Inverse Laplace transform of discrete numerical data input

File Exchange offers several procedures for Inverse Laplace Transfrom but - if I am not mistaken - all are for the case when F(s) is provided in an analitic form. This actually restricts their use to the input data in a noise-free form, close to 64 bits. Adding noise - which is always present in experimental data - usually makes Inverse Transfrom unstable even at quite high S/N ratio. Are there any published Matlab procedures to overcome the damaging presence of noise?

P.S. Historically, Povencher published a program, Contin, to perform this task, and there was (I can't recall the link now) a Matlab implementation but when I tried it with a 1:500 S/N ratio it failed fro me.

[Paul]

Hi Andrei,

If F(s) is provided in analytic form, then how can that be at all related to input data, where I'm interpreting data to mean numerical data.

In any case, if you have input data and the corresponding output data, in either frequency or time domain, for an LTI system, then the System Identification Toolbox might offer functionality your're looking for, e.g., tfest among many others. Also tfestimate in the Signal Processing Toolbox may be of interest.

Also, consider attaching some example data in a .mat file using the Paperclip icon on the Insert menu and posting the code being used to process the data so others can get an understanding of what you're trying to accomplish and perhaps provide some help.

[Andrei Dioumaev]

Hi Paul,

Sorry for the delay.

Yes, your first paragraph correctly identified my problem, that the available on the File Exchange procedures for the Inverse Laplace Transfrom focus on the case, when F(s) is provided in analytic form. However, I was asking for their equivalents that might be applied to (experimental) input data in numerical form which is discrete and potentiually noisy.

[Paul]

Again, it would be easier for someone to help if you post a simple example of the data you have that describes F(s) and explain what you'd like to do with that data as the end goal is not clear to me.