here

Report
on
CYCLOPS
Pre‐Commissioning
Tests
(Apr
24‐27)
Version
Control
1.0
First
version
–
CGT
8
June
2010.
1.
Introduction
Over
the
period
April
24‐28
Chris
Tinney
&
Guy
Monnet
(with
assistance
from
Jurek
Brseki
and
Steve
Lee,
in
addition
to
other
Site
&
Epping
staff)
carried
out
tests
of
CYCLOPS
mounted
on
the
AAT.
These
tests
aimed
at
addressing
the
items
5.1‐5.6
of
the
CYCLOPS
Commissioning
plan.
The
report
consists
of
an
overview
of
what
was
learned
during
the
commissioning,
and
specific
reports
on
the
commissioning
plan
items.
2.
Issues
to
address
1. Quartz
lamp
too
bright
–
needs
to
be
about
a
factor
of
10
fainter.
2. Relative
fibre
throughput
–
throughput
varies
by
20%
from
best
3
to
worst
3
fibres.
3. Filter
holder
–
design
needs
a
rework,
as
the
holder
holds
the
filters
very
well,
but
the
holder
flops
about
in
the
CYCLOPS
cass
unit
introducing
image
movement
on
the
guider.
4. Because
the
back‐viewing
mode
of
the
guider
focuses
on
the
front
face
of
the
microlens
array,
rather
than
the
fibres,
it
will
not
be
much
use
for
aligning
stars
onto
CYCLOPS.
CGT
will
have
to
develop
some
code
to
process
images
to
get
a
reconstructed
image
out
so
that
we
can
position
a
reference
star
on
the
centre
of
the
CYCLOPS
field.
5. Commisisoning
Tests
from
5.1‐5.6
still
to
do
a. 5.2.2
–
Slit
unit
removal
to
test
position
repeatability.
b. 5.5
–
Repeat
measurements
of
fibre
offsets
over
time,
and
especially
as
a
function
of
pre‐slit
area
temperature.
3.
Overview
CYCLOPS
Status
at
Commissioning
:
CYCLOPS
currently
has
a
fibre
bundle
in
whoch
3
of
the
15
fibres
are
broken
or
not
connected.
The
mapping
between
the
entrance
array
and
the
pseudo‐slit
formed
at
UCLES
is
shown
in
the
following
figure.
CYCLOPS
Cass
Feed
Layout
X3
X2
X6
1
5
4
7
9
8
11
13
10
12
15
14
•
orientation
currently
unknown
•
three
fibres
dead
(6,3
&
2)
CYCLOPS
Pseudo‐slit
Layout.
15
14
13
12
11
10
9
8
7
X6
5
4
X3
X2
1
Calibration
Lamps
:
both
the
internal
Quartz
flat
field
lamp
and
the
ThAr
arc
lamp
work.
However,
it
is
clear
that
nether
of
them
uniformly
illuminate
the
CYCLOPS
fibre
bundle,
as
the
both
show
significant
(and
different)
illuminations
of
the
fibres.
The
Quartz
lamp
is
actually
too
bright
(amazing
but
true….).
Observations
with
it
have
to
be
just
0.5‐1.0s
and
taken
with
an
ND
filter.
As
ND
filters
can’t
be
relied
on
to
be
truly
neutral
density
at
the
precisions
relevant
for
CYCLOPS
work,
we
need
to
be
able
to
take
flats
without
filters,
which
means
reducing
the
lamp
flux
by
about
a
factor
of
10.
Image
quality
:
Image
quality
is
generally
good.
When
images
of
arcs
are
taken
with
the
detector
unbinned,
the
individual
fibre
images
are
clearly
seperated.
Tests
indicate
we
can
get
a
focus
of
~2.5‐2.7
pixels
across
the
whole
field
(though
also
indicate
the
best
focus
surface
of
UCLES
is
tilted
relative
to
the
detector
–
fortunately
the
depth
of
focus
on
UCLES
is
good
enough
still
produce
good
images
over
the
field).
The
images
produced,
though,
have
the
'flat‐topped'
profile
you'd
expect
from
a
fibre,
convolved
with
the
instrument
PSF.
As
a
result
making
a
direct
comparison
between
the
70um
fibre
diameter,
and
the
63.4um
FWHM
(at
the
fibre
exit
plane)
that
2.55pix
at
the
detector
corresponds
to,
is
not
straightforward.
