2.2 Telescope
The 4m telescope will be operated by an Observing Assistant, who is familiar with the operation of both the telescope and the CCD system. The Observer Support section will have helped in the setup and will have checked out the instrument, and they will respond to any problems that might arise. The telescope should point to better than 10 arc sec rms, assuming accurate input co-ordinates. If the pointing is much worse than this then you should complain. The telescope control computer (TCS) is a VME bus computer with a program written in C, operating under VxWorks. The TCS computer will be operated by your Observing Assistant.2.3 PFCCD
All CTIO CCDs that are available for direct use can be installed at the f/2.8 prime focus imager of the CTIO 4-m telescope, which is shown below.
Click on the image for a
more detailed view.
The CCD that has been used for all 4-m PF direct imaging is a Tektronix 2048, in recent times #6. This had a powered dewar window to compensate for the curvature of the CCD. See section 4 for details.
The PFCCD imager must be used with a corrector. The only corrector currently used is the new four-element ADC (Atmospheric Dispersion Corrector). This corrector can compensate for atmospheric dispersion up to a zenith distance of 70 degrees, and by design can produce 0.25 arcsec images over most of a one degree diameter field. The corrector contains PSK3 and LLF1 glasses that have low UV transmission, while four of the eight surface have an AR coating that cuts off sharply below 3400Å. Transmission of the corrector is nominally 74% at 3650Å, 54% at 3500Å , 31% at 3400Å, 11% at 3350Å and near-zero at 3300Å. Thus, with a suitably blue- sensitive CCD such as the Tektronix 2048, Johnson U band photometry should be possible but Strömgren u photometry will be marginal.
Behind the corrector are two filter wheels and an-off axis pick-off mirror for the CCDTV camera, followed by the CCD dewar. Each filter wheel can hold up to five 4-inch square filters, which must be less than 1 cm thick. Inserts are available for holding 4-inch round, 3-inch square and 2-inch square filters. The Tek 2048 is normally used with 4-inch filters although 3-inch filters used in the upper filter wheel will not vignette. See Appendix I for filter information.
The CCD is placed on-axis. The tilt and rotation of the CCD can be adjusted so that the CCD is accurately perpendicular to the optic axis, and aligned NSEW. The scale is 18.00 arcsec/mm. The CCD is focused by moving the entire PFCCD camera head (including filters and TV camera).
A field approx 26 arc min North of the optical axis is viewed by a Peltier cooled CCDTV camera with leaky memory for acquisition and guiding. The TV is on a scan table which can be moved to acquire a guide star, and in the future will be used to implement short-scanning. The TV has its own filter bolt with 4 positions - the TV does not image through the CCD filter bolts. On moonless nights stars as faint as V = 21 can easily be seen in a few-second exposure.
All hardware functions in the CCD camera head (with the exception of the CCD shutter and preflash) are controlled by a STD bus computer (actually a small stand-alone PC with non-volatile RAM disk). The operation of this computer should be transparent to the user, and all motor commands are initiated from the Arcon-IRAF user interface. The dewar tilt and rotation will have been adjusted at the start of the run and should not need to be altered. The telescope focus control and readout are available to the data acquisition program. This means that when a filter is changed the telescope and TV camera will automatically refocus according to tables of offsets entered in the wheel1 and wheel2 parameter sets. Focussing, either by the classical method of moving the telescope between positions, or by moving the charge on the CCD (preferred), is an automatic procedure.
Below we describe recommended observing procedures. You are recommended to
use IRAF for reducing your data, and for writing your frames to dat tape or
exabyte.
Information on the use of CCDRED, and IRAF in general,
is given in the manuals and cookbooks kept near the SUN terminal. Observer
Support personnel and the Data Reductions Specialists can assist you with
IRAF. But you should not go to the telescope and expect to learn IRAF at the
same time as you observe!
THE MANUAL
"THE ARCON-IRAF INTERFACE"
DESCRIBES OBSERVING SOFTWARE AND PROCEDURES. YOU SHOULD REFER TO THIS MANUAL
FOR THE MECHANICS OF HOW TO TAKE DATA. Only a brief outline is given
below.
2. You should check you know where your home directory is (show home)
and where your pixel files are going (show imdir). Check that you have
lots of disk space (disks). Make sure that when you do an
observe that an IRAF image ends up where you expect it do be. Display
an IRAF image, using display. See below for instructions on how to
check whether your CCD system is functioning correctly.
3. Some of the CCDs operated by Arcon have more than one (either two or four)
output amplifiers available. The CCD is read out through each amplifier in
parallel, the data is then digitized through (1,2,4) 16-bit analog-to-digital
converters and then transmitted on a single serial fiber-optic line to the SUN
computer. A raw bi- or quad-amplifier picture looks a little unusual. Each
quadrant has a slightly different electrical offset (ie overscan level), and
the four overscan segments are in the center of the picture. A CCD read out
through two amplifiers can either have two overscan strips at the side of the
picture (parallel split) or two parallel strips in the center of the picture
(serial split). The real time display automatically removes the overscan and
applies an offset to each quadrant to normalize the DC level. In addition,
the real time display colors any saturated (ie 65535) pixel in red.
4. "Quad" CCD pictures are trimmed and overscan-subtracted in a separate pre-
processing task called quadproc (in the quad package).
