NI Circuit Design Community Blog

Community Browser
Labels
cancel
Showing results for 
Search instead for 
Did you mean: 

Temperature Variation - Test Your Models

OStep
NI Employee (retired)

Not all SPICE models are created equal.

This saying applies especially to a model’s dependence on temperature.

Multisim allows the user to set the operating temperature of the simulation (the default is 27°C).  Using the Temperature Sweep analysis, Multisim even allows the user to sweep across a range of temperature values, running an underlying simulation such as DC operating point or Transient  for each step in the sweep.  However performing such an operation only makes sense if the models of the critical components in your circuit are designed to support temperature variation.

Let’s consider two BJTs with Model A and Model B.

Transistors.png

Let’s configure Temperature Sweep analysis to run a sweep from -55°C to 125°C and monitor the effect on beta, which we setup using an expression with the format I(ProbeN)/I(IbN) in accordance with the above naming of the probes and current sources.

TempAnalysis.png

The following is the result for Model A:

ModelAResult.png

This beta variation with temperature matches the part’s datasheet reasonably well.

The following is a result for Model B:

ModelBResult.png

This result does not match the part’s datasheet, which shows a much greater variation in Beta. In this case, it can be concluded that the model does not support temperature variation.

Notice that a model’s lack of support for temperature variation does not necessary imply that the results produced by the model will not vary with temperature. Clearly, as shown by model B, the temperature may play a role in shaping the model’s curves. A model’s lack of temperature support implies that simulating the model using a temperature different from which it was designed for will likely lead to incorrect results.  Models that do not support temperature variation are typically designed for 27°C, the default setting in Multisim.

Therefore, it is good practice to run a simple, isolated test on the model and compare the results against expectations before assuming the model supports temperature variation and using the model in circuit-wide temperature analysis.

Max
National Instruments