Linearization gives accurate results with Simulink Control Design but not with linmod

I am using matlab r2010a. I have performed linearization of my simulink model using Simulink Control design, and using the matlab command linmod. The results I obtain from linmod are absolutely inaccurate. With simulink control deisgn, i obtain an underdamped step response as a result of linearization, while with linmod i get an unstable system going to negative infinity. What could be the reason for this unusual result?

[Guy Rouleau]

If your system is very complex, it is possible that SCD can do something that linmod cannot. However there could be multiple reasons. It is not possible to answer your question without the model and your linearization code. For example, are you sure you have the same operating point?

I recommend contacting technical support.

Answer 1

[Erman Korkut]

The underlying reason can be a lot of things for the difference:

1. If you have more than one input/output points, the channels of the results might be in different order and you might be comparing different channels.
2. If you have some blocks that specify their linearization, linmod/dlinmod would ignore these whereas Simulink Control Design linearization honors them.
3. If you have delay blocks in your model, you could be using exact delay models in SCD linearization whereas linmod/dlinmod cannot do that and uses Pade approximation instead.
4. If your model is multi rate, Simulink Control Design performs some postprocessing to handle the rate conversions better as explained in this demo:

http://www.mathworks.com/products/simcontrol/demos.html?file=/products/demos/shipping/slcontrol/scdmrate_codepad.html