Evaluate Project - the toolbar button

This function is available through menu items (right-click on the project name or choose the Evaluate menu in the menu bar) or through a toolbar button. This icon is used for similar processing functions of other Windows programs using multi-document interface, such as compiling the program source code and running the target executable in a C compiler.

In case of the T3Ster-Master program the button starts the run-time environment of the Transient Evaluation Engine using the NID method. First in the Evaluate window (see Figure 4-11) the project to be evaluated can be selected using the radio buttons under the Select project item. Under that, one can enter the Evolution Parameters for the evaluation engine:

Resolution (points/decade):	a value between 10 and 100 can be specified
Bayes iteration number:	a value between 128 and 65536 can be specified
Resolution tint-> Foster:	a value between 1 and 4 can be specified.

The first parameter determines how many data points are used in a decade during the evaluation process. If the resolution in the .rec file is different, interpolation is used to adjust to the resolution given here for the evaluation.

Figure 4-11: Specification of the Evaluation parameters in the Evaluation window

For imported simulated thermal transients if the data resolution of the simulation is between 10 and 100 points/decade the resolution specified in the import wizard should be the same so that no interpolation occurs. This way the evaluation uses the data points yielded by the simulation.

Note: Noise is present even in case of simulated transients with high data density – this noise is numerical noise (round-off errors, finite accuracy of PDE solver algorithms).

Figure 4‑12 shows the cumulative structure functions (see Section 3.3.8) of two measurements in the same setup, evaluated by 512 and 20000 iterations (Green.0001 and Green.0002, respectively). With higher iteration number more details of the structure can be seen. However, noise on the measurement can cause artifacts, false “steps” in the function.

Generally we recommend 512 to 2000 iterations for typical measurements. Simulated transients can be evaluated by 10000 to 40000 iterations.

Figure 4-12: Evaluation of two measurements with a Bayes iteration number of 512 and 20000

The third parameter Resolution tint-> Foster: is related to the way time constant spectra (Section 3.3.5) are produced.

Thermal structures can be represented by a behavioral model: a chain of parallel thermal RC elements. The evaluation procedure produces from the continuous time-constant spectrum or tau-intensity function a discretized one in equal intervals. By default it assigns a single RC stage to each discrete spectrum line. In this field is a reduction factor can be introduced for the discretization. E.g. value of 4 means a one-to-four reduction of discrete spectrum lines. For theoretical details of the evaluation procedure refer to Section 6.

The resolution and this reduction factor are not independent when time constant spectra are calculated. If high data density is specified then this reduction factor needs to be increased, otherwise serious numerical problems occur during structure function calculation. As a rule of thumb, if resolution above 40 points/decade is specified, this reduction factor usually has to be increased.
Evaluation is started when you press the Start button in the window. The parameters introduced in the input fields are forwarded to the evaluation engine's executable as command line parameters. The standard output of the evaluation engine is directed to the Evaluate window (Figure 4‑13) and it is also saved as a log file. When the evaluation is completed, one can browse this log in the Evaluate window with the sliders on the right hand side (Figure 4‑14). Press the Close button to complete the evaluation and to close the window.

Figure 4-13: The Evaluate window while the evaluation engine is running in the background

Figure 4-14: The Evaluate window when the evaluation is completed

Note that the implementations of the NID method in T3Ster-Master's results evaluation engine and in THERMODEL are different. Here Bayes iteration is used for the deconvolution while in THERMODEL this step is performed by Fourier inverse filtering. That is why – due to the different features of these algorithms – the same type of functions (such as e.g. time-constant spectra or structure functions) calculated by the different programs may show minor differences.

By selecting the appropriate checkboxes the following additional operations can be carried out:                                
                         

Selecting the Model generation checkbox a one-dimensional compact model of the device + measurement environment system will be produced (Figure 4-15). This is constructed of ladder-type RC stages, like the structure function, but the number of stages is reduced to the value specified in the field next to the checkbox.

The compact model is stored in SPICE subcircuit format (Figure 4-16). This format can be invoked by the FLOTHERM simulation program, in its Network Assembly.

Figure 4-15: Evaluation of two measurements, compact models of 3 and 6 stages generated (Green.0001 and Red.0001, respectively)

Figure 4-16: SPICE style subcircuit netlist of the 3 stages

In many cases besides the sample to be measured there are fixtures, clamps etc. adding a parallel heat conductance path to the actual system. It is relatively easy to make an estimation on the parallel path, removing the sample from the fixture one can measure the (usually high) thermal resistance of the holder.

The Correction with ║ Rth field is used to enter the estimated total thermal resistance of the parallel path (Figure 4-17, Figure 4-18).

Figure 4-17: Correction of the evaluation (parallel path of 96 K/W)

Figure 4-18: Structure functions with and without the correction of the parallel path