GCG
Previous Top Next



graphic

How to define the computing step

There are two ways to define the computing step.  The first is to use the interval when temperatures are written to the results file.  As defined below, this print interval can change with time during the solution.  If it does then the computing step will also change with time.

There will be times when the model will not converge when using the print interval for the computing step.  This is easily seen during the solution.  You will see the two convergence values either increase or simply oscillate endlessly.  In this case you will have to stop the solution and use the second method. 

The second method is to define a constant computing step by checking the "Use constant computing time step" checkbox.  In the Compute Time edit box, enter a smaller computing step. If the model still does not converge, stop the solution and enter a smaller value.

Start Time

Enter the start time for the solution in the Start Time edit box.  This is typically 0.0.  If some properties are time dependent, then it may be desired to start the model at a time other than zero.

Maximum Number of Iterations

During a single computing step, this defines the maximum number of iterations will go.  If it is met the temperatures at that calculation will be used even though the convergence criteria has not been met.  It is highly recommended that a very large value be used so the convergence values are met.

Defining a print interval

When a transient model is run, temperatures can be saved to disk at predefined points in time. The print times can be at constant and equal intervals for the entire solution or can vary during the solution.  It may be desired to print temperatures more often when a time dependent heat flux is at its peak.  The data is defined by defining a print interval and an end time for that interval.  To define a single print interval for the entire solution, enter the interval and the total run time for the model.  To define a time varying print interval, define pairs of data with the time value increasing. The last time value entered will be the total run time for the model.

To change an existing value, either double click on the row or select the row then hit the Edit button.  To add a data pair, hit the Add button.  To delete a pair of data, select the row then hit the Delete button.

Get Print Times

When you hit this button Solaria will go through all referenced arrays that are time or periodic dependent and adds the times to the print times.

Initial Temperature

The Initial Temperature edit box is used to define the initial temperatures of all nodes except boundary nodes.  The user can uncheck the Use Current Temperature checkbox and enter a temperature in the Initial Temperature edit box.  If the Use Current Temperature checkbox is checked, temperatures presently assigned to the nodes will be used as the initial temperature for the nodes.  This is useful when it is desired to start a transient run with steady state temperatures as the initial temperatures.  The model can be run for steady state temperatures, the heat loads changed or removed, then the model run to determine cool down rate.

Use Current Temperatures

When this is checked, the existing nodal temperatures are used as the initial temperatures for the solution.

Preconditioner

The default settings are usually good.  But if you want to try to speed up the solution, try other settings.  See below for more.


Transient solution detailed example

The figure below is a model of a Gallium Arsenide (GaAs) RF amplifier.  The smallest dimension in this model is 0.03 microns.  The minimum thermal time constant is 1.56e-11 seconds.  To run this model for 0.002 seconds would require more than 128 million iterations using the finite difference solver.  This is an ideal application for the GCG solver.
graphic
graphic

The model was run for nine different settings in the GCG dialog.  The three Preconditioner options were tried and the Termination Criteria was tightened.  A single node at the heat source was plotted with time and the time to solve the transient was recorded.  The following table summarizes the results.  The temperature plot of the nine cases is shown in the figure below.  It is easily seen that, with the incorrect settings, very incorrect answers can be calculated.  It is therefore very important to verify that the solution is correct.  For a certain type of problem, once proper settings have been determined, they can be used later, on other similar problems without having to do this exercise. 

For this case, the default of five sub iterations for the SOR preconditioner is best.  Changing to two or ten sub iterations resulted in worse results.  As would be expected, the termination criteria are of most importance.  From the first order look, adding a single decimal place to the termination criteria improves the results significantly.  On looking closer though, see the figure below, there are oscillations in the solution.  The SOR and Jacobi give about the same answer for a termination criteria of 0.0001 but the Jacobi solves more than a minute faster.  The Jacobi preconditioner with a termination criteria of 0.000001 solves in about twelve minutes and provides the best answers. 

Termination criteria                                                                                            
Case    MAX(Abs(R))    ABD(Sum(R))    Time to solve    Name               Pre Conditioner
                                                             (seconds)         
1          0.001                0.01                  138                   0.001                SOR 5 sub it.
2          0.0001              0.001                264                   0.0001              SOR 5 sub it.
3          0.00001            0.0001              694                   0.00001            SOR 5 sub it.                
4          0.001                0.01                  208                   0.001                SOR 10 SOR 10 sub it.
5          0.001                0.01                  141                   SOR 2              SOR 2 sub it.     
6          0.001                0.01                  137                   Jacobi 0.001     Jacobi
7          0.0001              0.001                189                   Jacobi 0.0001   Jacobi
8          0.00001            0.0001              335                   Jacobi 0.00001  Jacobi  
9          0.000001          0.00001            719                   Jacobi 0.000001 Jacobi

graphic