The TIFKAM User's Manual
Contents
- Introduction
- System Overview
- System Characteristics and Performance
- The HAWAII-1R HgCdTe Detector Array
- Observing Setup
- Observing Techniques
- Calibration
- Additional Information
Introduction
TIFKAM is a multi-purpose infrared imager and spectrometer designed for
operation at 0.9-2.5 micron wavelengths. The instrument currently
uses a 1024x1024 HAWAII-1R HgCdTe detector array built by Rockwell and
purchsed for TIFKAM by Dartmouth using funds provided by an NSF MRI
grant. Selectable re-imaging cameras provide a choice of two plate
scales and a variety of spectroscopic resolutions. Internal mechanisms
control the selection of filters, dispersers, focal-plane masks, and
cameras. All mechanisms are user controlled and can be used to rapidly
reconfigure the instrument, which allows for substantial observing
flexibility.
Why "TIFKAM"?
When originally designed, the instrument was named "MOSAIC", an acronym
for "MDM/Ohio State Active Infrared Camera". The intent was to have an
internal tip-tilt system to correct out first-order image motion
("active optics") to improve image quality. The tip-tilt system was
never implemented. We were then loaned a 512x1024 ALADDIN InSb array by
Kitt Peak National Observatory, in return for letting the KPNO community
use the instrument at the KPNO 4m and 2.1m telescopes. The problem was
that MOSAIC was also the name of the 8Kx8K CCD imager at KPNO, which
caused confusion. Mike Merrill at KPNO suggested the name "The
Instrument Formerly Known As MOSAIC", hence TIFKAM.
Aren't you glad you asked?
System Overview
TIFKAM is a general-purpose 0.9-2.5 micron imager and
moderate-resolution spectrometer built as collaboration between Ohio
State and the MDM Observatory. The instrument uses a 1024x1024 HAWAII-1R
HgCdTe detector array. Internal cameras provide three choices of pixel
scales (see below) or a pupil-viewing camera.
There are two 9-position filter/grism wheels in TIFKAM providing a
variety of filters and spectroscopic modes. There are 4 grisms
available, in approximate order of increasing dispersion:
- JHK grism - all of J thru K in one spectrum
- J+H grism - all of J thru H in one spectrum
- J/K grism - either J or K using order-separation filters to select
- H grism - H-band only
The population of the instrument filter,
slit, and camera wheels changes occasionally, but generally includes standard
broad band imaging filters (i.e. JHK) and a selection of grisms and slits.
The Light Path Through the Instrument The
control interface for the instrument is the Prospero package
written at OSU. More information on Prospero can be found on the
Prospero web page.
Please note that Prospero is not an image processing package,
but an interactive instrument control system. Simple commands for
inspecting images are included in the Prospero command suite, but
the observer is expected to use their favorite package to fully reduce
and analyze their data (e.g. VISTA, IRAF, IDL, FIGARO, MIDAS, ZODIAC,
HP-65, ABACUS, FINGERS). A useful summary of the Prospero
commands pertinent to TIFKAM is provided in the form of a 2-page
quick-reference card.
The instrument has many operational modes and has not been
extensively tested in each. There are also several known problems with
the instrument which we note below. Our intent is to work on them as
soon as possible, usually the next available opportunity to warm up and
open the dewar. For historical interest, we also provide a list of
problems the instrument has had in the past.
System Characteristics and Performance
Imaging
TIFKAM was designed to deliver good image quality over the entire field
of view of the 1024x1024 detector array with reasonably high throughput.
TIFKAM has three camera lens sets (f/5, f/7.5, and f/16.6) with
the following pixel scales:
Camera |
2.4m Hiltner |
1.3m McGraw-Hill |
f/5 | 0.30 "/pix | 0.55 "/pix |
f/7.5 | 0.20 "/pix | 0.37 "/pix |
f/16.6 | 0.05 "/pix | 0.09 "/pix |
A fourth camera is used to view the pupil and rarely used by observers.
The f/16.6 camera is primarily intended for spectroscopic use or to take
advantage of periods of unusually good seeing at the 2.4m. Most
observers taking direct images will use either the f/7.5 or f/5 cameras.
Spectroscopy
Long-slit spectroscopy is enabled by grisms inserted into the beam by
the filter wheel. Below is a table that summarizes the resolutions and
wavelength coverages of the various possiblities.
Grism | Blocking Filter |
Resolution (2 pixels) |
Wavelength Range (nm) |
Illuminated Rows |
J/K | J | 1450 | 1110 - 1360 (2nd order) | 400 - 982 |
J/K | K | 1325 | 1970 - 2420 (1st order) | 160 - 704 |
H | H | 1150 | 1480 - 1800 (2nd order) | 470 - 918 |
J+H | IJH | 720 | 950 - 1910 (1st order) | 1 - 900 |
JHK | JHK | 750 | 1220 - 2490 (usable to 2290; limited by blue leak) | 1 - 1024 |
Note that the actual wavelength coverage will depend on the filter
selected in the other filter wheel (i.e. if using the J/K grism and the
K filter as a blocker, the wavelength coverage will only be 2000 to
2400nm due to the transmission of the filter); we are currently trying
to obtain special filters to use specifically as spectroscopic
blockers. The table above gives the resolution assuming the 54-micron
slit is also inserted into the beam; this corresponds to roughly two
pixels FWHM for unresolved lines. There is also a 100 micron slit that
will give four pixels per resolution element and correspondingly lower
spectral resolution. Note that this slit is recommended when the seeing
is poor.
