Once
again the co-alignment of lenslets is under discussion.
The
scheme for operating the WFS assumes that the lenslet to CCD alignment is set
up for quad cell reading even if the spot pattern will not be read out in quad
cell mode.
For
quad cell operation the spots must fall onto the grid of pixels on the CCDs to
within ±1.2 mm across the field. This translates to a ±6.0 mm tolerance on the lenslet
positioning in X and Y and 1/100th of a degree in rotation. The CCD
is a fixed reference to which all lenslet arrays must map. This mapping must be
maintained as the carriages move up and down the WFS rails and not just for a
fixed position.
The
rails have a running parallelism specification of approximately 4 mm per 100 mm. The mounting of the rails will
probably degrade the parallelism, but by how much will not be known until the
carriages have been mounted and tested. To measure this scale of error reliably
is beyond the scope of the co-ordinate measuring machine, so interferometric
methods will be required. That means measuring tilts of a mirror at discreet
points along the travel and curve fitting to find displacements. The extremes
of travel of the carriages following the pick-off between the calibration source
and the far edge of the field is roughly 60 mm, so we can expect to have lost
half the allowance for positioning error of the lenslets.
Lenslet
alignment involves the manipulation of the lenslet arrays in an XYq-stage to position it in its holder correctly
aligned to the CCD. Once in position it has to be fixed in place. In order not
to disturb this level of alignment precision the lenslet will be fixed by
bonding using a UV curing cement. Hopefully curing stresses in the adhesive
will be sufficiently low, or symmetrical, that upon releasing the lenslet after
the curing of the adhesive it will not drift.
An
alternative approach is to mount the lenslets, reticule and full aperture
doublet into a plate. All four lenslets must be orientated in rotation, but
their positions are nominal. The plate is mounted onto an actuator driven
XY-stage of 50 x 25 mm travel. When a lenslet is required it is selected and
there will be a positioning procedure in which the spots on the CCD formed by a
calibration source will be read of and the lenslets repositioned in a closed
loop system to obtain quad cell operation alignment.
There
may need to be a look-up table of corrections for offsets to the lenslets to
accommodate out of straightness of the rails, but at least this would be
possible with an XY-stage. Calibration for this would be done here or at Durham. As things stand, if offsets are
required that’s just too bad.
The XY-stage would also be useful for white
light flattening of the DM as the required ½ X ½ subaperture offset could be
dialled up at the lenslet. That would mean we could leave out the offset array
giving one unobstructed aperture in the lenslet plate which would be a great
help for alignment checks. It would still be useful as a spare for the preferred
quad cell mode.
The
wheel would be discarded and replaced by a pair of stages mounted for XY
motion, the 50 mm travel stage being flat on the carriage where the base of the
wheel now mounts. The lenslets themselves would be mounted in a plate with a 3
X 2 grid of holes at 24.5 mm pitch.
From
the point of view of the mechanics and alignment there would be a probable
saving in time over the current system, but there would be a knock on for the
software. It may be that the code could be obtained from Electra or the Keck
group and modified. The current VME cards could be juggled to accommodate
stages with encoders as the slots for the ADC prisms do not need their
currently allocated capacity for reading encoders.
The
specification for the stages would be driven by repeatability required for
implementing offsets. This would be 3mm approximately. There are at least 2 options for
stages/motors from different suppliers that would more than meet the
requirements and fit in the space available.