Note that even after the pictures are trimmed and overscan subtracted the sky
levels in each quadrant will not be identical (unless sky level is zero).
This is because each of the four CCD amplifiers has a slightly different gain.
This can be measured using gfind (contributed package). Gain
variations between different amplifiers are removed when the object frames are
divided by a flat field frame.
5. There should be an exabyte drive, a 9-track drive and maybe even a DAT
drive available in your dome. The Arcon produces 16 bit data of type
'unsigned short integer' with valid data ranging from 0 to 65535. Thus the
unprocessed frame from a 2K CCD requires about 8.8 Mbytes.
It is your responsibility to check that the instrument is operating to your
satisfaction. A few tests are suggested below which should aid you in this
evaluation. You should also be on the lookout for any changes in the CCD
characteristics during the run, which may signal the onset of CCD or dewar
problems. Count rates from dome flat field (white spot) exposures should be
the same from day to day, to better than 10 percent. You should also divide
the night's flats by the previous night's and check to see that the division
is quite flat, with perhaps only the addition of one or two "rings" due to
specks of dust falling on the filters (ask for the filters to be cleaned if
these seem excessive, but don't try to use your old flats later on). Examine
zero (bias) exposures, they should not show horizontal or diagonally slanting
noise "stripes", which move position from frame to frame. Sometimes very low
level noise features are seen on zero frames, before complaining check whether
these are at a significant level to affect your data. Use of the IRAF
imhist command is also recommended. Histograms should exclude bad
columns, edges etc., and should be smooth with no missing codes.
It is worthwhile making a few test exposures before proceeding further. These
tests can be repeated as desired throughout your run. All can be done during
the daytime.
shutter delay time (s) = (C1/C2) - 1.
You can easily do a contour plot which is probably the best way of displaying
the result. You could also in principle correct all your data (see Stetson,
in "Highlights in Astronomy, 8, 635, 1988).
Shutter times less than one second are possible, but below about 0.1s shutter
operation is unreliable. However these very short shutter times are useful
when checking the telescope pointing using a very bright star, for instance.
Non integral shutter times are possible, and are correctly recorded in the
image header. They will most likely be seen when using pause and
resume.
3.4 Calibration Frames.
The following calibration exposures are essential so that your data can be
properly reduced. Consult the CTIO IRAF manual for full details.
Most of the calibration frames can be taken during the afternoon. Almost all
observers prefer to take calibration frames each day, and we also recommend
this, but it is not absolutely essential. Observer Support personnel will
help you to get set up if required. If you have special requirements (dome
flats in the morning etc) please consult with Observer Support in advance so
that arrangements can be made. In general, it is assumed that astronomers
will not wish to use the telescope for calibrations etc from approx 08h00
until at least 14h00. Note that in order to ventilate the domes so as to
minimize "dome seeing" it is normal for the shutters to be opened
approximately one hour before sunset. Please try to finish your calibration
frames prior to this time.
3.4.1 Zero (Bias) Exposures.
These zero exposures can be started using the commands zero or
observe. You should obtain between 5 and 25 exposures, if you
are doing really low S/N photometry through narrow band filters then you
should aim for the higher number. These can be combined into a single frame
with an "average sigma clipping" algorithm (IRAF), the idea being to produce a
"zero level" frame from which any cosmic ray events have been removed. You
will get plenty of cosmic rays on the Tek 2048 frames due to the long readout
time, so make sure you have at least 5 zero frames for this CCD. Combine
these using zerocombine which you will find in the quad
package.
3.4.2 Dome (white spot) Flat Fields.
These exposures map the sensitivity variations of the CCD, and since the
response alters as a function of wavelength it is necessary to prepare a flat
field for each filter. Ultimately, the quality of your data will depend on
how well you prepare your flat fields. You can use the dflat or
observe commands to take dome flats. Broad band (eg BVRI) dome
flats require use of a color balance filter. Do not use the color balance
filter for U, or for medium band (eg Strömgren, DDO) or narrow band
filters. You should aim for flat levels of several thousand (say 5000 -
40000) counts. Take a sequence of several exposures per filter.
The resulting frames can then be averaged to produce a single dome flat field
frame for each filter. Data frames should successfully be flattened with dome
flats to 1-2 %, perhaps a little better for redder passbands such as V,R,I.
To do better than this there are two alternatives -either to use twilight
flats alone, or to correct the low spatial frequencies in the dome flats by
using a dark sky exposure(s) or with the twilight flats. However you are
advised to take dome flats anyway, just in case the twilight or dark sky flats
are unsuccessful. You may find the dome flats are good enough for your
purposes. NOTE - You can get excellent pixel to pixel statistics with dome
flats, and in some cases this may be more important than low spatial frequency
errors.
3.4.3 Twilight Sky Flat Fields.
Broadband B and U (especially) twilight sky flat fields appear to flatten data
frames better than do dome flats. However they will not have particularly
good pixel to pixel statistics, particularly if you are working near 1 e-/adu.
You can either use a smoothed twilight sky flat to take the low spatial
frequencies out of your dome flats, thus preserving the good high spatial
frequency S/N of the dome flats, or else dispense with dome flats and use the
sky flats directly. Consider taking 2x2 or 4x4 binned frames and later use
them as smoothed corrections to your dome flats in order to correct the low
spatial frequency errors in the latter.
