[PIPE2D-584] Dithered images are not properly dithered Created: 15/May/20  Updated: 29/Jan/22  Resolved: 13/Jan/21

Status: Won't Fix
Project: DRP 2-D Pipeline
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Task Priority: Normal
Reporter: ncaplar Assignee: ncaplar
Resolution: Won't Fix Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: PNG File Dithering_Aug_14.png     PNG File Dithering_Aug_14_Updated.png     PNG File Dithering_May13_Arnaud.png     PNG File Dithering_May14.png     PNG File Dithering_May19.png     PNG File Dithering_May20.png     PNG File Dithering_pixels_added_simulation.png     PNG File Single_example.png     PNG File SM1_dither03_20200424_Exp225.png     PNG File SM1_dither03_20200424_Exp225.png    
Issue Links:
Blocks
blocks PIPE2D-581 Create the postage stamps for the dit... Open
Relates
relates to INSTRM-1051 Ensure that hexapod moves always move... Done
Story Points: 4

 Description   

When looking at creating dithered images (PIPE2D-581) I noticed that dithered images are looking suspicious. arnaud.lefur and I did some investigative work, presented in the following figures:

1) Dithering_May14.png - shows by how much the center of flux changes between images after dithering. I created results for the sequence of only 4 dithers (4440-4463) and the sequence of 9 dithers (4720-4773), taken at Subaru. Also added is the data from the last LAM dithered acquisition (4 dither positions, 21256-21279). I conclude 3 things:

a) Subaru and LAM results are different 

b) final LAM dataset is also not perfect - at some point in the past the values for dithering were deduced by arnaud.lefur so that dithering was ``perfect''. 

c) there is also potentially the difference between the results when depending on the sequence of moves (the difference between yellow and black data), but this would have to be further investigated in order to make a certain claim.

 

I am also attaching Figure Dithering_May13_Arnaud, which shows the analysis did by Arnaud which mimicking the black points from the previous plot. The results are consistent.

Finally, I am also attaching Figure Dithering_pixels_added_simulation.png, which was the original Figure showing this effect that was discussed on Slack.

 

 Comments   
Comment by ncaplar [ 20/May/20 ]

I have added some further images. Dithering_May19 shows a complete list of all dithering images that we ever took at LAM or Subaru. The points show the median difference after a dithering movement, deduced from many HgAr spots across the detector. The error bars show the 16%-84% quantile of the results.

 

Single_example.png shows the example of the same spot in the July LAM data and Subaru data. The title above the figure shows the centroid of the 20x20 image. I am showing only the center 10x10 pixels so that the images are clearer. You can clearly see the difference between dithered images, even though the movement of the slit is nominally the same for two locations. 

Comment by ncaplar [ 18/Jun/20 ]

On June 18, Fabrice and I plan to take some additional data to characterize the problem in blue (as red shutter does not function at the moment).

Comment by arnaud.lefur [ 18/Jun/20 ]

Did you already checked visit_set(368): ditheredArcs, PU dataset, 17998, 18051
I took those for software validation purposes.

Comment by ncaplar [ 18/Jun/20 ]

No, I did not! What kind of dithering did you do (9 positions, 4 positions?). I think that Fabrice and I are thinking of taking datasets which would be different than until now to better understand the problems. At least on my mind, it will be dithered by 1/3 and extending dithering by more steps, i.e., further than 0.5 pixels.

Comment by arnaud.lefur [ 18/Jun/20 ]

There are 9 positions.
sure, I get it. In any case, if you want to check it out, it's there.

Comment by ncaplar [ 14/Aug/20 ]

Added results from the first semi-successful run on Subaru in August. As I have written in slack: ``I definitely see that the movements are not the same in both directions, but it did not reproduce the suspicious large difference in one direction, as sometimes seen before.''

Comment by ncaplar [ 18/Aug/20 ]

I have added an updated version of Aug 14 experiment, with all of the positions available. The main conclusions are unchanged from the preliminary version. I have improved my centroid estimation which lowered the uncertainties of measurements. In the estimates of 1d uncertainties on the right-hand side of the Figure, the uncertainties do not include uncertainty on the estimation of the position at 0,0,0 position, i.e., 0,0,0 position is assumed to be known with infinite precision. 

Comment by ncaplar [ 13/Jan/21 ]

This issue has been moved into the domain of instrument tickets (e.g., INSTRM-1051 and following tickets).

Generated at Sun Apr 06 22:39:16 JST 2025 using Jira 8.3.4#803005-sha1:1f96e09b3c60279a408a2ae47be3c745f571388b.