CPE value greater than 1

[MatlabBeginner]

hi every one

i have a question about fitting a circuit to my data.

i use both software: "MEISP" and "Zview" but a common problem appears.

when i want to use constant phase element (CPE) in my circuit,actually its value must change between 0-1 . but the software may obtain value greater than 1.

( phi value in MEISPvalue and CPE-P value in Zview  may become >1)

in such cases i manually fix its value to for example .85 , .9 or other values and continue fitting procedure.

is there a way to get the value automatically between 0-1 and not grater then 1???

thank u all in advance

[Yevgen Barsukov]

In most Newton-method based fastest descent non-linear fitting routines it is problematic to put limits because they interfere with the logic of the optimization. I try using such limits in mathcad non-linear fit, and what happens in this cases is that value keeps changing until it reaches the limit, after which optimization stops. So it is the same as fixing the value at the limit in the first place.

Situation is somewhat different if you do "global optimization" for example using genetic algorithms. In this case you can have upper and lower limit for all parameters and these limits define the range where "parent" initial guesses are going to be generated. I am not aware of a genetic optimization based impedance fit program that is freely available, although there are research papers on it. Generally it will be much slower because it does not descent along fastest path to minimum as Newton methods, but it will be less likely to be stuck at a local minimum. This is usually less a concern if you have good initial guess, that you can get with pre-fit in MEISP.

Regards,
Yevgen

--
You received this message because you are subscribed to the Google Groups "Impedance Spectroscopy" group.
To unsubscribe from this group and stop receiving emails from it, send an email to impedance-spectro...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[MatlabBeginner]

Thank you Yevgen!

as i understand it doesn't have a solution

i have two choices:

use prefit prosedure (and hope it return a value smaller then one)

or

i should fix this value and go ahead

is it right? and what is your suggestion?

To unsubscribe from this group and stop receiving emails from it, send an email to impedance-spectroscopy+unsub...@googlegroups.com.

[MatlabBeginner]

hi

i need help

i cant use meisp to fit my data to my model

the value of CPE confused me!

i attach one of my data set and schematic of my model

i would be so thankful if you explain the procedure to obtain model parameters(specially CPE values)

thank you in advance!

[Søren Koch]

Well I couldn't open your model file, but I have fitted your data in elchemea analytical successfully using the following model

R: R=446.2
RQ: R=117, Y=7.189e-08, n=1
RQ: R=4.745e+04, Y=2.331e-07, n=0.8739

that is 1 series resistor and 2 RQ elements (RQ being a resistor in parallel with a constant phase element)

Unfortunately your data have some noise making it hard to get good start guesses even when using 'find values', however by manually finding some values which were 'good enough' the fitting converged to the above values.
I had to lock the first n value to 1 to avoid it getting above 1 (free fitting resulted in n=1.09)

I hope that you can use the result

Søren Koch

[Yevgen Barsukov]

MB,

your problem is that spectum that you have has very little information, it looks like it can be fited

well with a single R and CPE, but your cirquit is very complex so most of its parameters become undefined.

Even though fit is good, you cant get anything aout of it. Phi values are just becoming zero so your cpes

are just disappearing.

I would try with much simple circuit with single CPE, or even better, with a circuit that has physical meaning, for example

uses transmission line for your film that you are trying to describe, although it is going to be an inhomohogenous one, since there is no 45degree angle.

For example, if you just use single CPE, here is you get a good fit. It means that there is no additional information in this data,

although you could use a physically relevant circuit with the same number of time constants and same degree of distribution.

Regards,
Yevgen

--

You received this message because you are subscribed to the Google Groups "Impedance Spectroscopy" group.

To unsubscribe from this group and stop receiving emails from it, send an email to impedance-spectro...@googlegroups.com.