IRIS2 Flexure and Focus
- Slit Positioning
- Camera Focus - Sensitivity to Camera Defocus
- Camera Focus - Imaging
- Camera Focus - Spectroscopy
- Astrometric Distortion
The primary flexures of interest for IRIS2 observers are:
- Flexure of components in the SLIT wheel with respect to the rest of the IRIS2 optics : You need to know what pixel position the spectrograph slit correponds to on the detector, so that you can move an object there (with the SLIT wheel at its OPEN position), and then move the SLIT wheel to one of the slits, and be confident light from the target object is passing through the slit. You also want to be confident the slit is staying put at that location during an exposure, so your slit doesn't move out from underneath your target.
- Flexure of the spectrograph components within IRIS2 : if this happens your night sky and arc lines, etc. will get smeared out over the course of an exposure. In practise, single exposures with IRIS2, or even pair of exposures being subtracted for sky-subtraction are unlikely to be seperated in time or telescope position on the sky by more than 20-30 minutes. So for precise sky subtraction, you need to know the optics are stable over that sort of periods.
- Variation of best focus of the IRIS2 cameras with telescope position
Lets deal with each in turn.
Tests have been carried out with both the MATRIX mask and the slits in the SLIT wheel. The conclusions from these tests are that components in the slit wheel move by ~0.5 pixels when the telescope is moved to a Zenith Distance of 60 degrees in the East-West direction, but not at all when the telescope is slewed to any North-South position.
These movements are in the sense that the positions of spots from the MATRIX mask, or the centroid of the slit profile, moves by +0.55 pixels when the telescope is moved to ZD=60 in the WEST, and -0.55 when moved to ZD=60 degrees in the EAST.
The same shifts are seen when the telescope is moved to these positions, and then the wheel is positioned, or when the wheel is positioned and then the telescope is moved. In other words, these shifts are repeatable. They are believed to be due to different gravity vectors affecting the way that the IRIS2 wheels sit in their worms.
At present, observers are recommended to re-determine the centre of their slit for the purposes of acquisition, when they make large telescope movements E-W.
These flexures have no impact on imaging observations.
Focus sequences performed with the matrix mask in the SLIT wheel at a variety of telescope orientations indicate that best camera focus seems to decrease by ~35 encoder units when the telescope is moved off the vertical (in any direction) by ZD=41 degrees. That is best camera focus decrease by (0.834 encoder units / degree ZD). Examination of the sensitivity of camera focus (discussed below) indicates the impact of this on observing is very small, and it can be neglected.
Tests of slit re-positioning show that the slit re-position to <0.1" at all instrument orientations. However, as noted above, the slit position that it precisely repositions to, is a gentle function of telescope orientation (i.e. Zenith Distance E-W).
There is also a small amount of backlash in the repositioning. That is if you position the slit wheel by moving the wheel in different directions, it positions to a slightly different place. However, in normal acquisition operations, you will always move the slit wheel from OPEN to one of the slits by moving the SLIT wheel in the same direction, so this backlash is of academic interest only.
Focus - Sensitivity to Camera Defocus
Focus sequences obtained with the matrix (pinhole) mask enable us to determine how much image degradation results from a given amount of deviation of the camera from best focus. The relevant quantities are:
Error in Camera Focus
25 <0.02 <0.022" 50 <0.1 <0.045" 75 <0.2 <0.09" 100 ~0.4 ~0.18"
In other words errors in camera focus of less than 50 encoder units have essentially no effect on image quality, while errors of 75 units have only a marginal effect in all but the very best seeing conditions. Errors of 100 encoder units are clearly at a level which should be of concern.
Camera Focus - Imaging
The plane of best focus for IRIS2 (as observed using the matrix mask) is curved. Matrix mask focus sequences have been processed by fitting each point with an elliptical Gaussian. The major (green) and minor (yellow) axis FWHM are then fitted with parabolas to determine the best value of camera focus at that location on the array.
In the case of the configuration above the 'mean' camera focus across the detector is 275 encoder units. Examination of the 'best' focus curves shows that this mean focus implies that no region of the detector is more than 75 encoder units away from best focus, and so (from the sensitivity discussion above) image quality due to the curved focal plane varies by < 0.1" across the field of IRIS2.
The J filter has been adopted as the filter to be used for measuring IRIS2's "base" focus, when the telescope is pointed at the zenith. The following table shows the results of measuring this base focus over the course of time. In general we have found the base focus generally does not change over time, unless the IRIS2 internals are removed or adjusted. Unless your support astronomer indicates the IRIS2 internals have been adjusted, you can assume the base focus is the same as the last listed value in this table. The AAO_Inc.tcl file called by all standard AAO observing sequences sets this base focus appropriately.
If you want to check this, acquire a sequence of images of the SLIT_150um slit with the J filter in 25 encoder unit steps around the last measured value. Determine the focus which delivers the best image quality averaged along the slit.