The HAWAII-1R HgCdTe Detector Array
TIFKAM Detector Characteristics
Full Format | 1034x1024 |
Pixel Size | 18 microns |
Active Region | [11:1024,6:1019] (1014x1014) |
System Gain | 2 electrons/ADU |
Read Noise | 15 electrons RMS |
Full-Well Capacity | ~105 electrons |
Minimum Integration Time | 4.29 s |
The HAWAII-1R array has a series of reference pixels appear as a dark
strip around the outside margins of the raw image. While these look
superficially like CCD overscan pixels, they are instead physically
separate pixels in the CMOS readout multiplexer that are not connected
to the HgCdTe detector layer, and insensitive to IR light. The actual
active area of the detector is 1014x1014 pixels with 5 lines of
reference pixels above and below the active area veritically and 10
columns of reference pixels on either side.
Dark Current and Read Noise
The dark current in the array is low. We currently measure <0.4
ADU/pix/sec dark current, some of which may be due to background
radiation in the dewar proper. The dark current will be higher if the
instrument has been cold for less than 36 hours.
The read noise of the array is ~15 electrons.
Maintenance
The only maintenance task required during an observing run is to keep
the LN
2 topped off. Our experience is that filling the
instrument once per day is sufficient.
Linearity and Saturation
The array becomes significantly non-linear before the full-well capacity
of the detector is reached. Further, an accurate non-linearity
correction depends on the signal rate as well as the total signal
collected. However, non-linearity is <1% for <11,000 ADU at signal
rates ranging up to ~600 ADU/sec (about the background rate at K on a
warm night with the f/7.5 camera). We are working on obtaining more
information about the reproducibility and signal-rate-dependence of the
non-linearity; we currently suggest that interesting signals be kept to
<8000 ADU so that non-linearity corrections should be <0.2%.
There may be a small offset between the integration time requested by
the observer (and entered in the image header) and the actual time
interval between the two reads of the double-correlated sampling cycle.
This will result in an apparent non-linearity in the measurement of a
constant signal as a function of integration time. Until we understand
this better, we advise against the use of bright standard star
observations at very short integration times to photometrically
calibrate object fields at much longer integration times.
Hard saturation of the detector array occurs at about 25,000 ADU. The
array becomes seriously non-linear (>5%) at about 17,000 ADU.
Bad Pixels
There are many dead, hot, unresponsive, and just plain bad pixels on the
array.
After Image or Residual Charge
The detector array in TIFKAM exhibits a residual signal from bright
sources. The magnitude of the residual image seems to depend on the
brightness of and total signal recorded from the source. For typical
observations the residual will be 0.5-2% of the originally detected
signal. Reading the array several times reduces the magnitude of the
residual image to <<1% of the original signal.
We suggest that if residual images will seriously compromise your
results, you dither the telescope faithfully and, perhaps, read the
array several times (set the exposure time to 0 seconds, execute an
mgo 3 command, and throw those images away) between each
science exposure.
Observing Setup
Focusing
Both the internal optics and the telescope should be in focus for the most
optimal images.
The internal optics can be focussed by inserting one of the
spectroscopic slits into the beam and then running the camera focus
through a range of values. Approximate values (June 2008) are
f/7.5 - camfocus 1340
f/5 - camfocus 2730
f/16.6 - camfocus 400
The focus curve for the f/16.6 camera is very shallow, so if remeasuring
you should use big steps (200) in camfocus, whereas for the faster cameras
steps of 50-100 are best.
For spectroscopy, we recommend that the optimum camera focus be
determined using a calibration line source. The value of the the
various camera foci do not change unless the detector array is removed
or the optics are realigned, both of which require opening the dewar
and are (hopefully) rarely done.
The telescope can then be focused in the normal manner. We generally
find a rough focus by starting movie (which reads out the
array continuously) and rapidly running the telescope focus in and out
until the image looks good, then taking a series of images at different
focus settings.
Note that all the filters and prefilters are in near-collimated light,
so there is no change of optimal focus with wavelength.
Observing Techniques
Imaging Observations
There are five (5) mechanisms that you control from within the Prospero
program: a slit/focal plane mask wheel ("slit"), two filter wheels
(called "filter" and "prefilt"), the choice of camera ("camera"; which
sets the imaging plate scale and spectral resolution), and the internal
focus position of the camera ("camfocus"). The slit, filter, and camera
selections can be most conveniently found by typing
print
slit or
print prefilt etc., to list the populations of
that particular mechanism.
The command syntax for moving these mechanisms is
[mechanism] [value]
For example, to set up for J band imaging with the PK-50 blocker,
for example, one would execute
prefilt 2
filter 6
in any order.