You will need to point the telescope towards the east (HA -3 hrs) in order to
lessen polarization effects (or just leave it at the zenith, polarization
effects seem to be minimal). Do not use color balance filters. As the sky
darkens after sunset, take one second exposures (with U if you are using it)
until the sky is dark enough so that an exposure of a few seconds gives about
20000-50000 counts. Use the preview command, and look at the counts in
the ARCON STATUS window.
You should have time to take several flats in each filter, and if so these can
be combined later (IRAF flatcombine in the quad package). To
make it easier to remove star images from your flats, take the flats with the
TELESCOPE DRIVE ON, and jog the telescope a little after each exposure. If
you wait until it's too dark, then your exposures will be longer and the
chances of getting significant contamination by stars is greater, particularly
in R and I. These will be removed in the combining process as long as you
have 4-5 de- registered frames. Twilight flats should be done in the order
UBVRI and exposures should be at least 5-6 seconds long in order to reduce the
center- corner variation due to the shutter to negligible proportions. Start
the test U exposures at sunset or just a few minutes after. DON'T ZAP THE
THOMSON CCD BY DOING (FOR INSTANCE) I FLATS JUST AFTER SUNSET- this CCD
suffers from after- images!.
3.4.4 Dark Sky Flats.
Exposures of "star-less" high galactic latitude fields can be used to provide
an "illumination correction" to flat fields. There are several options. One
is to take several stepped (e.g. 20 arc sec) exposures of the field then
combine the exposures in such a way (e.g. median filtering) that faint stars
are removed. Since the final frame will be heavily smoothed a sky level of
200-300 ADU over bias is sufficient. Be sure to remove any bad pixels before
smoothing. The resulting smoothed frame can then be combined with the
corresponding dome flat to form a master flat. That is, the sky flat removes
low spatial frequency errors in the dome flat. The advantage of this
procedure over using twilight sky flats is that you retain the good pixel to
pixel statistics of the dome flats, and the color match is improved. In
practice the latter advantage does not seem to be very important. The
disadvantage of using a combination of dome flat + smoothed sky flat is that
any high spatial frequency errors in the dome flats will not be removed,
however in general these are few. A second option is to use the combined,
star-less flats directly. You need lots of counts in the sky level to get
good statistics, so this is really only feasible in the R and I bands. A
third option is to prepare flat fields from combinations of your object
frames. This is generally only possible if they contain few bright stars, and
even then it may be necessary to "patch" out the brighter objects prior to
combining the frames. But if you have lots of relatively long-exposure object
frames then this is a very efficient method.
Note that dark sky flats are recommended so as to allow removal of any fringes
due to night sky emission lines causing interference in the thinned CCDs
(Teks). None of these CCDs fringe excessively (<1% in the I band is typical)
and so you may not need to de-fringe.
With care, you should be able to reduce systematics in your flat fields to
below 0.1 percent, and better over smaller areas. It is possible to determine
sky to maybe 0.01 percent, in sparse fields. To reach this degree of accuracy
requires considerable care, and if you are unfamiliar with these rather
specialized techniques you are encouraged to consult with your CTIO support
astronomer prior to your observing run. (See J.A. Tyson, J. Opt. Soc. Am. A
3, 2131, 1986).
Positions where there are a few less stars than usual:
RA 01 20 22.4 DEC -30 49 04 (1986.5)
Your Observing Assistant is familiar with focussing methods, offset
determinations, operation of the TV and Guider, and will help if required.
3.5.2 Focussing the Telescope. On the first night set
the telescope focus to the "nominal" value. Ask your Observing assistant or
look amongst old Observing Record Sheets to ascertain the relevant value. Try
pedestal (telescope) focus = 27000.
Accurate focus is found by taking several snapshots of the star, each at a
different focus position. Use the focus command, with the relevant
obspars and instrpars parameters set. At the 4-m, the focus mode
should be set to "auto". This will allow multiple
exposures on a single CCD frame. Please save your focus frames and write a
note on the night report sheet saying where they are, they will be written to
tape and kept as a database for seeing and optics improvements. For broadband
filters at the 4m you can just point the telescope at any random field, a 10
second exposure in V will give you several stars suitable for determining the
focus. Offsets for the standard UBVRI sets are known (ask Observer Support)
so you need only focus one filter.
When focusing, always focus against gravity: move out (to larger readings of
the pedestal). The procedure is as follows:
Once rough focus is known, a focus sequence can be done automatically. You
specify the middle = approx focus, plus focus step size, plus number of
measurements to make. Choose either telescope or detector (recommended)
method. A step size of 50-100 units is suggested, use the smaller number if
the seeing is good. The CCD will be read out after the last exposure. The
sequence will automatically do a double space in offset AFTER THE FIRST
exposure (be careful with multi-amplifier CCDs, the orientation of the
sequence may be reversed when comparing the upper and lower halves of the
CCD).
