| T O P I C R E V I E W |
| Wojtek. |
Posted - 01/29/2015 : 05:10:18 AM
Origin Ver. and Service Release (Select Help-->About Origin): OriginPro 2015 32-bit Srl b9.2.257 evaluation
Operating System: Win7 32-bit
Hi,
I have some problems with nonlinear implicit curve fitting. I have dark measurement data of solar cell (in that case it is a normal pn diode) and I would like to fit these data with Shockley equation taking into account the series and shunt resistances. I have define my fitting function:
f = Is*(exp((q*(V-I*Rs))/(n*k*T))-1)+((V-I*Rs)/Rsh)-I, where Is is a saturation current, V, I are voltage and current, n - ideality factor of diode, Rs and Rsh are series and shunt resistances, T - temperature and k,q are constants.
as a result of fitting I obtained results as follow:
Fitting curve looks quite goods, but if I zoom it I observe something like that:
The fitting curve seems to "jumping" on my experimental data. Moreover if I plot the current in log scale I obtain another different plot:
Do you have any ideas why my fitting curve does not fit perfectly with experimental data. I think that my equation is correct. The R^2 parameter is almost 1.
Regards,
Wojtek |
| 5 L A T E S T R E P L I E S (Newest First) |
| Wojtek. |
Posted - 02/09/2015 : 06:07:07 AM
Hi Christy,
thanks you for help.
I have obtain the same results as you:
when I selected Data Type of X: Fitted Point for Input Data in Setting: Fitted and Initialize parameters (which are my guesses) using button marked as 1. Then I click Fit button and I obtain this weird graph and extremely strange values of parameters:
So I started fitting again (using my standard procedure, which I used to use for fitting data but with explicit equations): I initialize all parameters, fixed all of them and perform iterations until converged. Then I deselect one parameter and again perform iterations until fit converged, then I deselect second parameter and perform iterations until fit converged and so on... It works so far, but in this case (implicit equation) it does not work...
To sum up, everything works until I perform iterations to determine the fitting parameters.
Regards,
Wojtek |
| Wojtek. |
Posted - 02/03/2015 : 04:13:57 AM
Hi Christy,
thanks you a lot for help.
Your fitting result look good (according to your last posted graph), but take a look on the fitting parameters which you obtained. The Is parameter (saturation current of pn diode) is extralegally small and negative, moreover it looks like you fixed the values of others parameters (n, Rs, Rsh and T) because their standard error is equal to 0.
Note, that for all fixed parameters (after performing even 1 iteration) the fitting curve perfectly fit to the data, but the fitting parameters are initial (my guess values).
I plotted the I-V curve again, changed current axis to log and started the fitting procedure from the beginning, then I changed Data Type of X: from Uniform linear to Fitted Point for Input Data. I fixed all fitting parameters and I obtain this graph (note that all standards errors are equal to 0):
Next, I deselect Is parameter, perform fitting and obtain this graph:
then I deselect Rs and Rsh parameters and obtain this weird graph and message: Fit did not converge - reason unknown.
So, I have setting which you recommended and there is still a problem with fitting |
| juliopjbn |
Posted - 02/03/2015 : 01:07:16 AM
Thank you for sharing
|
| Wojtek. |
Posted - 02/02/2015 : 03:36:38 AM
Hi Christy,
thanks you for help.
As you posted, after some changes in Settings the ripple problem dissapear:
But if I change the current axis from linear to log, some discrepancy between experimetal data and fitting curve will appear:
It even happens when I started fitting procedure from the beginning with initial parameters obtained from fitting in linear scale of current (first figure).
Do you have any idea what is wrong with my fitting procedure ? |
| Wojtek. |
Posted - 01/29/2015 : 06:39:25 AM
Hi Christy,
I have tried connect your FTP, but it seems to be offline. I have also tried send the opj file using General Contact Form, but the button for file attachment is inactive...
May I send the opj file to you directly by e-mail ?
Wojtek
EDIT:
I sended the file Wojtek_opj file.zip using General Inquiry Form. |
|
|
