Thanks a lot, it helped me to give me a little heads up and save some time. However, I am not sure how to define my input data. I have e Step Response as output. (Angle) Do I use a single step as input? Some more background: I stimulate a muscle via electrical stimulation. The stimulation has 3 parameters: The current, the frequency and the pulse width. Out of those 3 only the pulse width changes. I fixed the pulse width on the maximum value to get the step response. Thanks a lot again and have a nice day!
Respected sir I have five values as input x= [5.42 10.62 15.13 20.41 25.23] and five values as output y=[0.05 0.09 0.15 0.25]. Can I get the transfer function from this data? What should I try for this- continuous time or discrete time inside transfer function? Thanks.
For the 1st question: No. 5 samples is too small a dataset for any meaningful identification of any system. In principle, you need a dataset large enough that it covers both the transient and steady state behavior of your system, sampled with a sufficient sampling rate ( 10 times higher than the rise time of the system). For the 2nd question: This is pretty much your choice, do you want a continuous-time or a discrete-time representation of the system? Note that you can convert from one form to another using MATLAB's functions 'c2d' and 'd2c'.
Hey Ventsislav, There's no reason, theoretical or practical, for this method not to work with 1st order transfer function. Could you explain to me what exactly is the problem you encountered?
@@h511345 My thoughts exactly. There should be no mathematical reason for this not to work. When I reproduce your experiment with 2-nd order transfer function it works exactly as intended. However when I try it with 1st order transfer function the identification comes up with different result.
System Identification - An Introduction : www.springer.com/gp/book/9780857295217 Also, System Identification toolbox MATLAB - Getting started guide: nl.mathworks.com/help/ident/getting-started-1.html
A concise and easy to follow tutorial...Thank you....you are contributing to the world.
A tip: watch series on Flixzone. I've been using it for watching loads of movies lately.
@Layton Tucker Yup, been watching on flixzone} for months myself :D
I watched this in a public PC. I signed in only to write my comment and say thank you!
This is a truly excellent lesson you've given. Thank you.
This is a good explanation. You should also add few thoughts on estimation data and separate testing data for completeness.
Thank you for a clear, informative video
You're welcome Seonghoon!
You saved me man you are my hero. Thanks a lot!!
the best video man. you saved us all.
The best tutorial ever. Thank you very much!
Thanks for tutorial. What is the command to get transfer function from frequency response?
Finally a useful video, thank you so much
Very clear and helpful, thanks!
Thanks a lot mate!! You've just saved me!
Cheers mate! Enjoy
Thank you sir. Very helpful.
Perfect tutorial!!! It was of great help.TFS!
Good work!! thanks, I really wanted this kind of help !! Thank you
Thanks for the positive feedback!
why l got error like this "The input and output must have the same number of row" ?
same, did u find the solution?
very good, thanks
nice tutorial
Thanks a lot, it helped me to give me a little heads up and save some time.
However, I am not sure how to define my input data. I have e Step Response as output. (Angle) Do I use a single step as input?
Some more background: I stimulate a muscle via electrical stimulation. The stimulation has 3 parameters: The current, the frequency and the pulse width. Out of those 3 only the pulse width changes. I fixed the pulse width on the maximum value to get the step response.
Thanks a lot again and have a nice day!
Thanks a lot mate!!
thanks a lot
Sir...how about if i have 4 input and 2 output?
Hello Hatim Mala! my input and output data don't have same number of row, how can i solve this problem?
Amazing !
hi, i made the same thing but the estimated tf was not like that i have although i entered correct number of poles and zeros
thanks alot mate
thanks goat
Respected sir I have five values as input x= [5.42 10.62 15.13 20.41 25.23] and five values as output y=[0.05 0.09 0.15 0.25]. Can I get the transfer function from this data? What should I try for this- continuous time or discrete time inside transfer function? Thanks.
For the 1st question: No. 5 samples is too small a dataset for any meaningful identification of any system. In principle, you need a dataset large enough that it covers both the transient and steady state behavior of your system, sampled with a sufficient sampling rate ( 10 times higher than the rise time of the system).
For the 2nd question: This is pretty much your choice, do you want a continuous-time or a discrete-time representation of the system? Note that you can convert from one form to another using MATLAB's functions 'c2d' and 'd2c'.
@@h511345 Thanks a lot Sir.
how can i design a mimo system using this app?
Is there a reason for this not to work with 1st order transfer function?
Hey Ventsislav,
There's no reason, theoretical or practical, for this method not to work with 1st order transfer function. Could you explain to me what exactly is the problem you encountered?
@@h511345 My thoughts exactly. There should be no mathematical reason for this not to work. When I reproduce your experiment with 2-nd order transfer function it works exactly as intended. However when I try it with 1st order transfer function the identification comes up with different result.
thanks
What is the name of the book that you recommend?
System Identification - An Introduction :
www.springer.com/gp/book/9780857295217
Also, System Identification toolbox MATLAB - Getting started guide:
nl.mathworks.com/help/ident/getting-started-1.html
Thank you!!!
mas ada tau code gelombang permukaan di laut yang bisa dimasukkan di matlab? kalau ada tau mas, silahkan kirim ke email nuraudina@gmail.com,
thanks