The quickest way to examine your images is to use imexam to plot
profiles. Best focus will be that with highest peak intensity and minimum
distortion in the wings. This last criterion is perhaps the most stringent
one. Make sure you check the focus for several stars on the CCD frame, and
best over-all focus is achieved by using a star some ~ 1/3 of the way center
-> corner of the CCD. The focus will change during the night, as it is a
function of temperature and telescope position. In good seeing (FWHM <1.5")
frequent focus checks are recommended especially if you move the telescope
more than 10 or 20 degrees, thus you should re-focus at your fields for best
results.
3.5.3 Checking the dewar tilts and rotation.
Do this (on your first night) using a focus frame. Compare the focus position
(ie the value of the focus for best image quality) for stars near the four
corners of the CCD. It is best to average results for 2-3 stars at each
position. You should find that the focus variation corner-corner is no more
than ± 100 microns. Since the Tek 2048 CCDs are not flat you should not
waste time striving for perfection! At the time of writing, and with the Tek
2048 #4 CCD (Arcon 3.6) the tilts are set to
xdewarti = 600
for best results. The rotation is set at its nominal value (zero). You can
adjust the tilts and the rotation using setdewar. You can review all
the motor positions including tilts and rotation with motorstat. The
diagrams below represent three different examples of what you might find when
you do a focus run. The
numbers on the diagrams correspond to the focus at that position on the CCD.
Beside each diagram is the prescription for returning the CCD to a position normal
to the focal plane.
3.5.4 TV Camera.
A Peltier-cooled CCDTV camera serves both for field acquisition and for
guiding. Field of view is approximately 4 by 3 arc minutes.
On the CCDTV control panel integration time can
be switch selected and gain/offset controls are available. Try integration
time of 1-2 seconds, gain = 2 and offset so that you see star images.
On the
Leaky Guider (0.9m, 1.5m), hardware windowing allows the user to select the contrast and
brightness of the displayed image optimum for detection and/or guiding. The
autoguider does real time centroiding of a designated star in the TV field.
At the 4m PF, there is a TV filter bolt with four positions, directly in
front of the CCD camera. It contains V,B,R and clear(fused silica) filters
(positions 1-4) of identical thickness.
Focus the TV after focusing the CCD. Watch for smallest images, preferably
with the Leaky Guider switched to "digital". This focus is not that critical,
as the guider will work fine on slightly out of focus images.
Normally the determination of the TV-CCD position offset will be carried out entirely
by Observer Support personnel on the first night only. In any case, in these
days of large CCDs it is more efficient to find objects and center up using
the CCD in preview mode. But if you want to determine the offset, see
the Software Manual.
3.5.5 Guider Operation.
At the 4m, the CCDTV camera digital frame is fed directly into a PC
running a guiding program written by Steve Shectman (Carnegie). This
has a very nice display showing guide errors, etc.
At the other telescopes the CCDTV signal is fed into a leaky guider.
Best results are with a fairly contrasty image, and with the leak time set to
1/2 or 1 second leak. Longer times will not correct for rapid telescope
motion, while no leak results in trying to guide too fast and may result in
oscillation.
Set window first and identify a guide star on the TV monitor. Then enable
only the "rectangular box" cursor. Do NOT try to adjust windowing while
guider is enabled.
The guide box always remains in view on the TV
screen.
The guider, if it gets confused, can confuse you even more. Never try to move
the telescope with the guider enabled. If the guide star vanishes for any
reason (clouds, TV turned down, telescope moving, cursor A turned off), then
the guider will try to center on noise. This could cause very large guide
corrections resulting in unpredictable telescope motion.
3.6.1 Finding Your Object:
The quickest procedure is to point the telescope at the correct position and
take a preview or movie. Alternatively the telescope can be
positioned such that the object is on the TV, and then offset back to the CCD.
Then find a guide star. If need-be, move the TV scan table
(setscantable) and refocus (settv).
3.6.2 Exposure times:
All the CCDs bleed charge down columns, and to a lesser extent across rows,
when an integration is long enough for a star image to become saturated. All
recover well from over-exposure, however excessively bright stars should be
avoided. All the CCDs have very low readout noise so there is an advantage in
dynamic range by splitting a long exposure up into a number of shorter ones.
You should also beware of bright stars just outside the CCD field. These can
easily scatter light onto the CCD frame and make photometry difficult.
Cosmic rays tend to be either single or double hot pixel events. The standard
way to remove them is to do at least 3 exposures (4 or more give better
statistical results) of the same field and median (or use avsigclip option in
the IRAF combine command) filter to remove transient events. The IRAF
cosmicray and imedit commands can be effectively used to remove
cosmic rays from single frames.
You should be careful not to over-exposure your object(s) of interest.
Saturated pixels are colored RED on the real-time display.
3.6.3 Bad pixels:
There are two types: 1) hot (above local background) and 2) cold (less than
local background). Sometimes hot pixels tend to be found at the beginning of a
column of cold pixels - it appears that most of the charge finds its way into
just one pixel. Neither type are transient events: they are both permanent
defects in the CCD. However, the number of bad pixels is a function of
intensity: there are many more at low light levels. The Tek 2048 #3 CCD has
hot columns and large charge traps. These are obvious, and should be avoided.
The Thomson CCDs have about 20 short traps visible only at low background
levels; at higher backgrounds almost all of these shorten to a single low
pixel.
