Hello

check below link:

https://de.mathworks.com/help/ident/ref/tfest.html

Hello Dear Ali

I tried use this command in matlab but returned result is very bad and don't match with my data.if you check attached file you seen I use tfest but!!!

Ok , Asghar  I will try running you code and see ! Only catch I need some time If it ok  with you.

Ok Dear Aparna I waiting for your feedback, Thanks a lot.

Well  Asghar , here you go : Changes in the Calculate_z.m

opts = tfestOptions('SearchMethod','grad','Display','on');

opts.SearchOption.MaxIter=30; or 40 and so on

First the snapshots of the results . {A1,A2}

Next: as far as the ('WeightingFilter'  , 'inv') goes this has no field by that name in the datastructure "opts" (snaptshot  attached) , so I can't set any of the  fields in the opts .

I guess , you have to look at which option can be used for 'inv'

opts.SearchOption.Advanced are all numeric fields not strings.

Try this in the commandline .

Hello Dear Aparna,

Weighting Filter  added in matlab 2017a ind i use this option for improve my result but also i dont know how can i select number of poles and zeros! what's you'r idea for select number of poles and zeros?

I attached my new code for you if its possible for you please read this link and give me some feed back.thanks alot

https://de.mathworks.com/help/ident/ref/tfestoptions.html?searchHighlight=tfestOption&s_tid=doc_srchtitle

Dear Asghar , will go through and let you know in a day or two.  For poles and zeros , you seem to have chosen : opt.InitOption.N4Weight='auto';  then for SVD to have Poles-Zeros some options needs to be looked into ...

Dear Aparna, I really do not know where is my problem, I have read MATLAB Guide, Link that I sent to you, Writing to data in the frequency domain use ' EnforceStability '  option but when i use this option result is very poor and do not match with my data, and writing' InitMethod' applicable only for estimation of continuous time transfer function using time domain data!!

I just read the tfest , you have the 9 there !!! You change it to any # , the # of poles will change accordingly Asghar. The tfest always takes (p ) # of poles  and (p-1) # of zeros . That is the standard the degree of the denominator has to be > that numerator .

Ok we will try this , is the data in the time domain ?

If yes then get the transfer function by tfest(data,np) , here np is say is some arbitrary # or if you want take 9 .  Prepare the data in the format it (tfest)  wants .

Next , use all the options and other stuff get the new transfer function and check . The  raw vs simulated with options.

forget about InitMethod it is not of much use to us.

try opt.Focus='stability' for the 2nd tfest

ok, my data is the frequency domain, I start doing your suggest and give back some feedback

 Thanks my dear friend. 

Progress !!! any direction Asghar ...