How can i plot frequency response from 25000 time domain Data?

Hello every body, I have near 25000 time domain data, they are from a idustrial plant, I have both input and output data. I want plot frequency response by using this time domain data, please help me how it is possible? Thanks alot.

[dpb]

What's your end objective? When one has to ask the question, I suggest need some background first...

A list of references using Matlab for illustrative purposes is at System Identification Bookss; it appears the 2nd and 3rd would likely be apropos to your query as likely are several others...

[mostafa]

Hello dpb, my end object is select best sampling time from frequency response and band width. My first sampling time is 25 seconds, but i think this not good sampling time, i have plan to plot frequency response by time domain datas using input and output datas. I was trying using DFT, but i cant use this method carfully in matlab, next i saw this book "The Scientist and Engineer's Guide to Digital Signal Processing, Second Edition, by Steven W. Smith" in page 171 i saw that after multiple discrete time domain data by Hamming window, earn a new data as "windowed signal", after by DFT produce frequency response. I dont know what is Windowed signal, and why multiple time domain signal to Hamming window? Please help me how can i do it?

[dpb]

If you've already sampled at 25 second intervals, you have no valid frequency content above 0.02 Hz by the Nyquist sampling theorem (you do know what that is, at least, I hope?) The problem in determining a "best" sampling time is that you've got to ensure you've sample at a rate at least greater than twice the maximum frequency content possibly of interest. And, of course, you need to ensure that there isn't frequency content above that in the sampled data that is being aliased into that sampling range and that if so, is absolutely not removable by any means whatsoever once the data have been sampled.

I'm reluctant to suggest you just blindly apply FFT techniques to these data until you have a better understanding of the problems; the consequences of doing so are potentially too serious.

Is this an academic or commercial exercise? If academic consult your advisor, if commercial is there some access to consultant in the area you can confer with? This is a Matlab programming forum, not a consulting service, unfortunately.

[mostafa]

Hello , dbp and thanks alot for your attention. This is an academic execerse and my advisor dont want help me about this and fully reject to me!!!!!!!!!!! I am trying use matlab for this purpose, by many command such as "spa, fft and so on" and toolbox, such as "System identification toolbox", but any of them dont reply me good response then i am trying this methode by myself using m-file. Before i was trying use DFT methode and hamming window, but i think i didnt use them carefully, Can you help me abviously which methode can i use for improve my work and if matlab has better command than which one i said, please tell me for use. Best regards.

[dpb]

I think you're barking up the wrong tree first here; you need to learn something about signal processing and data collection basics first to ensure your data have any meaning before just applying tools. Perhaps the reason the results aren't what you expect is that the data aren't any good...

But, to see what frequency content is contained in the sampled data, there's a very good, concise example of an amplitude spectrum computed via the fft at

doc fft

The Signal Processing toolbox also has the function tfestimate for transfer function estimation and the spectrum function which has multitude of options for estimating spectra.

Again, however, unless there's something actually in the dataset you've collected, none of these can fix corrupted data obtained without adequate protection during that process, whatever it was.

[mostafa]

Thanks alot Sir...

Answer 1

[Image Analyst]

Look up pwelch() in the Signal Processing Toolbox: http://www.mathworks.com/products/signal/features.html#spectral-analysis

[mostafa]

Hello Image analyst, Thank you so much for your attention.

[mostafa]

Hello image analyst, I dont know PSD and welch theory, please if it is possible introduce to me quick refrence for it and use it in matlab. Thaks alot.