| Author |
 Topic  |
|
|
TheNew
Thailand
4 Posts |
 Posted - 06/15/2017 : 03:53:40 AM
|
Origin Ver. and Service Release (Select Help-->About Origin): originpro 64-bit 9.0
Operating System:windows 10
I tried to use non linear implicit fitting to fit the modified diode equation like this
f = Is*(exp((q*(V-I*Rs))/(n*k*T))-1)+((V-I*Rs)/Rsh)-I
T : temperature
Is : saturation current
Rs : series resistance
Rsh : shunt resistance
q : electron charge
I : current
V : bias voltage
the fitting is converged. The graph looks nice in the whole picture.
In positive region it seem perfectly fit.
But when I observed in negative region, It seem the fitting is under estimated compared to data.
When I change to log scale, the huge difference occur in negative region.
How can I solve this problem.
Thank you
|
Edited by - TheNew on 06/15/2017 06:32:40 AM |
|
|
arstern
USA
237 Posts |
Posted - 06/15/2017 : 09:37:14 AM
|
Hi
You can set the initial y value to be fixed. With the NLFIT() dialog box open select the parameters tab and input your initial y value. Make sure to select fixed. Also make sure to select the tab highlighted in the figure that has the forward sign. This performs multiple iterations until fit converges.
If this does not work, could you please provide the values you are inputting for your constants? And if your parameters are set within a range, please provide that also.
Thanks,
Aviel
OriginLab |
 |
|
|
TheNew
Thailand
4 Posts |
Posted - 06/15/2017 : 11:39:10 AM
|
Thanks you for helping, Aviel. But I still have the problem when I followed your suggestion.
I tried to set Is (saturated current) to be fixed at 1.25E-11 and performed iteration. I got weird result. The graph cannot fit in threshold region.
And in negative region, it still underestimate.
Also in log scale
Here, I attached my setting on the builder.
Thanks you
Mongkol |
|
|
|
arstern
USA
237 Posts |
|
|
TheNew
Thailand
4 Posts |
Posted - 06/16/2017 : 02:18:28 AM
|
Thanks you Aviel,
I followed the link you gave by using "Data Type of X: Fitted Point for Input Data" but it still can't be fitted
So I attached the opj file to tech@originlab.com
Hope you can help
Thanks you
Mongkol Sapankaew |
|
|
|
arstern
USA
237 Posts |
Posted - 06/16/2017 : 4:33:13 PM
|
Hi Mongkol,
To shift the fitting curve upward so that it fits your data, you will need to add a y initial (Io) into your equation.
-Set Io parameter to be 1E-11.
-I also changed your n value to be less than what you had input into your equation.
-For the constants, I input the same values that you show in your figure.
-Also Select Fitted Curve and set Data Type: Fitted Point for Input Data
OriginLab
Here you can see that the fitted curve now is much more aligned with your data.
Figure with y-axis log scale
Hope this helps,
Aviel
OriginLab |
|
|
|
TheNew
Thailand
4 Posts |
Posted - 06/17/2017 : 06:11:53 AM
|
Thanks you, Aviel. Your solution help me so much. Now I can fit the curve with my data.
Thanks again
Mongkol |
|
|
|
yuki_wu
896 Posts |
Posted - 01/01/2019 : 9:34:05 PM
|
Hey guys,
In Origin 2019, we have a new app Fitting Function Library, in which you could search and find the fitting functions you desired:
https://www.originlab.com/fileExchange/details.aspx?fid=490
so you don't need to build a user-defined fitting function for some general equations by yourself.
Regards,
Yuki
OriginLab
|
|
|
| |
Topic |
|
Problem with nonlinear implicit fitting on diode