I
show
on
the
following
pages
a
full
unbinned
ThAr
arc
image
(left),
and
a
zoom
on
the
central
regions
(right)
Throughput
:
Test
observations
of
the
interior
of
the
dome
(ie
through
the
telescope
and
CYCLOPS
and
UCLES)
have
been
used
to
measure
the
relative
throughput
of
the
fibres
in
the
CYCLOPS
system
(we
can’t
measure
absolute
throughput
until
we
get
on‐sky).
The
worst
3
fibres
are
about
20%
down
in
throughput
compared
to
the
best
3,
with
the
remainder
sitting
in
between.
This
is
not
great.
Operations
:
a
few
operational
issues
became
evident
•
the
bayonet
kinematic
mount
fitting
for
the
CYCLOPS
fibre
head
into
the
cass
unit
is
a
great
idea,
but
needs
work
on
the
implementation.
At
the
moment
one
has
to
remove
~12
non‐captive
screws
(at
least
one
of
which
has
already
become
cross‐
threaded)
to
plug/unplug
the
fibre.
This
is
insane,
as
it
makes
a
5
minute
job
take
almost
an
hour
to
get
done.
•
When
removed
the
bayonet
kinematic
mount
fitting
for
the
CYCLOPS
fibre
head
needs
a
safe
and
secure
location
to
store
it
in
the
cage
where
it
will
be
immune
to
bumps
people
and
hardware
in
the
cage.
It
also
needs
a
better,
bayonet
fitting
cover
for
the
fibre
head,
which
has
optical
surfaces
that
must
be
protected.
Fringing
:
at
the
level
of
2%
(peak‐to‐peak)
was
found
in
the
flat
field
exposures.
The
period
of
the
fringing
is
quite
short
at
~10
pixels.
We
have
not
yet
determined
whether
this
comes
from
the
lamp,
the
lamp
mirror,
the
fibre
unit
itself,
or
the
neutral
density
filter
we
had
to
use.
Analysis
is
proceeding
–
the
most
likely
culprits
are
the
ND
filters
and/or
the
lamp
or
its
mirror.
Software
:
a
number
of
infelicities
were
identified
with
the
GUI/software
that
controls
the
CYCLOPS
calibration
unit.
Some
of
these
are
being
addressed
by
Minh
already
•
The
calibration
hardware
has
timeouts
built
into
it
at
a
low‐level
(ie
the
Quartz
and
ThAr
lamps
time
out
and
switch
off
after
quite
a
short
period
of
time).
The
problem
with
this
is
that
because
the
CYCLOPS
GUI
is
not
integrated
with
the
detector
controller,
there
is
no
way,
when
you
start
an
exposure,
for
the
ODC
to
check
whether
the
lamps
have
timed
out
and
turn
them
back
on!
Until
CYCLOPS
is
integrated
into
the
ODC,
these
timeouts
needs
to
be
set
to
some
timescale
long
enough
that
they
cannot
impact
on
observing
or
calibrations
(~2h).
•
The
control
of
ThAr,Quartz
and
their
mirrors
is
clunky.
Astronomical
users
will
only
ever
want
one
of
three
things
:
take
a
flat,
take
and
arc,
observe.
So
there
only
needs
to
be
three
buttons
for
these
states
“Flat”,
“Arc”,
“Star”.
Fibre
positioning
(Y=dispersion
direction):
the
fibres
forming
the
pseudo‐slit
at
the
entrance
of
UCLES
have
a
peak‐to‐peak
range
of
position
in
the
dispersion
direction
of
0.1pixel,
which
corresponds
to
1.35um
in
the
detector
focal
plane,
or
17.9um
at
the
UCLES
entrance
slit
(for
an
UCLES
de‐projection
factor
in
the
spectral
direction
of
13.25).
This
corresponds
to
2.5um
at
the
physical
fibre
exit
(for
an
assumed
magnification
of
5/36=0.139).
This
means
the
fibre
positioning
in
the
dispersion
direction
has
a
peak‐to‐
peak
variation
of
3.5%
of
a
70um
fibre.
Analysis
of
arc
exposures,
where
care
is
take
to
‘derotate’
each
pseudo‐slit,
show
that
these
offsets
can
be
modelled
as
being
the
same
over
the
whole
field.
Fibre
positioning
(X=spatial
direction):
The
fibres
forming
the
pseudo‐slit
have
a
peak‐
to‐peak
variation
in
the
X
(or
spatial)
direction
of
0.52
pixels,
or
8.4um
at
the
fibre
exit
plane
(for
an
UCLES
spatial
de‐projection
of
8.63,
and
a
CYCLOPS
slit‐unit
magnification
of
1/7.2=5/36=0.139.
