**5G and IoT Supplement**

Fingertip Gesture Control Interface for "Invisible" Devices

Testing 5G: Time to Throw Away the Cables

NB-IoT and eMTC Make 4G Networks Ready for IoT

This ad will close in seconds. Skip now

Advertisement

Advertisement

Recent Searchesmobile learning / southeast university / rto2000 / soc / jac capital

Advertisement

Advertisement

Industry News
# Advanced Large-signal Models for Transistor Parameter Extraction Software

#### An upgraded version of the LASIMO software that can conveniently make MESFET measurements and fit the data to advanced large-signal models

August 1, 1999

One Comment

**Advanced Large-signal Models for Transistor Parameter Extraction Software**

**Optotek Ltd. Kanata, Ontario, Canada**

Accurate model generation is becoming a progressively more important part of the computer-aided-design process. Of special importance are large-signal models capable of accurately predicting the performance of nonlinear circuits. LASIMO large- and small-signal modeling software^{1} facilitates the development of large-signal models by simplifying procedures for the extraction of MESFET and high electron mobility transistor (HEMT) model parameters. The program provides the designer with a set of proven transistor models^{2} as well as the capability to incorporate user-defined models and subsequently modify these as required.^{3}

Model parameters are optimized to match actual transistors by fitting the measured and modeled bias dependence of the device characteristics. In the user-defined model version of LASIMO, provisions were made for five DC and five capacitance user-defined large-signal models. The user-defined models were implemented as dynamic link libraries (DLL) created and added outside the program and linked at run time. Each of the DLLs is generated from a set of project files written in the C language and loaded into a C compiler. The default DC model is the Curtice model, and the default capacitance model is the Basic Semi-Junction Model. The designer has only to edit a single function that computes the nonlinear parameters. This function receives a set of arguments from the program and returns the required nonlinear parameters. The user is able to define and access up to 13 large-signal parameter coefficients to be optimized for each DC model and up to 15 large-signal parameters for each capacitance model. These DC and capacitance large-signal parameters can be selected by the user to suit various nonlinear simulators.

**The LASIMO Upgrade**

An upgrade to LASIMO version 2.1 has been released that incorporates additional advanced models, including the Alpha Own Model (AOM); TriQuint Own Model, Level 3 (TOM3); and TOM3, Modified. These models are implemented in the user-defined configuration to accommodate future refinement or user modification. The program's model complement is listed in **Table 1** . The reconfigurable models can be modified by the user, or other user-defined models can be substituted.

| |

Built-in Standard Models | |

I | |

Model 1: | Curtice |

Model 2: | Statz(Raytheon) |

Model 3: | Materka-Kacprzk |

Model 4: | TriQuint Own Model |

Model 5: | Advanded Curtice |

Model 6: | Curtice-Ettenberg |

Model 7: | Lehovic Zuleeg |

C | |

Model 1: | Junction Model |

Model 2: | Statz(Raytheon) |

Model 3: | A physically based model |

I | |

Model 1: | Curtice |

Model 2: | Advanced Curtice |

C | |

Model 1: | a physically based model |

Reconfigurable Models | |

Model 1: | Curtice and Basic Semi-Junction Capacitance |

Model 2: | TOM3 (no temperature dependence) |

Model 3: | Alpha Own Model |

Model 4: | TOM3 (with temperature dependence) |

User-defined Models | |

Models 1-5 | User assignable as substitutes for the reconfigurable models |

**The Alpha Own Model**

AOM is a comprehensive model for GaAs MESFETs, which expands upon aspects of the TOM to account for dispersion, self-heating effects and charge conservation. A set of capacitance and charge equations are used for consistent small- and large-signal models. Transconductance and output conductance dispersion are modeled by combining a feedback network and subcircuit that describes the self-heating effects. The new model accurately predicts the I-V, CV, bias-dependent S-parameter, waveform, power and linearity characteristics of the MESFET. This model has been implemented in PSPICE.^{4,5}

**The TriQuint Own Model, Level 3**

TOM3 is also a comprehensive model for GaAs MESFETs. It was developed to improve existing MESFET capacitance models for SPICE using conservation of charge in the implanted layer.^{6} The capacitance model calculates the gate charge from the drain current and gate capacitances from the drain conductances. Relating the gate charge to the channel current creates gate capacitances dependent upon the channel current derivatives linking the small-signal model to the large-signal equations. Drain dispersion and self-heating effects are modeled by a GD model using a set of device equations and a specific subcircuit in SPICE.^{7,8}

**The TriQuint Own Model, Level 3, Modified**

A variant of the TOM3 model is also provided where the parameters assigned to temperature are not included. For many applications the model can be used quite effectively without the temperature parameters. In addition, speed of extraction is improved.

