RE: LEVM/LEVMW -Version 8.11 for DRT-ii
Macdonald, James R
You should ask yourself whether your impedance-spectroscopy frequency-response data extends over a wide range and shows much structure, including possibly several peaks. If so, perhaps 19 points for a DRT would indeed be insufficient. You might find the attached paper of interest. The method of Tuncer, illustrated there, is able to provide a DRT estimate with many more points than 19. If you find that your data suggest the need for more, you might be able to get him to help you: he controls and uses his method, and I do not have it available. He is currently at GE in New York state.
Ross Macdonald
------------------------------------------------------------------
Dr. J. Ross Macdonald is William R. Kenan Jr. Professor of Physics, Emeritus; Department of Physics and Astronomy of the University of North Carolina at Chapel Hill, and a member of NAS and NAE. Website: http://jross macdonald.com
________________________________________
From: impedance-s...@googlegroups.com [impedance-s...@googlegroups.com] on behalf of Taesik [taes...@googlemail.com]
Sent: Thursday, May 17, 2012 2:00 PM
To: Impedance Spectroscopy
Subject: Re: LEVM/LEVMW -Version 8.11 for DRT
Dear Prof. Macdonald,
It is my honor to have a word with you.
In fig 4 of this Journal of the electrochemical society paper (http://
dx.doi.org/10.1149/1.2801372), one can find a distribution function
plot.
The plot is obtained from nonparametric method (Tikhonov
regularization).
Do you think 19 elements will be enough to handle this kind of
impedance data?
In fig 4, they have more or less 10 points per a peak, and there are
five peaks shown.
I just thought it would be nice to have some number close to that from
parametric approach.
In my opinion, parametric approach is more straightforward involving
less mathematical subtlety than nonparametric one does.
Hopefully this can inspire some gifted Fortran programmers in the
group since I am totally lost about how to do the recompiling.
I wonder if there is any inherent programming difficulties regarding
this issue?
Thank you for your kind attention and time.
On May 17, 5:56 am, "Macdonald, James R" <m...@email.unc.edu> wrote:
> To change Circuit R in LEVMW to allow determination of more than 19 relaxation times you would have to have available a Fortran Compiler program, and the task would be onerous. I do not reccoment that route. It is not clear to me why one would need more than 19 relaxation times; I have usually found that a maximum of 9 to 13 well defines the distribution.
>
> You might find the examples of DRT determination in the FITTESTS file of LEVM instructive. You might also benefit from reading my papers #'s 184, 206, and 220, which you can freely read and download from my website,http://jrossmacdonald.com.
>
> Good luck,
>
> Ross M.
>
> ------------------------------------------------------------------
> Dr. J. Ross Macdonald is William R. Kenan Jr. Professor of Physics, Emeritus; Department of Physics and Astronomy of the University of North Carolina at Chapel Hill, and a member of NAS and NAE. Website:http://jrossmacdonald.com
>
> ________________________________________
> From: impedance-s...@googlegroups.com [impedance-s...@googlegroups.com] on behalf of Taesik [taesik...@googlemail.com]
> Sent: Wednesday, May 16, 2012 8:08 PM
> To: Impedance Spectroscopy
> Subject: LEVM/LEVMW -Version 8.11 for DRT
>
> Hi, all,
>
> I am trying to calculate the distribution of relaxation times from
> Circuit R in LEVMW.
>
> From the manual, it is clear that I can have up to 19 Voigt / Maxwell
> elements.
>
> However, it also says there is a way to increase this number by re-
> compiling(?).
>
> I am not very familiar with computer programming; thus, would anyone
> be able to explain how I can increase this number?
>
> I hope that this is not too complicated.
>
> Thank you for your attention.
>
> Best regards,
Taesik Oh
I attached the reference paper.
I will contact Dr. Tuncer, but his method seems much more complicated to me. (I never did a Monte Carlo calculation)
One question regarding LEVMW.
There is an option to use Tikhonov regularization in your LEVMW. (p(33) is the regularization parameter)
I tried RTST303.REG and RTST303.NOR, and LEVMW returns only 11 time constants for both.
If your code already contains Tikhonov regularization, then can I use LEVMW to do nonparametric analysis as in Fig 4 in the attached paper?
I have no idea on how actual programming goes, but I believe this should be already possible using your LEVM code.
I know that you prefer parametric approach from your journal of computational physics paper in 2000, and I agree with you on that.
However, it will be nice for LEVMW to handle nonparametric approach also.
I think that will make LEVMW even more valuable.
Thank you very much for your kind attention.
Best regards,