| Author |
 Topic  |
|
|
Yan_Mikoyan
Ukraine
7 Posts |
 Posted - 06/23/2022 : 1:53:05 PM
|
OriginPro 2022, Win 7
Please tell me how to solve the problem. I have a long time series (over 100000 points) with a lot of (about 50) pulses. I want to approximate each pulse with by two decaying exponents and then calculate the average of the rise and fall constants.
Can this action be done in one go? Personally, I don't think it's possible. I have another idea. It is possible to (1) split the entire time series into a large number of columns, where each column will contain one pulse, and (2) apply the "sequential fit" + "piecewise fit" applications. How to optimally perform partitioning (1)
and are the fitting operations in (2) compatible? |
Edited by - Yan_Mikoyan on 06/23/2022 2:03:33 PM |
|
|
YimingChen
1652 Posts |
Posted - 06/23/2022 : 3:54:57 PM
|
You can probably use the Gadget Replicator app.
https://www.originlab.com/FileExchange/details.aspx?fid=359
1. Define your piece-wise fitting function.
2. Plot out the time series.
3. Apply the Quick Fit gadget (Menu Gadgets: Quick Fit: Open Dialog...) on the plot using the defined fitting function.
4. Run the Gadget Replicator app, the app can run the curve-fitting on multiple data ranges. You have to determine the range of each peak and save the peak-begin and peak-end values into two columns as the input of the app.
James |
 |
|
|
YimingChen
1652 Posts |
Posted - 06/23/2022 : 4:10:30 PM
|
Also, would you mind sharing your data here for us to check?
|
|
|
|
Yan_Mikoyan
Ukraine
7 Posts |
Posted - 06/24/2022 : 03:59:41 AM
|
Thanks for the advice. I overlooked the Gadget Replicator app. But question (2) remains. Each individual pulse is approximated by a piecewise shape function: Constant-expdec1-expdec1 in the Piecewise Fit app. But this function is not in the main library, which can be accessed from the Quick Fit gadget. Does this mean that it has to be created manually, since there is no way to combine the use of the Piecewise Fit app and the main program?
The attached file shows part of the signal on which I create the processing algorithm.
https://my.originlab.com/ftp/forum_and_kbase/Images/Probe%20TS.opju |
|
|
|
YimingChen
1652 Posts |
Posted - 06/29/2022 : 5:28:05 PM
|
Hi,
Thank you for sharing the data. We checked it and Origin doesn't have the analysis to detect the pulses and calculate the rising time and decaying time. We will make an app (APPS-1188) to do this analysis. Thank you.
James |
|
|
|
YimingChen
1652 Posts |
Posted - 06/30/2022 : 09:01:44 AM
|
By the way, you can try the fitting function ExpGrowDec or ExpGrow3Dec2 under the Exponential category. Thank you.
James |
|
|
|
Yan_Mikoyan
Ukraine
7 Posts |
Posted - 06/30/2022 : 12:40:19 PM
|
| Hi! I am very touched by your attention to my particular problem. Your advice helped me move forward in solving it, but new nuances were discovered. The Gadget Replicator app does not work well with long strings. Firstly, it is very difficult to determine with the necessary accuracy the base time interval on which the approximation is carried out, therefore, by the end of the series, an inevitable “desynchronization” occurs. Secondly, in the experimental sequence of pulses, the distance between them is usually not constant, which makes it difficult to accurately indicate the working interval. If you intend to create a special application for pulse train analysis, then it is highly desirable to take into account the features of real experimental data. |
|
|
|
YimingChen
1652 Posts |
Posted - 06/30/2022 : 2:55:57 PM
|
Yes, we need to build an app that can detect the location of each pulse and determine its left/right bounds. We will let you know when the app comes out. Thank you.
James |
|
|
| |
Topic |
|
Time series analysis