Re: Not able to get the matlab equivalent output from tfestimate function (transfer function estimate)

[Mike Miller]

On Thu, Feb 14, 2013 at 9:55 AM, sarbjit wrote:
> I have already submitted a bug "http://savannah.gnu.org/bugs/?38303";, but
> unfortunately I need to use this piece of code at earliest and there is no
> response as of know. I looked at the tfestimate.m and found it is calling
> pwelch.m, I tried reading the file but could not found a way to make it
> working.

Ok, sorry I missed that it was up there already.

> Anyways as as far as data is concerned, I am attaching matlab workspace for
> my portion of code which is having "Txy", I loaded it in Octave and found
> the following results :-
>
> octave-3.6.2.exe:11> max(W)
> ans =  3.1416
> octave-3.6.2.exe:12> max(txy)
> ans =  1.00000

Yes, if this is your data computed in matlab, then my point is that
Octave tfestimate is working and you are essentially getting the same
results. Matlab's auto-scaling is deceiving you. If you don't see what
I'm saying, try zooming your y-axis scale out in matlab, or set the
scale explicitly with

axis([0, pi, 0.8, 1.2])

Or plot the log of the transfer function

plot(W, log10(txy))

and you'll see that what you are calling variations in matlab are negligible.

-- 
mike