Fibre
spacing
:
The
same
fitting
process
enables
us
to
look
at
how
the
fibre
spacing
varies
over
the
field.
The
fibre
spacing
(ie
distance
fibre
centre‐fibre‐centre)
has
a
mean
value
of
5.8
(though
it
is
linearly
correlated
with
X
position
on
the
detector
(due
to
the
use
of
a
prism
cross‐disperser),
and
ranges
from
5.60
to
6.0
pixels.
This
means
that
the
spacing
between
fibre
centres
is
~144um.
3.
Reports
against
Commissioning
plan
items
5.1 Obtain test exposures with new ThAr and FF lamps in CYCLOPS
• Verify useful exposure times with these lamps, and that they meet the FPRD.
o 1-2s exposures with the ThAr lamps produce useful exposures. 10-20s will see
much fainter lines, but saturate brighter ones.
o 0.5-1.0s exposures with the Quartz lamp will saturate UCLES even when the
detector is used unbinned. The lamp needs to be about 10 times fainter.
•
Verify image quality from arc lines meets that expected from UCLES when illuminated
with the CYCLOPS exit fibres.
o Image quality is acceptable – a focus can be obtained that delivers 2.5-2.7 pixel
images over the whole field.
5.2 Map Fibre Output in Direction along the Slit
• Each of the individual fibres at the input end of CYCLOPS must be illuminated with a
continuum lamp source or laser, so that the position of the output spectrum along the
pseudo slit at the entrance to UCLES can be mapped. (2 days to set-up and implement)
o Individual fibre illumination could not be achieved – however, we were able to
determine that arc exposures could obtain good position information across the
whole detector, and so this test was not necessary.
•
This measurement will need to be repeated following a cycle of removing the CYCLOPS
slit unit, and remounting it. (1 day)
o A cycle of slit removal and replacement was not done, and needs to be done
during on-sky commissioning.
5.3 Map fibre relative throughput
Using the mapping between position along the slit obtained above, a uniform flat-field
illumination of the fibre input end will be used to determine the relative throughputs of the fibres,
both at a nominal wavelength (~5500A), as well as as a function of wavelength.
o
Dome flats were obtained, but gave adequate counts for relative throughput
measurement at just a few emission lines present in the dome fluorescent lights. I
analysed three lines (at 5757,5779 & 5525A). I treated each fibre image for each
line as a ‘star’ in a 2D image, and used Sextractor to identify the positions of the
flux coming from each fibre. These were then used as the starting positions for a
DAOPHOT analysis of all these lines (this was necessary as all the fibre images
overlap, and so a simultaneous solution for all 12 fibres in a pseudo-slit is critical).
The fluxes from each fibre in each of these three lines were then normalised to
the flux from the fibre with the highest throughput, to produce the following plot.
The green trace (line at 1075,3082) is the faintest line and so has the least
weight. The fibres are numbered from the left (1,4,5,7,8,9,10,11,12,13,14,15) –
fibres 2,3 and 6 are broken.
The fibre throughput variations clearly repeat, with the worst three fibres (ie the worst
quartile of the 12 available) ~20% worse than the best three fibres.
o
Data to determine fibre throughput as a function of wavelength this was taken (ie
internal flat exposures, and dome exposures that uniformly illuminate CYCLOPS),
but has not yet been analysed.
5.4 Map CYCLOPS wavelength formats
Using a set of UCLES echellograms pointings (ie. ET,EG pointings) that provide echellograms
covering the full useful wavelength range of UCLES (ie 350-1000nm), obtain mappings between
echellogram x,y pixel position and wavelength. Flat field exposures at the same echellogram
pointings wil indicate the wavelengths at which no useful data can be obtained from
UCLES+CYCLOPS.
o
Observations were acquired in three wavelength settings that I will hereafter refer
to as “blue”,”mid” and “red”. “mid” was specifically set up to place the key
wavelength range for planet search work (5000-6100A) near the centre of the
wavelength format.
A full analysis has been done to date for only the “mid” setting – in general, all
this shows is that we achieve exactly the wavelength formats and coverage
expected, given we are using the 79l/mm echelle. The processing done consisted
of straightening the images using the FIGARO sdist & cdist functions, by tracing
the well-seperated fibre 1 image in a flat-field, and then doing a standard
extraction and wavelength calibration. Dispersions obtained were was expected
for the UCLES 79l/mm echelle, and had typical scatter about the 5th order
dispersion calibration polynomial of 1/10th of a pixel (in each order).
