Mextram 504

1
Mextram 504

• Doug Weiser
• October 13, 2005

2
Bug Support

• Problems with SiMKit 2.0.1
• Reported 4/2005
• Self-heating examples do not work.
• Resolved in SiMKit 2.1.1 released 4/22/2005 but
  could not be verified until 10/7/2005
• Multiple values of Gmin
• Reported 4/2005
• These are in the solver (two values) and CMI code
  (two values, one mapped to simulator Gmin).
• Verilog-A from Delft only has one value set as a
  model parameter.
• This still remains an open issue.
• Solver and self-heating
• Reported 6/22/2005
• Starting with SiMKit 2.1.1, the solver added the
  dt node with the capability to add a more complex
  thermal network or thermal coupling between
  devices.
• However, it is not clear how to turn this off so
  that a single device can be simulated with
  self-heating controlled by the model RTH/CTH
  values.
• Resolved with information from Philips 10/7/2005

3
Simulator Support

• TI supports many different simulators beyond
  TISPICE
• ADS 2005A.400 SmartSPICE 2.17.9.B
• Eldo v6.5_2.1 Spectre 6.0.1
• HSPICE W-2005.03 Verilog-A 504.6 10/2005 Beta
• Testing of available versions (not necessarily
  the latest) shows problems from small (not
  significant relative to process variations) to
  large (results not suitable for circuit design).
• Before TI can make wide use of Mextram, simulator
  versions with consistent, reasonable
  implementations are needed.
• TI has worked with two simulator vendors and
  Delft to correct implementation issues and will
  work with other vendors in order of business
  priority.

4
Enhancement CCSO and SW_ET

• Add support for CCSO
• Represents oxide trench isolation capacitance
• No bias dependence
• Analogous to CBCO and CBEO which are supported
• Add support for SW_ET
• Switch to control self-heating at the instance
  level
• Default is on with control by RTH value
• Being supported in the CMC R2 model
• Suggest following same methodology
• Only need self-heating for some instances

5
Enhancement Internal Base-Collector Model

• Without a buried layer, lateral voltage (IR)
  drops in the Collector forward bias the internal
  Base-Collector junction.
• This makes fitting DC and AC in quasi saturation
  difficult.
• Without a buried layer, collector resistance is a
  strong function of both Vcb and Vcs.
• Code in simple bias-independent splitting first.
• Develop bias dependent physical model later.
• Even in high voltage devices with a buried layer,
  vertical voltage (IR) drops in the Collector
  forward bias the internal Base-Collector junction.

6
Enhancement Internal Base-Collector Model

• Modify latest Mextram 504.6 Verilog-A model from
  Delft to add bias independent intrinsic/extrinsic
  Base-Collector diode splitting and show impact on
  an oxide isolated device.
• New parameter XIBIC with default of 1.0
• Modified equations
• IexI (1-XIBIC)f(Vb2c2) Iex XIBICf(Vb2c1)
• Also need to update
• Qex
• Shot Noise
• Power dissipation
• May need to split Ib3?
• Add current contribution from b2 to c2

7
Enhancement Internal Base-Collector Model
8
Enhancement Internal Base-Collector Model