HVEKV Model Review

1
HV-EKV Model Review

• Richard Cherne
• Dec 2006

2
Comparison of HV-EKV vs Intersil Composite Models

• Obtained dataset from lateral 15-volt
  Extended-Drain PMOS Device
• Tox 12.5 nm, Leff 1 um
• Six device curves from 2 wafers used for opt.
  target
• Intersil composite model (sub-circuit) was fit
  using standard extraction and optimization
  routines.
• HV-EKV Model was fit to same dataset and compared
  to Intersil composite model fit.

3
Intersil Model vs Silicon Overlays

• Intersil model based on BSIM3v3
• External elements added to handle drift region,
  quasi-sat, etc.
• Model fit over body biases of Vb 0, 1, 2, 3, 4
  volts simultaneously.
• Linear and sat region Id-Vg optimized
  simultaneously followed by Id-Vd optimization.
  Some iteration required.

4
Intersil Model vs Silicon Overlays
5
Intersil Model vs Silicon Overlays
6
Intersil Model vs Silicon Overlays
7
Intersil Model vs Silicon Overlays
8
Intersil Model vs Silicon Overlays
9
Intersil Model vs Silicon Overlays
10
HV-EKV Model Fit

• Subthreshold region fit OK.
• Model gave reasonable fit when optimized over
  body biases of Vb 0, 1, 2, 3, 4 volts
  simultaneously for Id-Vd.
• Output resistance off somewhat.
• Parameters used in Id-Vd fit E0, KP0, res,
  UCRIT0, LAMBDA, teta_acc0, vsat0, GAMMA0.
• Had trouble fitting over both Id-Vd and low-Vds
  Id-Vg.

11
HV-EKV Model Fit
12
HV-EKV Model Fit
13
HV-EKV Model Fit
14
HV-EKV Model Fit
15
HV-EKV Model Fit
16
HV-EKV Model Fit

• Low-Vds region fit was poor after Id-Vd fit.

17
HV-EKV Model Fit
18
HV-EKV Observations

• HV-EKV model fit for Id-Vd was encouraging.
• Self-heating routine seemed to work as expected.
• Impact Ionization functioned as expected.
• Need help with optimization routine
• -- Appropriate target?
• -- More degrees of freedom would have liked
  BSIM-ish ub, uc, keta, pdiblc

19
AC Analysis
20
Cg Ramp
21
Cgd Ramp
22
AC Analysis Comments

• Need help with ac model fit.
• Cgd ? 0 under nominal drain bias?
• Are D/B and S/B diodes available, or must they be
  added externally?

23
Conclusions

• Cursory review only.
• Verilog code ran with no problems in spectre.
• Single device simulations only
• No information on convergence with large number
  of devices.
• Model seems to work as reported.
• Liked the simplicity.
• Reasonable fit considering limited number of
  fitting parameters.
• Intersil would consider possible use of model if
  fitting issues over all regions of operation are
  resolved.