| Author |
 Topic  |
|
|
Boubera
Portugal
2 Posts |
 Posted - 01/27/2016 : 1:58:37 PM
|
Hello
I defined a curve fitting with the following equation:
y = FthN + (Fth1-FthN)*exp(-k*(x-1))
So I need to compute FthN, Fth1, k.
I used the following parameter settings:
InitialValues = --(V),--(V),--(V)
Meanings = saturation threshold,single-shot threshold,accumulation factor
LowerBounds = 0.000000(X, On),0.000000(X, On),0.000000(X, On)
UpperBounds = --(X, Off),--(X, Off),--(X, Off)
NamingMethod = User-Defined
NumberOfSignificantDigits = -1,-1,-1
and parameter initialization:
Fth1 =max(y_data);
FthN = min(y_data);
and the Function form is Origin C.
This fitting fits pretty fine a first set of X and Y values I have and if I introduce the Y_err values, it keeps working fine. However, for a second set of XY values (shown in the image below), the fitting does not converge to the points IF I plot the values with the the Y-error/uncertainty (which is of about 5 to 8%). The final fitting of these values is the same as that of after one iteration.
Can anyone help me find the problem?
Thanks in advance! |
|
|
SeanMao
China
288 Posts |
Posted - 01/28/2016 : 03:21:28 AM
|
Hi,
I did a sample fit using the function you mentioned and there is no problem when I fit data with Y errors.
Would you mind sending your OPJ file to tech@originlab.com so that I can try to reproduce your problem on my side?
Regards!
Sean
OriginLab Tech. |
 |
|
|
SeanMao
China
288 Posts |
Posted - 01/29/2016 : 02:20:07 AM
|
Hi,
The reason for this is that you have fewer data points on the start part and the weighting of the start part is smaller compared with the tail part where the error bars are smaller, so that the start part will be less considered during fitting.
The solution is to use better initial values, we have deduced a set a initial values using built-in ExpDec1 fitting function and after inserted into your case, it worked well now.
Fth1 = 0.606
FthN=0.29
k = 0.12
Regards!
Sean |
|
|
|
Boubera
Portugal
2 Posts |
Posted - 01/29/2016 : 09:21:50 AM
|
Thanks a lot Sean, it worked!
Regards |
|
|
| |
Topic |
|
|
|
user defined non linear curve fitting