both of
  the applications.
• The simulation results of PSP are more
  reasonable, based on our knowledge of circuit of
  these two types.
• We did not observe any high computation overhead
  compared to BSIM. Probably the circuit under
  evaluation is simply too small (lt15 transistors)