3.6.4 Standard Stars: In broadband UBVRI work
excellent results have been achieved by using standards from Landolt (AJ 104,
340, 1992). There are several fields (eg T Phe, SA 98, SA 110) which contain
several stars in a 3-4 arcmin diameter field with a wide range in color, and
so are invaluable for use with large CCDs. John Graham's uncrowded E region
standards (PASP, 94, 244, 1982) in the magnitude range of 10 to 16, the older
Landolt standards (AJ, 78, 959, 1973; AJ, 88, 439, 1983) and some of the
fainter (10-11 mag) Cousin's stars (J.W. Menzies et al., SAAO Circ., 13, 1)
are useful. Copies of these lists are kept in each of the console rooms.
Look carefully at the quoted errors on the standards before observing them.
Residuals in the photometric solutions are a few percent at worst, and
standard stars (properly done) repeat to 0.005 mag or better from night to
night. It is preferable not to mix Landolt standards with those of
Cousins/Graham if you are interested in getting systematics much below the
0.02 mag level, particularly for stars of extreme colors, and for U-B.
For other systems (Washington, Strömgren, Gunn, etc.) standards are fewer
and less conveniently situated. If you are uncertain, consult your CTIO staff
contact prior to your observing run.
For narrow band imaging one can try using spectrophotometric standards (R.
Stone, J. Baldwin, 1983. MNRAS, 204, 347 and 1984. MNRAS, 206, 241, D.
Hayes, 1970. ApJ, 159, 165, M. Hamuy et al., PASP 104, 533, 1992, PASP 106,
566, 1994).
The biggest mistake made in observing standards is to underexpose: try for a
peak in the star of at least a few thousand ADU above background (exposure
times of 5-20 seconds for V = 12-14 at the 4m). Shutter clocking is accurate
enough to permit 1 second exposures but, as discussed above, you will need to
correct for the shutter delay and for the uneven field illumination.
Use of faint photoelectric sequences (particularly those in crowded fields)
requires caution; these often have proven to have systematic errors, and
require long exposure times in each filter to reach the desired accuracy in a
large measurement aperture.
The care with which one observes standards will be reflected in the accuracy
of one's photometric calibration. It is common to observe several standards
at the start of a photometric night, several more in the middle, and still
more at the end. If you have fields with only one or two standards, try
taking two exposures of a standard field per filter, stepping the telescope by
20 arcseconds between exposures. The extra frame takes very little time, and
insures against your star landing on a bad pixel or cosmic ray. For the
highest accuracy one should definitely observe enough standards at a range in
airmass to solve for the extinction coefficients. Tololo extinction can be as
low as kV = 0.12, or > 0.2 if conditions are dusty, so one should not rely on
the nominal values (for instance in 1991-2 volcanic dust caused elevated
extinctions). The CCD camera has shown itself capable of doing better than 1%
photometry: it seems a shame to skimp on the external calibration necessary to
achieve this. Furthermore, to make proper use of such an accurate calibration
will require the existence in the object frames of uncrowded stars exposed to
the same level as the standards.
CTIO operates several types of CCDs. There is, as yet, no general purpose CCD
and a CCD suited to your program will be the one scheduled. Loral CCDs
require periodic UV- flooding in order to provide enhanced sensitivity below
5000Å, and must be kept cold in order to retain the improved response. The
output signals from CCDs are extremely small (remember 1 electron = 1.6 x
10^(-19) Coulomb!) and the electronics must not allow significant extra noise
to degrade the performance. In case of problems, electronic components can be
replaced (at the board level) and if necessary, the CCD can be changed. If
you are dissatisfied with the performance of the CCD system, please consult
with Observer Support personnel.
At PFCCD the CCD is mounted in a downward looking liquid nitrogen filled
dewar, and is kept at a constant temperature by a small heater. This
temperature is displayed in the console room and should remain constant. If
the temperature appears to be rising immediately alert either your Observing
Assistant or Observer Support personnel. After initial power on and the dewar
has been filled with liquid nitrogen for the first time, it requires at least
4 hours for all temperatures and operating voltages to stabilize. The dewar
hold time is over 12 hours -filling at the start of the night will generally
do for the whole night, except in winter time when, near midnight, the night
assistant will ask for permission to return the telescope to the access
platform in order to fill the dewar.
The analog signal (which can be considered to be some number of electrons)
from each pixel of the CCD is first amplified, then passes through an
integrator which discriminates against various noise sources, and is then
converted into a digital signal by an analog-digital- converter (ADC). Hence
the signal at this stage is measured in "counts" or " adu's" (analog-
to-digital-units). The ratio adu/electrons is called the gain, although by
convention we usually talk about the inverse gain, electrons/adu. The gain is
an adjustable parameter on the CTIO CCD systems. The range of data numbers
possible extends from 0 to 65535 (0 to 216-1, 16 bits).
A summary of the
characteristics of the detectors normally used for direct imaging is as
follows:
1. The full well capacity is the limit above which charge "spills"
out of a CCD pixel into adjacent pixels. At some level below full
well the response of the CCD becomes non-linear to incident light.
Some CCDs retain excellent linearity right up to full well, but
for others the departure from linearity becomes severe well below
full well. This will define the practical maximum charge capacity
per pixel. It may depend on details (slope, amplitude, levels) of
the clocking voltages and for some of our devices there is some
hope of improving the values given.
