[BUG] Pake preview (and fit) not working #308

New Issue

2025-03-13T08:04:35Z

saeckech commented

2025-03-13 08:04:35 +00:00

Current behavior

In order to check D2O coupling parameters, I wanted to (manually) fit a Pake spectrum to partially frozen water data (appended 'data'-data). This worked fine some time ago (appended 'preview_old'-data and image). Today, trying to repeat everything with changed fit/preview parameters (eta=0 instead of 0.1), the preview does not appear and a fit (levenberg-marquard) gives the error message 'residuals not finite at the initial point'. Fitparameters were fixed or (reasonably?) restrained. Nelder-Mead and ODR calculate something upon fitting, but yield only NaN.

Version

2025-02-13

Expected behavior

Concisely, I expected a Pake-shaped something to appear

Steps to reproduce

No response

Log messages

13/03/2025 08:09:52 - ERROR - root : Residuals are not finite in the initial point.
NoneType: None

Anything else?

the python-2 version of the auswerteprogram at Mathilda's computer manages a preview, but the preview looks weird. I'm guessing it has too few x-values.
I tried fits and preview with and without excluding the narrowed line of the spectrum

### Current behavior In order to check D2O coupling parameters, I wanted to (manually) fit a Pake spectrum to partially frozen water data (appended 'data'-data). This worked fine some time ago (appended 'preview_old'-data and image). Today, trying to repeat everything with changed fit/preview parameters (eta=0 instead of 0.1), the preview does not appear and a fit (levenberg-marquard) gives the error message 'residuals not finite at the initial point'. Fitparameters were fixed or (reasonably?) restrained. Nelder-Mead and ODR calculate something upon fitting, but yield only NaN. ### Version 2025-02-13 ### Expected behavior Concisely, I expected a Pake-shaped something to appear ### Steps to reproduce _No response_ ### Log messages ```shell 13/03/2025 08:09:52 - ERROR - root : Residuals are not finite in the initial point. NoneType: None ``` ### Anything else? the python-2 version of the auswerteprogram at Mathilda's computer manages a preview, but the preview looks weird. I'm guessing it has too few x-values. I tried fits and preview with and without excluding the narrowed line of the spectrum

Data.txt

6.8 KiB

pNIPAM-CN0E_2H-NMR_Spectrum-SE-14dB_250K_2024-01-09T1858_PTS46704500Hz.png

13 KiB

Preview_old.txt

3.4 KiB

pNIPAM-CN0E_2H-NMR_Spectrum-SE-14dB_250K_2024-01-09T1858_PTS46704500Hz.png

saeckech added the Kind/Bug label 2025-03-13 08:04:35 +00:00

dominik was assigned by saeckech

2025-03-13 08:04:35 +00:00

dominik commented

2025-03-13 09:00:49 +00:00

Hm, works for me. Most often a problem of ill-chosen start parameters and bounds. What did you use as initial values?

saeckech commented

2025-03-13 12:53:17 +00:00

I used the values which worked previously for A, delta, eta, Sigma_B and t_pulse. I also tried different values for A over several orders of magnitude, eta=0.0 and Sigma_B=1,0.1 and tried out both t_pulse=2.5 and t_pulse=2.5e-6.

dominik commented

2025-03-13 16:44:29 +00:00

I see ze problem: Sigma_B is too small, anything below a couple 100 Hz breaks the calculation.

saeckech commented

2025-03-13 16:46:07 +00:00

also solves the problem for me :)
is it a pake- / histogram-specific issue, or could it happen for normal functions as well?

also solves the problem for me :) is it a pake- / histogram-specific issue, or could it happen for normal functions as well?

dominik commented

2025-03-13 16:58:12 +00:00

This one is due to Pake weirdness.

Sign in to join this conversation.

2 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: IPKM/nmreval#308