The wavelength coverage provided in this configuration when using the EEV2 whole
detector mostly complete (though gaps appear as you go further to the red). There
will be no coverage gaps short of 5500A.
Order
49
48
47
46
45
44
43
42
41
40
39
38
37
36
35
34
33
32
31
Coverage
4542 4652
4635 4749
4735 4850
4838 4955
4945 5065
5057 5180
5174 5300
5297 5426
5426 5558
5562 5697
5704 5842
5854 5996
6012 6157
6178 6328
6354 6508
6541 6699
6739 6902
6949 7117
7172 7346
Missing
4
7
12
16
21
26
33
40
47
55
However, there is significant vignetting of the camera that appears at the red and
blue end of each order. This manifests as a sharp inflection point or ‘kink’ in the
throughput, which can be easily seen in extracted flat-field exposures, like the
following
Note also the fringing pattern, which shows a 2% peak to peak with a period of about
10 pixels – I still don’t know the source of this fringing.
This vignetting is well known in UCLES, and is present in all 31l/mm and 79l/mm data
taken with a slit. The wavelength coverage if one clips each order when this
vignetting sets in for the “mid” configuration is then
Order
49
48
47
46
45
44
43
42
41
40
39
38
37
36
35
34
33
32
31
Coverage
4560-4637
4655-4733
4754-4835
4858-4940
4966-5050
5078-5165
5195-5284
5320-5410
5450-5540
5585-5680
5728-5825
5880-5980
6038-6140
6205-6310
6380-6490
6569-6680
6765-6882
6980-7100
7200-7322
Missing
18A
21
23
26
28
30
36
40
45
48
55
58
65
70
79
85
98
100
5.5 Map Fibre Output Offsets in Direction Orthogonal to the Slit (3 days + 1 day)
Each of the individual fibres at the input end of CYCLOPS must be illuminated with a laser lamp
and/or a ThAr arc spectrum.
•
•
•
•
The resulting spectra must be analysed to determine the offset of each fibre from the
mean central axis of the slit (ie. in the direction orthogonal to the length of the slit) can
be determined. This offset will need to be determined to a level of at least 1/100th of a
spectral PSF, and preferably higher.
The spectra must be analysed to verify that each fibre meets the image quality expected,
and to determine the level of variability in PSF (if any) between fibres.
This measurement will need to be repeated following a cycle of removing the CYCLOPS
slit unit, and remounting it.
This measurement will need to be repeated on a night when the slit-unit temperature has
been changed significantly (ie. by at least 5 degrees, and preferably by 10 degrees),
compared to the first test.
As noted above individual illumination did not turn out to be feasible. However, the arc
spectra are well separated, and so analysing the images as an individual ‘star’ for each
arc line in each fibre turned out to be feasible. Sextractor was used to identify arc lines –
its deblending and simultaneous fitting allowing reasonable centroids to be obtained even
when images overlapped.
I then analysed these individual images to look at how the individual arc images in each
pseudo-slit were offset relative to the slit. Each pseudo-slit was fit with a linear slope, to
take account of the changing distortions of the slit produced by the UCLES optics. These
manifest as a changing slit angle on the detector over the field.
Some time was spent coming up with a simple geometric model that could model and
remove these slit angle variations (which have a magnitude of +- 4 degrees over the
field). Amazingly a model consisting of two convergent points off the edge of the detector
(presumably corresponding to the two off-axis reflections/beam deviations produced by
the (1) Echelle and (2) prisms) and one radial distortion (presumably produced by the
Schmidt camera optics) was able to model this slit angle variation with residuals of just
0.01 deg over the whole field. (While this is not terribly exciting for CYCLOPS, it suggests
that UCLES Echelle reduction could be massively simplified by re-mapping the spectra
using a very simple geometric transformation, rather than having to trace orders and arc
lines ……)
The following figure shows the effect removed by each succeeding component of the
model – top left are the raw slit angle data, then top right are the residuals after a
convergent point off the bottom of the detector is removed, then lower left are the
residuals after a convergent point off the left edge is removed, and then lower right are
the residuals after a radial distortion is removed.
After rotating all the pseudo-slit X-Y data (using either the fitted slopes or a the model) it
is then possible to look at the systematic offsets we see for each fibre about the pseudoslit mid-line. The results for the X and Y offsets for the entire field are shown below (red
dots are individual measures, squares and error bars are the means and starndard
deviations.