Typically a factor 20 more charge must be accumulated
before noticeable bleeding occurs. This bleeding is stronger in the column
direction for our CCDs. With normal gain settings the data system limit of
65535 counts is reached prior to entering the CCD non-linear region. Care
should be taken always to operate the CCDs in the linear part of their
response.
2. The readout noise is somewhat dependent on gain, larger values of readout
noise correspond to larger values of e-/ADU. Similarly, the read time is
given for gains of 3-4 e-/adu.
We have two Tek 1024 CCDs, #1 and #2. At the present time only #2
(a quad amp devices) is being scheduled. Read noise is very low. Cosmetics are
superb. QE is poor in the UV and U band photometry is not recommended. Note
that these CCDs do not have the serial registers and amplifier areas shielded
from light, consequently on high light level flat field exposures the
overscans show an exponential decay and during reduction should be fitted with
high-order splines. See the software manual for more details.
We have three Tek 2048 CCDs (#3, #5,#6).These are thinned, AR coated devices
like our Tek 1024s. Read noise is very low, and QE is similar to the Tek 1024's
except the U band response is much improved.
Tek #3 has several column defects, with bright columns on each
side and three blocked columns near the center. These should all be
avoided. Tek 2048 #6 is the CCD normally scheduled at the 4-m PF, while
#3 is dedicated to the 0.9m., #5 is usually at the Schmidt.
#3 & #6 have four low-noise amplifiers and thus have short
read time, typically 30 seconds. #5 has one amplifier noisier than the
rest, type "ccdinfo" to see the characteristics of the CCD you are using.
.
RETIRED!!
The STIS CCD was made by Tektronix as part of the development of the
CCDs for the Space Telescope Imaging Spectrograph. It has slightly smaller
pixels than regular Tek 2048's, and our example is front-illuminated
with a metachrome coating, thus its QE is just like the Thomsons (see above).
The CCD is very low noise, and reads through all two amplifiers.
It has a few column defects. Used as spare.
The observer can control and change several CCD parameters. These are: the
CCD readout format, the binning, the preflash time, and the gain.
Observers should think carefully whether they need all the field (if
not read a ROI) or the resolution (if not, bin 2x2). Even though the
CCDs with quad readout have short read times by big-CCD standards,
substantial gains in efficiency are possible by reducing the format.
(1) Pixel sizes (arc sec)
(2) Field Sizes (arc min per side)
The PFCCD has two filter wheels each holding 5 filters. F1 is the wheel
furthest from the CCD and normally contains only a color balance filter for
use when doing dome flats. F2 usually contains U,B,V,R,I filters. Other
filters can be installed when required. The nominal filter size is 4x4
inches. However 3x3 inch filters in wheel F2 do not vignette CCDs up to a Tek
2048 (49 mm square) in size. We have adaptors to accommodate 3x3 inch
filters.
The (4x4 inch) filters we have available for the 4-m PFCCD are as follows:
1. Take a sequence of 2 bias frames. Make sure that the preflash is set to
zero.
2. Take a sequence of 2 dome flats or, better yet, 2 bias frames with a big
preflash value (giving a total signal of approximately 1/2 saturation).
3. Subtract these frames as follows:
BIAS1-2 = BIAS1 - BIAS2 (use imarith)
4. Use the IRAF imstat task to determine MEAN(FLASH1),
MEAN(FLASH2), MEAN(BIAS1), MEAN(BIAS2), RMS(FLASH1-2), and RMS(BIAS1-2). Note
that it is very important that the stats box be positioned on a region of the
CCD free from traps, bad columns, or any bad pixels.
5. Calculate the gain,
e/adu = (MEAN(FLASH1)+MEAN(FLASH2)) -
(MEAN(BIAS1)-MEAN(BIAS2))
RMS(FLASH1-2)2 - RMS(BIAS1-2)2
6. The readout noise in electrons is
RON(e) = RMS(BIAS1-2) * (e/adu) / 1.41
2.4 HELP!
Observers needing further information should consult the following persons, in
the first instance contacting people who are actually on Tololo at the time:
3. DATA-TAKING PROCEDURES.
3.1 Introduction
The IRAF-based user interface allows observing commands to be sent to Arcon
from within the IRAF cl. This results in a single uniform user interface for
data taking and data reduction and allows the Arcon user to employ features of
the cl such as the parameter mechanism and the history editor. It also allows
advanced users to write cl scripts which freely mix data acquisition and data
reduction operations. We have tried hard to make the user interface look very
similar to "ICE" as used at KPNO. However, users should be aware that it is
NOT identical and parameter lists, etc. will differ.3.2 Starting Up, and Some Miscellaneous Advice
1. You should login on the data acquisition computer using your assigned
username as password. You will use the OPENWINDOWS environment, with
XIMTOOL as the default diaplay. Several windows will open
automatically, including a (blue-colored) IRAF ARCON user-interface window via
which you interact with the CCD. There is also a (brick red-colored) IRAF
reduction window, an ARCON console window, an ARCON status window with
countdown timer, and an XImtool display window. The transputer
boot-up is automatic (watch the ARCON console window). See the Observing
Software Manual for procedures. IRAF should automatically start in the two
IRAF windows. In the IRAF Acquisition window the arcon and the
astronomer packages load automatically, correctly configured for
the PFCCD. You will be asked whether tou wish to synchronize parameters (ie
force the controller to the parameters that are set in the User Interface.
