New Philips MOS20 LDMOS Model in SmartSpice

Introduction

Philips MOS20 was released in January 2004. Its purpose is to provide a high-voltage compact model to describe both operation of the channel region and drift region under the thin gate oxide. It can be used as a replacement for the macro-model composed of MOS9 and MOS30 to describe Lateral or Vertical Double-diffused MOS (LDMOS or VDMOS) or Extended-Drain MOS devices (EPMOS).

 

A Compact Model to Replace Two Macro-Models

Before MOS20 was released, simulating LDMOS devices required the use of macro-models. To do this, SmartSpice provided the following models: MOS9, MOS30 and MOS40.

Figure 1. Macro-model using MOS9 and MOS30.

 

One was used to achieve channel region simulation (MOS9) and another was required to account for the drift region (MOS30 or MOS40). Both were assembled at the netlist level, for example in a subcircuit.

MOS20 accounts for both regions, and above all internally computes the voltage at the transition between channel and drift regions. This improvement is very important with regard to convergence. It does not require any external node to link two models together, and this particular voltage is available to model equations since it is internally computed.

The only case when associating MOS20 and MOS40 is still necessary is when very high voltages devices are simulated. In this case, a macro model will help accounting for the drift region under the thick field oxide.

 

MOS20 Takes Advantages of all Philips Models

The model is based on all the best achievements in compact models. Model core has been derived from the SOI-LDMOS model from University of Southampton.

Figure 2. The region under the thin gate
oxide of LDMOS device (n-channel).

 

MOS20 equations are based on surface potentials. With this technique, it provides an accurate description of all operating regimes. Only one equation is computed for all regimes, ensuring that no discontinuity appears, and no smoothing function is used. The equations implemented are based on MOS11 equations. Surface potentials are calculated using an approximation of Poisson equation, in order to get explicit expression of surface potential with regard to node voltages.

This way, MOS20 benefits from all the improvements and experience acquired with those models.

In addition to these core equations, MOS20 also accounts for :

  • Mobility reduction

  •  
  • Velocity saturation

  •  
  • Drain-Induced Barrier Lowering (DIBL)

  •  
  • Static feedback

  •  
  • Channel length modulation

  •  
  • Weak avalanche current

 

Silvaco Implementation

MOS20 implementation in SmartSpice takes advantages of both Philips’ original equations, and SmartSpice common MOSFET features such as :

  • Transient Noise analysis using Philips detailed noise model

  •  
  • RF analyses with SmartSpiceRF

  •  
  • Speed improvements : VZERO and BYPASS options, parallel architectures

  •  
  • Convergence control and user-friendly hints

  •  
  • Out-of-bound guards, both for parameters and internal values

  •  
  • Extrinsic elements : bulk diodes, Source/Drain resistances

All these features allow the user to make the most of MOS20 model.

Examples
The plot shown in Figure 3 is a simple Id=f(Vd, Vg) characteristic. The uprising part of the curve is due to the weak-avalanche current (or impact ionization). This current variation is important because it has an influence on dissipated power, that results in temperature elevation. Since this is a power device, temperature is critical.

Figure 3. Id=f(Vd,Vg) characteristics,
showing avalanche current.

 

Another interesting plot (Figure 4) is the potential at the transition between channel region and drift region. These curves help to investigate how the model is computed, giving an inside view of the specific LDMOS structure. This curves shows how VDi potential varies and reflects where the transition lies, between VD and VS potentials.

Figure 4. Potential at transition between
channel and drift regions vs Vg, Vd.

MOS20 also provides an accurate charge description. For Gate and Bulk charges, the nodal charges are simply decomposed into drift and channel contribution. The Drain charge is computed a different way, distinguishing two cases : well above threshold, and below threshold. The following example is a plot of CGG capacitance, with regard to VD and VG (Figure 5).

Figure 5. Capacitance charge CGG and VD and VG.

Conclusion

The MOS20 model is relevant for its ability to account for high-voltage devices without using macro-models. Furthermore, it is based on all the latest modeling techniques such as surface potentials. For these reasons, users can rely on an accurate model with good convergence properties.

 

References

  1. Detailed model parameters and equations can be found in SmartSpice modeling manual vol 1.
  2. Philips’ website also contains all the documentation and literature about MOS 20 : http://www.semiconductors.philips.com/Philips_Models