Donut global stability study¶
I have studied the stability of the Donut measured decenters over a period of several nights, using images taken in the supernova C1,C2,C3 fields. These fields are single ra,dec positions and the three of them are quite close in position.
This plot shows the Azimuth vs. Altitude of the images taken in these SN fields from 12/7 through 1/3. The color scale shows the exposure id - with expid from 159000 to 165000 or so. These images were taken in sequence, so points with the same color show you images taken in a sequence on a given night.
This plot shows the value of the x decenter (dx) as a function of Altitude for these images (units here and for all decenters are microns). During this period the AOS controlled only focus not decenter or tip/tilt, so this shows just the telescope LUT value. You can see that the LUT moves the hexapod over a wide range for this range of Alt.
Next we show the value of the Donut measured x decenter (dodx) again as a function of Alt, and with the color showing the expid. This is the offline measured dodx, which is very consistent with the online version, but is able to use more donuts. For a given night the dodx values do show a clear trend, but from night to night there are clearly changes. For instance the green points show dodx around zero, while the light blue have dodx starting around 200 microns and moving up to values around 1000 microns. For the same Alt values they differ by around 1000 microns, but these images were taken at the same Alt,Az and same RA,dec just 3 nights apart (night of 12/15 vs. 12/18)
Now what about the BCAMs? The same kind of plot is shown for BCAMs x decenter here:
The BCAMs show very nice tight bands of values, modulo a few anomalous points. But there is large variation from night to night in these values - again even though these are taken at the same Alt,Az and RA,dec. It is curious that the changes do seem to move in time order, by the way.
Lastly we can look for any correlation between the Donut dodx and the BCAM value here:
and you can see that over all these images there really isn't any correlation. We have definitely seen some correlation - good in dy and not as good in dx - between Donuts and BCAMs, but generally for a single sequence of images. What this shows is that over different nights that correlation is washed out.
Here are the same set of plots for y decenter. They show less overall variation, but the last conclusion remains - any short term correlation between donuts and bcams is washed out.
So what is going on? I think this shows clear evidence that the optical alignment is different than the
mechanical alignment - Donuts and BCAMs are just different. And it shows evidence that the absolute alignment is changing - both the optical alignment and the mechanical alignment, and moreover that these are changing differently. The problems with the pointing model may indicate another source of changes - but since 1000 microns = 18 arcsec - the alignment issues here could only introduce 10-20 arcsec size effects.
One final comment, to set the scale of this issue. A variation in decenter of 1mm corresponds to a change in Coma of just 0.2 waves, or 1.2 microns OPD. The variation of Coma across the field of view ranges from -0.2 to +0.2 waves or so, so this shift should have an effect on image quality but not an enormous one.