Normally, the answer is "yes". In the
IRAF Reduction window load the arcon and quad packages, and
anything else you want. Back in the axquisition window, the
astronomer package contains all the
commands needed for observing. Some of these are:
3.3 Instrument Setup and Checkout
At the start of your observing run, Observer Support personnel will check that
the CCD is operating correctly, and that all the hardware functions. They
will fit your selected filters into the filter wheels. They will also have
set the CCD gain and binning, via setdetector. See what they have done
by typing ccdinfo. The gain and read noise are calculated using
the gfind task which is in the contributed package. You
will first need to do two low level (bias) exposures, and two high level
exposures (both at the same light level).3.3.1 Zero (Bias) exposures.
These zero illumination exposures should show no noise patterns down to about
the 1 adu level. They should be approx. flat although bright columns will
show up (Tek 2048 #3). Now do an exposure (dome flat) that adds a small
signal. Any exposure with some illumination should show the overscan strip.
Don't forget that the Real Time Display automatically clips out the overscan
so you will need to display the image within IRAF. The overscan strip may
appear down the left side, right side, or center of the picture depending on
the number and position of the CCD amplifiers in use (note that, to avoid
confusion, the orientation of the CCD frame on the Real Time Display (RTD) and
in IRAF does NOT change when a different readout amplifier(s) is selected.
The overscan is on the right hand side of the picture (single channel readout,
right-side amplifier), or down the left hand side of the picture, or down the
center (quad readout). If you are reading through more than one amplifier the
overscan strip will be made up of sub-strips, one for each amplifier. 3.3.2 White Spot exposures.
Ask for the telescope to be pointed at the white spot on the dome, and turn on
the quartz halogen lamps. Make sure the telescope focus is approximately
correct for the filters you will be using during the night. Put the color
balance filter in the beam. This filter makes the 3000 C quartz lamps look
more like the 5500 C night sky. Take a trial exposure, typically a few
seconds through a broadband (B,V,R,I) filter. Display the frame, and do a row
plot across it. The response should be reasonably flat (say ±10 percent in
general). After trimming and subtracting off the zero (bias) exposure (use
quadproc) you can try dividing frames at two different count levels
whereupon all the structure seen in the dome exposures should disappear to
better than a few 0.1%. For CCDs with more than one amplifier in use, the
"quadrant" edges should all disappear, to better than ~0.2 percent. Note that
in general comparisons between parts of the CCD read out via different
amplifiers is a powerful test for correct functioning of your CCD system.
ARCON has separate data paths from the CCD until after the ADC, and it is very
unlikely that a controller fault would affect each data path in the same way.
In the above case we are testing for differential non-linearity, and can
detect it far easier than we can measure absolute linearity, which for our
CCDs is also at the 0.1-0.2 percent level. 3.3.3 Dark Exposure.
A five minute dark exposure (lights off in dome) should only show the hot
columns (Tek 2048 #3) and a few cosmic rays and hot pixels. The general
background should be at the same level as the overscan. For all our CCDs, the
dark rate is so low in comparison to sky background rates as not to warrant
doing long dark exposures, however if you insist, you should do several dark
exposures at least as long as your longest object exposure. Hot pixels tend
to be non-linear in their dark rate as a function of time so may not subtract
out perfectly. 3.3.4 Shutter Timing.
The shutter unit has a 100mm clear aperture. The effective "open" time is
always longer than the nominal number of seconds requested for your exposure
time. This is a delay prior to opening and closing. The speed of the shutter
blades is high and the corner-center exposure difference is only a few ms for
the smaller CCDs. For the Tek 2048 the center-to-corner difference is some 60
ms (ie exposures are 60 ms longer than nominal at the center, and 0 ms longer
at the corners). So beware when doing short exposures for flat fields and
standards. Sometime during your run, it is a good idea to determine this
correction. The easiest way of doing this is to compare (for example) a 20
second dome flat with the sum of 20 1s dome flats. The latter can be done
extremely quickly using the focus procedure but not bothering to move the
focus (or the telescope), whereupon you can avoid reading the CCD out 20 times
but the shutter will have opened and closed 20 times. Bracket this exposure
with normal 20s exposures (just choose a filter and exposure time to give
several thousand adu total counts), the bracketed exposures can be averaged to
take out any drifts in the lamp output, but in any case it is wise not to
begin this test until several minutes have elapsed after turning on the lamps.
Subtract bias off the two frames, then if the counts are C1 for the
sum of the 20 1s exposures, and C2 for the single 20s exposure, then to first
order we have
RA 12 56 07.7 DEC -76 56 39 (1950)
RA 16 49 42 DEC -15 21 00 (1982) 3.5 Night Time Procedures.