Date J Camera Focus
Comments July 2002 275 October 2002 420 Camera optics removed between July and Oct 2002 runs January 2003 420 April 2003 411 July 2003 430 Jan 2004 416 Mar 2004 425 Jun 2004 410 Jun 2005 430 Engineering-grade array re-installed. May 2006 340 New science-grade array installed. May 2007 200 After full thermal cycle of main dewar. June 2009 225 October 2009 210
Focus sequence tests have been carried out in all the IRIS2 filters in January 2002. The important entry in the following table is the offset between the base focus at J and the focus for each filter. Even if the base focus moves in the future, these offsets should remain constant, so this table can be used to predict optimal camera focus for each filter.
Filter Camera Focus
Camera Focus Offset
(relative to J filter)*
J 419 0 H 392 -30 K 420 0 Ks 410 -10 CH4s 415 -10 CH4l 407 -10 Jcont 443 +25 PaBeta 421 0 FeII 395 -25 Hcont 401 -20 Z 420 0 He I 420 0 H2 v=1-0 432 +10 H2 v=2-1 447 +25 BrGamma 455 +35 Kcont 454 +35 CO 2-0 454 +35 Jlong 455 +35 Jshort 454 +35 Hspect 370 -50
* Rounded to nearest 5 units.
None of these focus offsets are very large, so neglecting them should not produce significant image quality deterioration. If you use a standard observing sequence (i.e. one that sets the instrument configuration using the MOVEImaging task defined in AAO_Inc.tcl) and you (or your support astronomer) have checked the base focus is set correctly at the start of your IRIS2 observing run, then you will automatically have these offsets applied.
Focus - Spectroscopy
Getting a good focus across the field with the 2.2-pixel slits used in IRIS2 is not hard. Generally you can get a focus which delivers 2.2 pixel FWHM lines at all wavelengths. Focus variations with telescope position are as for imaging.
Focus variations across field
This is hard to get a good feel for with 2.2 pixel slits and the focus variation as insensitive as IRIS2's is. The best we can say is that it looks similar to that seen with the imaging set-up. A 'mean' focus delivers good images across the field, but there is a small amount of focal plane curvature.
Focus variation with spectroscopic format (i.e. atmospheric window)
Focus sequences have been obtained for IRIS2 using on-sky measurement (i.e. focusing using actual sky lines).
Setup Camera Focus
Camera Focus Offset
(relative to J imaging focus)
J + Sap240 + OFF_150 500 +80 Jshort + Sap240 + SLIT_150 470 +50 Jlong + Sap240 + OFF_150 465 +45 H + Sap316 + OFF_150 368 (July2003)* -60* Hspect + Sap316 + OFF_150 368 (July 2003)* -60* K + Sap240 + SLIT_150 365* -55* Ks + Sap240 + SLIT_150 365* -55*
If you use a standard observing sequence (i.e. one that sets the spectrograph configuration using the MOVESpectro task defined in AAO_Inc.tcl) and you (or your support astronomer) have checked the base focus is set correctly at the start of your IRIS2 observing run, then you will automatically have these offsets applied.
The IRIS2 plate scale is around 1% different at the field corners from that at the field centre. This means that if you dither by more than a few arcminutes, you will not be able to register images for the purposes of making a combined image with a simple shift or rotation (this is true for almost every imaging system on a large telescope which delivers wide field images).
This distortion can be measured with some precision using the IRIS2 matrix mask. We know the holes in the matrix mask have been placed on a 3mm grid with great precision. For the focal reduction of IRIS2, if the field had no astrometric distortion, this would correspond to a regular spacing on a 44.84 pixel grid. We can therefore develop a mapping between this 'ideal' object spacing, and the observed spacing. A non-linear minimisation analysis has been used to determine a radial distortion function which (together with an arbitrary shift and rotation, and a small amount of tilt of the IRIS2 detector relative to its focal plane) gives us results like the following.
You can see that a smooth radial disortion correction (centered on pixel 516.857,515.015) allows the entire field to be mapped onto a square co-ordinate system with residuals of 0.06 pixels in radius or 0.09 pixels in X/Y.
Should you wish to convert your images to a square co-ordinate system (for the purposes of mosaicing images correctly) the recommended form for the correction is
Let R' = the measured radius from a central pixel DX0,DY0
and R = the radius from DX0,DY0 in an ideal 'undstorted' co-ordinate system.
R = R'(1 + A1*R' + A2*R'*R' + A3*R'*R'*R' + ... )
It is a general property of astrometric distortion in astronomical systems that they are either pincushion, or barrel. That is the plate scale either gets uniformly smaller, or uniformly larger with radius from the distortion centre. It is therefore useful to set all the even terms above to zero (i.e. A2=A4=...=0), and seek solutions in which A1, A3, ... all have the same sign.
We are gaining experience however, at present it appears that
R = R'(1 - 2.4988e-6*R' - 4.4466e-11*R'*R'*R')
A1 = -2.4988e-6
A3 = -4.4466e-11
delivers good results.