The data values that arise from this for each of the 12 fibres available are
Fibre
Xoffset +- RMS
Yoffset +- RMS
===================================================
Fibre 1 :
0.319 +- 0.018
0.021 +- 0.011
Fibre 4 :
-0.083 +- 0.034
-0.061 +- 0.010
Fibre 5 :
-0.052 +- 0.021
0.028 +- 0.015
Fibre 7 :
-0.202 +- 0.017
-0.031 +- 0.013
Fibre 8 :
-0.106 +- 0.015
-0.017 +- 0.012
Fibre 9 :
-0.125 +- 0.020
0.020 +- 0.011
Fibre 10 :
0.003 +- 0.023
0.049 +- 0.017
Fibre 11 : -0.024 +- 0.021
0.048 +- 0.017
Fibre 12 : -0.000 +- 0.025
0.028 +- 0.020
Fibre 13 :
0.057 +- 0.017
-0.052 +- 0.017
Fibre 14 :
0.053 +- 0.022
-0.069 +- 0.016
Fibre 15 :
0.162 +- 0.017
0.036 +- 0.012
===================================================
The peak-to-peak variation in the Y (or spectral dispersion) direction is 0.1pixel, which
corresponds to 1.35um in the detector focal plane, or 17.9um at the UCLES entrance slit
(for an UCLES de-projection factor in the spectral direction of 13.25). This corresponds to
2.5um at the physical fibre exit (for an assumed magnification of 5/36=0.139). This
means the fibre positioning in the dispersion direction has a peak-to-peak variation of
3.5% of a 70um fibre. The precision with which
these offsets can be measured is ~0.6% of a
spectral FWHM.
The fibres forming the pseudo-slit have a peak-topeak variation in the X (or spatial) direction of 0.52
pixels, or 8.4um at the fibre exit plane (for an
UCLES spatial de-projection of 8.63, and a
CYCLOPS slit-unit magnification of
1/7.2=5/36=0.139.
The same fitting process enables us to look at how
the fibre spacing varies over the field. The fibre
spacing (ie distance fibre centre-fibre-centre) has a
mean value of 5.8 (though it is linearly correlated
with X position on the detector (due to the use of a
prism cross-disperser), and ranges from 5.60 to 6.0
pixels. This means that the spacing between fibre
centres is ~144um.
5.5 GUIDER calibration (1 day)
Back illuminate CYCLOPS bundle and determine bundle centre on guider camera as a function
of
•
Guider filter
The guider back-viewing mode turns out to not be terribly useful for aligning anything onto the
guider. This alignment will have to be done using stars and code to reconstruct and image from
an UCLES spectrum (as for SPIRAL).
•
Telescope DEC and HA (ie use back-illumination to examine flexure between guider
camera and fibre entrance)
The guider back-viewing mode was useful (using back-illumination) to measure flexure of the
lever-arm in the Cass unit between the guider camera and the fibre entrance (or at least the
microlens front surface).
An initial flexure test carried out in this way showed significant flexure – however, this was
potentially due to the (many!) screws on the Cass unit not being installed and tightened. After
tightening these, another test was done. This showed some ‘flexure’ however it was flexure with
hysteresis (ie that ‘flip-flopped’ as the telescope slewed from one side of the sky to the other).
This flexure is suspected to be due to the flopping of the filter-holder for the guider – as this
holder sits in the collimated beam, movement of it will move images on the guider CCD. A test
without the filter was carried out, and showed greatly reduced flexure. A detailed analysis of this
data has still not been done, though image centroids of the files were obtained as the data was
written at the AAT.
Images saved as g27may0001 …
Run
Position
ZD
X,Y(pixel)
============================================
2,3
Up
0
121.7,151.9
4
dec=-75
43.7
5
dec=-90
58.7
121.6,151.8
6
dec=+30
61.7
122.1,152.4
7
dec=+15
46.3
122.1,152.0
8
Zenith
0
121.9,151.9
9
+5:00 (ie W)
62.7
121.5,152.3
10
+4:00
50.6
121.7,152.0
11
+2:00
25.6
121.9,151.7
12
-2:00
25.6
122.6,152.0
13
-4:00
50.6
123.1,152.3
14
-5:00
62.7
123.3,151.8
15
Zenith
0
121.9,151.8
============================================
It would appear that when the filter is NOT present, that whatever flexure is present is
repeatable and without hysteresis. However, there is evidence that when the telescope went into
the East, that almost a pixel of flexure occurred. That corresponds to a 0.3” shift. Most of this
flexure seems to happen between 2:00h and 4:00h. More data and a more detailed analysis is
needed to map this effect, so we can determine whether it is real, and worth modelling.
It should be able to be taken out in software in the guider.