3.5.1 Start of the night procedures.
About an hour before sunset the dome shutters will be opened unless
calibration frames are still being taken. All doors should be closed and
lights turned off. During twilight, your Observing Assistant will fill the
dewar with liquid nitrogen, open the ventilation doors, and check the
telescope pointing on a bright star near the zenith. If it is the first night
that the CCD has been installed then the offset between the CCD and the
acquisition TV camera will also be determined, and the focus of the TV camera
checked.
ydewarti = -400
E
______________________
|26500 26200|
| |
| |
| | Problem 1: tilted about E-W axis
| | N
| | Solution: increase X by 1000 microns
| |
| |
|26500 26200|
|______________________|
E
________________________
|26200 26200|
| |
| | Problem 2: tilted about N-S axis
| | N
| | Solution: decrease Y by 1000 microns
| |
| |
| |
|26500 26500|
|______________________|
E
________________________
|26300 26200|
| |
| | Problem 3: tilted about both axes
| | N
| | Solution: increase X, decrease Y,
| | each by 1000 microns
| |
| |
|26500 26300|
|______________________|
You are cautioned that if the PFCCD motor-controller has its power cycled, or
is manually reset, or the commands motorinit or motorreset are
executed, then the tilts
will be set back to their nominal (ie zero) positions. You will then need to
reset the tilts using setdewar. The same is true for the other PFCCD
motors, but their positions are generally rather more obvious and not so
damaging to the data. Positions for all motors can be displayed by the command
motorstat. If the Arcon software needs to be re-loaded, then answering
"yes" to the question re synchronizing the controller will force both controller
and the instrument to the configuration set by the User Interface parameter files
(obspars, instrpars).3.6 Observing.
4. DETECTOR CHARACTERISTICS
4.1. General Principles
4.2 Detector Options
Tek 1024 Tek 2048 STIS 2048
Pixels 1024x1024 2048x2048 2048x2048
Pixel size (microns) 24 24 21
Pixel size (arcsec) 0.3 0.43 0.38
Field size (arcmin 7.3x7.3 14.7x14.7 12.9x12.9
Readout noise (electrons) 4-6 3-5 3-5
Electrons/ADU (typical) 1-4 1-4 1-2
Cosmic ray rate (per min) 20 100 ?
U: Sensitivity at the 4m 10 30 20
B: ( U=B=V=R=I ) 180 239 90
V: ( = 20 mag ) 350 348 150
R: ( star, in ) 380 380 200
I: (electrons/sec) 200 205 130
Read time (quad, secs) 12 30 30
Notes:4.3 Detector Notes
4.3.1 The Tek 1024 CCDs
4.3.2 The Tek (SITe) 2048 CCDs
4.3.3 The Thomson 1024 CCDs
4.3.4 The STIS 2048 CCD
4.4 Detector Control Options
4.4.1 CCD Readout Format
A single region-of-interest can be read out, positioned at an arbitrary place
on the CCD. These operate very efficiently, for instance a 1024x1024 ROI
centered on a Tek 2048 takes under 10 seconds to read out, while a 256x256
takes just 2 seconds.4.4.2 Binning
The ARCON controllers are able to bin pixels in various ways. The readout
noise per pixel remains the same as in the unbinned case. The gain
calibration is unchanged from the unbinned mode. For most direct imaging
applications at CTIO the CCDs tend to undersample the images, so there is
little advantage in binning. The main advantages of binning are that the
readout time is shorter, and less data is created.4.4.3 Preflash Time
Default is zero. All our CCDs operate without preflash. At the 4-m PFCCD the
preflash illumination is rather uneven and thus of little use.4.4.4 Gain
Generally you will find for direct imaging a gain of about 3-5 e-/adu is a
good compromise between sampling the read noise and achieving good dynamic
range. Excessively high gain (ie small e-/adu) does little good since
digitization noise is soon masked by CCD readout noise and photon statistical
fluctuations. Different CCDs are somewhat different in their photon/adu
conversion, noise and points at which non-linearities begin to occur. The
method used to calculate the gain is described in
Appendix II. 4.5 CCD scales at various foci
Tek Tek Thomson STIS
1024 2048 1024 2048
4-m f/2.85 0.43 0.43 0.34 0.37
4-m f/7.5 0.16 0.16 0.13 0.14
1.5-m f/7.5 0.44 0.44 0.34 0.37
1.5-m f/13.5 0.24 0.24 0.19 0.21
0.9-m f/13.5 0.40 0.40 0.32 0.35
Schmidt f/3.5 2.0
Tek Tek Thomson STIS
1024 2048 1024 2048
4-m f/2.85 7.34 14.7 5.81 12.9
4-m f/7.5 2.70 5.40 2.13 4.73
1.5-m f/7.5 7.41 14.8 5.87 13.0
1.5-m f/13.5 4.09 8.18 3.25 7.15
0.9-m f/13.5 6.76 13.5 5.35 11.8
Schmidt f/3.5 68
APPENDIX I:FILTERS FOR CCD IMAGING
A
filter list
is available and can also be found in
NOAO Newsletter
No. 34. A complete CTIO filter list is also available via ftp. See the WWW
home page.
APPENDIX II: GAIN AND READOUT NOISE CALCULATION
The gain in electrons/adu is usually accomplished using the gfind
command (in the contributed package). Here is a cookbook recipe for
determining the gain in electrons/adu, for anyone who wants to do all the
steps one by one.
FLASH1-2 = FLASH1 - FLASH2
Footnote 1
At time of writing, use smcinit
instead of motorinit and instrument instead of
motormove.