Open mgalloy opened 7 months ago
I've gotten Steve's code running and produced D matrices for 20221118 and 20221119.
I ran 20221026 as a test date to compare Steve's results to mine to make sure I am getting the same results as Steve on a date where I had Steve's results. The demodulation matrices are all exactly the same, but the temperatures are wrong, mine are showing up as 0.0.
For example, for me:
Wavelength: 637.400
Unfiltered Temperature: 0.00000
and for Steve:
Wavelength: 637.400
Unfiltered Temperature: 32.7730
This will affect creating the temperature dependent fit in the next step.
I have fixed the issue with the temperatures. Mine now reads the same as Steve's:
Wavelength: 637.400
Unfiltered Temperature: 32.7730
The UCoMP_Pol_Cal_Temperature
fit is failing with:
2 sav files
4 total calibrations at 637.400
1 cals rejected due to bad chisq
0 points identified as outliers
... large array of all zeros omitted ...
1 cals rejected due to recurrent bad points
% Compiled module: POLY_FIT.
% POLY_FIT: Singular matrix detected.
I did check to see if the temperatures were different and they were:
33.1440 33.1550 33.0770 33.0790
I don't think we have enough reading to fit the temperature. Using UCoMP_Pol_Cal_Temperature
I had to include the August-November calibrations to get a 1074 nm temperature dependence, about 48 1074 nm calibrations — the 35 calibrations from September-November was not enough. We only have 2-4 calibrations for the new wave regions.
The DMatrix
coefficients, it seems we have agreement to maximum of about 0.003 (1.40181 - 1.40454 = -0.00273).
For 20221118:
Dmatrix:
-0.25433 0.66068 0.05861 0.53552
0.33624 -0.80948 0.53646 -0.06347
-0.28015 0.15453 1.15413 -1.02808
1.40181 -0.73843 -0.25481 -0.40767
For 20221119:
Dmatrix:
-0.25433 0.66139 0.05836 0.53522
0.33627 -0.81005 0.53625 -0.06264
-0.28197 0.15620 1.15623 -1.02889
1.40454 -0.73967 -0.25390 -0.40889
IDL> d18 - d19
0.0000000 -0.00071001053 0.00025000051 0.00029999018
-3.0010939e-05 0.00056999922 0.00020998716 -0.00083000213
0.0018199980 -0.0016700029 -0.0020999908 0.00081002712
-0.0027298927 0.0012399554 -0.00091001391 0.0012200177
Looking at the de-modulation coefficients across the wave regions. It seems like there is a fairly smooth quadratic relationship between the wavelength and coefficient. So we should be able to safely project to new wavelengths.
https://github.com/NCAR/ucomp-studies/blob/master/demodulation/de-modcheck.ipynb
@bberkeyU How much variation of temperature do you expect now? If there is no (or very little) variation, we can make a constant demodulation matrix even from the few readings we have.
@mgalloy I pulled the tu_mod values from the database for Dec 2021->Nov 2022. The results were not exactly what I expected. Somewhere in April 2022, it looks like the temperature started slowly rising, moving toward 32.5C. We also have some big excursions over the course of a couple of days.
We need to check (for the old lines) if the new demodulation matrix is consistent with the old demodulation matrices. After so much time closed, this is not something we can take for granted.
We should check when was the last time Steve updated the demodulation matrix. I would update it again, especially if the last update was done a while back. If the new polarimetric calibration looks strange we will not use it for the update and need to understand what that means.
The 20220409 polarimetric calibration coefficient for the old wavelengths, especially for 1074 and 1079 nm are not in agreement with the old demodulation coefficients (see closed ticket https://github.com/NCAR/ucomp-pipeline/issues/265). In some cases they differ by 10% or more. Plots were produced by Ben to compare the new coefficients with the old ones. The plots can be found in this notebook:
We do not have an explanation for why there is such difference and it is hard to draw conclusions with only one data point. We need to take more polarimetric data when the observatory reopens.
Create a demodulation matrix for each of the new wave regions from data observed in late November 2022.
Basic steps:
UCoMP_Pol_Cal.pro
to create a demodulation matrix for each day with polarimetric data after 20221026.UCoMP_Pol_Cal_Temperature.pro
to combine days results to give a demodulation matrix with a temperature dependence.Eventually, this should move into the pipeline infrastructure, but for now could just change the hard-coded paths from Steve's code.
Conclusions
My conclusions right now:
Useful information
Tasks
UCoMP_Read_Catalog_File.pro
to itUCoMP_Pol_Cal
on 20221118UCoMP_Pol_Cal
on 20221119UCoMP_Pol_Cal_Temperature
to add new wave regionsUCoMP_Pol_Cal_Temperature
Routines to find
openpost.pro
,closepost.pro
, and dependencies (doesn't seem to be the UCOMP subversion or in SolarSoft) found inDocuments/idl/Plotting/
in the backup of Steve's data -- there are other versions, but I can tell this is the correct versionmueller_retarder.pro
,mueller_polarizer.pro
, and dependencies (doesn't seem to be the UCOMP subversion) found inDocuments/idl/Plotting/
in the backup of Steve's data, but there are other versions that I can't tell if they are the correct onesSummary
The 20220409 polarimetric calibration data, especially for 1074 nm, is not in line with the coefficient fits (see #265). Waiting for more polarimetric calibration data when we reopen to create these demodulation matrices.