3.4 Practical Use

For the use of the analysis functionality of the SIESTA framework the following assignments are necessary. First the model to be analyzed must be specified. Such a model in general consists of a series of simulations which generates the queried results. Then the strategy is defined how the results should be obtained. This depends on the type of analysis, e.g., manufacturability analysis using Gauß distributions for the process parameters.

Figure 3.8: Experiment file for the DoE.

(doe (model      "analyze.mod")
     (design     CCI)
     (free       ew vb)
     (target     current))

In Figure 3.8 the experiment file of an analysis using a Central Composite Inscribed design is shown. In this file the first keyword describes the type of analysis, a DoE (doe) followed by some parameter assignments. The experimental points are arranged in a Central Composite Inscribed (CCI) design. The free parameters are the work function (ew) and the voltage of the bulk contact (vb). The response is the drain current (current) of the transistor.

Figure 3.9: Model file for the extraction of the drain current ("analyze.mod").

(process-model imeter
  (inputs 
   (ew          bound-float -0.56 -0.62 -0.5)
   (vb          bound-float  0.0  -2.0   2.0))

  (outputs      (current     float))
  
  (output-mapping (current mmnt.result Id))

  (tool-flow
   (mmnt (tool
           (controls (command    "mmnt")
                     (cmdline    "ipd")
                     (ipd        (file      "mmnt.ipd")))
           (results  (result     "mmnt.crv"  operating-point))))))

The model function "analyze.mod" is listed in Figure 3.9. Here the function for obtaining the result (current) from a set of input variables (ew and vb) is defined. In this model, called imeter, a process is executed for generating a result (process-model). In the inputs section of the file the names and the settings of the control parameters ew and vb are defined. These two variables are of the type bound-float which allows three settings: the nominal, minimum and maximum value.

The executed tool is the device simulator MINIMOS-NT (command) with the command line options given in cmdline. Here the only option is the filename "ipd" which is a copy of the MINIMOS-NT input deck file "mmnt.ipd" located in the same directory as this model file Figure 3.9. For passing the values of the control parameters to the program the file "ipd" is prepared by a pattern replace mechanism. Each input parameter enclosed in <( and )> (for example <(ew)>) is replaced by the current value of the parameter. The input deck for the device simulator is shown in Figure 3.10 (more details about the features and the Input Deck Library keywords of the MINIMOS-NT device simulator can be found in [3]).

Figure 3.10: Input deck for the MINIMOS-NT device simulator ("mmnt.ipd").

#include "$VROOT/defaults/defaults.ipd"

Device : ~DeviceDefaults {
  +Source = 0.0 V;
  +Drain  = 2.0 V;
  +Gate   = 2.0 V;
  +Bulk   = <(vb)> V;

  Input { 
    file  = "mmnt.pbf"; }

  Output { 
    file = "mmntout.pbf"; }

  Phys  {
    +Gate { 
      Contact { 
        Ohmic { 
          Ew = <(ew)> eV; }}}}}

Curve {
  file = "mmnt.crv";    
  Response {
    +Id = output("Device", "I", "Drain");}}

The result of the device simulation is the file "mmnt.crv" consisting of several current and voltage values. In the output-mapping the drain current (Id) is selected from the file and stored in the variable current which is assigned to be a float (see outputs). For each evaluation of the model function the MINIMOS-NT device simulator is started and calculates a new operating point.