**A Large-signal Model Extraction Example**

As an example of the use of the new user-defined models, the fitting of the AOM to a discrete GaAs MESFET is demonstrated. The transistor, a low power, low noise device,^{9} was selected from the process control monitor in a three-inch GaAs wafer. Relevant data include a gate width of 300 mm (four interdigitated 75 mm fingers), gate length of 0.5 mm, ion implantation energy of 150 Kev and open channel current after recess of 100 mA. The transistor was measured using an on-wafer probe station and a Wiltron 360 vector network analyzer (VNA). The device was powered using two Keithley model 236 programmable power supplies. The VNA and power supplies were controlled via the IEEE-488 GPIB bus using the program's data acquisition module. The FET's S parameters and drain current were measured at Vds = 0.5, 1.0, 1.5, 3.0 and 5.0 V and Vgs = -1.8, -1.2, -0.8, -0.4 and 0 V. A total of 25 data points were thus generated to represent the device's I-V curve. **Figure 1** shows the small-signal model chosen to represent the device.

The device parasitic resistances were measured using in-house PCM characterization procedures,^{10} resulting in Rs = 1.3 W, Rd = 2.7 W and Rg = 1.1 W. The parasitic inductances were then extracted using the program, producing Ls = 36.35 pH, Ld = 4.21 pH and Lg = 77.46 pH. With the bias-invariant parasitics, LASIMO was used to extract the intrinsic device models at all bias points. In this process, the program gathers in memory the arrays of transconductance Gm, output conductance Gds, gate source capacitance Cgs and gate drain capacitance Cgd. These arrays are a function of the intrinsic biases Vgsi and Vdsi. The intrinsic biases are calculated by subtracting the impressed voltages from the parasitics.

The program then is used to optimize these small-signal data arrays to the AOM equations representing the nonlinear Ids, Gm, Gds, Cgs and Cgd. The initial DC model values for optimization were chosen with care so that the optimization process converges rapidly. As an example, key DC parameter values were roughly calculated to produce the device's saturation current Idss. The rest of the model parameters were chosen from default values. A combination of random and Levenberg-Marquardt optimizers were selected and care was exercised to ensure the values were realistic and within limits. The initial and final AOM model parameters are shown in **Figure 2** .

LASIMO graphs of the fitted DC parameters demonstrate the versatility of the new AOM model in achieving a best fit for Ids, Gm and Rds, as shown in **Figures 3** , **4** and **5** , respectively. Similarly, the AOM model demonstrates a good capacitance fit for the transistor, as shown in **Figures** **6** and **7** .

**Conclusion**

LASIMO software can be used to conveniently acquire MESFET measurements and fit these data to advanced large-signal models. The system requirements to operate the program properly include a Pentium PC with Windows 3.1/95/98/NT and a minimum of 16 MB of RAM. LASIMO Version 2.1 is priced from $7500 and is available immediately. Additional information can be obtained from the company's Web site: www.optotek.com.

**References**

*1. Optotek Ltd., Kanata, Ontario, Canada K2K 2A9.*

*2. "GaAs MESFET and HEMT Model Extraction Software," Microwave Journal, Vol. 38, No. 4, April 1995, pp. 274-276.*

*3. "Large-signal Modeling of MESFETs and HEMTs," Microwave Journal, Vol. 40, No 11, November 1997, pp. 162-166.*

*4. C.J. Wei, Y.A.Tkachenko, D. Bartle, S. Dindo and D. Kennedy, "A Compact Large-signal Model of a GaAs MESFET," Microwave Journal, Vol. 40, No 12, December 1997, pp. 22-34.*

*5. C.J. Wei, Y.A. Tkachenko and Dylan Bartle, "An Accurate Large-signal Model of GaAs MESFET Which Accounts for Charge Conservation, Dispersion and Self-heating," IEEE Transactions on Microwave Theory and Techniques, Vol. 46, No. 11, November 1998.*

*6. R.B. Hallgren and P.H. Litzenberg, "A TOM3 Capacitance Model: Linking Large- and Small-signal MESFET Models in SPICE," IEEE MTT, Vol. 47, No. 5, May 1999, pp. 556-561.*

*7. R.B. Hallgren and D.S. Smith, "TOM3 Equations," TriQuint Report, September 14, 1998.*

*8. D.H. Smith and A.J. McCamant, TOM Model Library User's Guide, Version 2.0 for PSPICE 4.04, TriQuint Semiconductors Inc., December 3, 1990.*

*9. S. Dindo, R. North and D. Madge, "A Manufacturing Process For Gallium Arsenide Monolithic Microwave Integrated Circuits," Canadian Journal of Physics, August 1987, pp. 885-891.*

*10. H. Fukui, "A Determination of the Basic Device Parameters of GaAs MESFET," BSTJ, Vol. 58, No. 3, 1979, pp. 771-797. *

**Optotek Ltd., Kanata, Ontario, Canada (613) 591-0336.**

Get access to premium content and e-newsletters by registering on the web site. You can also subscribe to Microwave Journal magazine.

Advertisement