The slits are roughly 480 pixels long. The 54-micron pixel slit is
slightly rotated with respect to the array, by about 2 pixels over the
length of the slit, whereas the 100-micron slit appears aligned with
respect to the array to within 1 pixel.
There are three ways of taking exposures:
- go takes a single exposure
- mgo n takes n integrations and writes n frames to disk
- avego n takes n integrations and writes 1 averaged frame to disk
The array becomes significantly non-linear beyond ~8000 ADU that it is
advisable to keep the integration times short enough to keep the
background less than this level. At this point the background noise
overwhelms the read noise anyway. The upper right quadrant seems to
become non-linear at a somewhat lower level than the rest of the array.
Spectroscopic Observations
Use imaging mode to acquire spectroscopic targets, offset the telescope
to move the target onto the array location of the slit (which can be
found by taking an image of the slit against the night sky; without the
grism in place of course), acquire a guide star to "lock" the telescope
onto the object, put in the slit, check that the telescope did not move
while acquiring a guide star by taking another image (and correct the
position if it did move), put in the grism, and start taking data. Note
that the first few spectra may have residual image artifacts due to the
previous imaging observations. Executing several short "dummy"
observations prior to taking science data will greatly reduce the
magnitude of these artifacts. We recommend taking spectra of point
sources at several positions along the slit to provide sky measurements
and to aid in removal of systematic effects such as bad pixels which may
occur in any one spectrum. Because the slit is so narrow, we strongly
recommend the use of the autoguider at both the 2.4-m and 1.3-m
telescopes.
NOTE: Although the grisms are in the nominally collimated beam,
we have observed that the lines from a comparison source are in best
focus at a camfocus value of approximately 250, whereas a direct image
of the slit is best focused at a value of approximately 100. We
recommend that the optimum camera focus for spectroscopy be empirically
determined using a comparison lamp spectrum.
Telescope Focus
There is a fairly strong focus dependence of the telescope on the
ambient air temperature at both the 2.4-m and 1.3-m telescopes, and the
1.3m telescope can go out of focus at large zenith distances.
Prospero Scripts
Although
Prospero commands are normally entered from the
terminal, it is possible to execute a list of commands stored in an
external text file. These scripts make it possible to execute complex
or repetitive observing sequences, including loops and conditional
branching from the vocabulary of individual
Prospero commands.
These may be written and stored on disk prior to observing. The
Prospero Command Procedure
Scripts Manual provides comprehensive coverage of this subject.
Calibration
Standard Stars
Photometric Standards
The best near-infrared standards are those defined by Elias et al (1982,
AJ, 87, 1029), but these stars are all too bright to observe with TIFKAM
without neutral density filters (i.e. they saturate the detector array
in the minimum available exposure time). There are several sets of
fainter standards including those measured by Carter & Meadows, the
UKIRT Faint Standards, and a set of stars being measured to support
NICMOS. The Carter & Meadows measurements appear to be excellent
quality, but the stars are relatively bright (e.g. K = 9-10 mag) and may
not be observable with TIFKAM without a neutral density (ND)
filter. Since we do not know how "neutral" an ND filter is in the near
infrared, this makes this set of standards somewhat problematic. Note,
however, that an ND filter is currently installed in TIFKAM . The UKIRT
standards are probably fine for measurements requiring no better than 5%
accuracy or so.
We suggest that the list of standards prepared
for NICMOS (Persson et al. 1998, A.J., 116, 2475) be used when the
most accurate photometry is desired.
Spectroscopic Standards
The near-infrared region has many strong absorption features due to
various molecules in the atmosphere. One of the best ways to remove
these features is to ratio the spectrum of the target object with the
spectrum of a featureless source observed with the same instrumental
setup and airmass. In the J and K band, A stars provide good atmospheric
standards, since they have only H absorption features at 2.17 microns
(Br-gamma) and 1.28 microns (Pa-beta). In the H band, A stars have many
Br-series absorption features that make them somewhat problematic. Note
that Kurucz models of A stars actually seem to reproduce the
near-infrared spectrum of these stars reasonably well, so these may be
used to correct for the intrinsic absorption in the stellar atmosphere.
Wavelength Calibration
In the JHK region, the OH airglow lines which are such a nuisance do
have the virtue of providing a useful grid for spectral calibration.
However, at both the 2.4-m and 1.3-m telescopes, the spectral comparison
sources built into the MIS boxes provide an excellent Ar spectrum, with
a few He and Ne lines thrown in. One may obtain a high S/N calibration
spectrum with 30 to 60 sec exposures (followed by an equal dark
observation for bias subtraction) using the lamp sources, so this
technique is recommended.
Flat Fielding
The HgCdTe detector array is reasonably flat; we have measured ~1%
photometric accuracy for standard stars (at K) without flat-fielding the
data. The best flat fields seem to be made by observing the dome flat
field screen with the lights on (at a low level in imaging mode) and
with the lights off, forming the flat from the difference of the two.
This technique seems to remove any significant scattered light or
thermal emission component from the resulting flat.
Additional Information
Web Pages:
TIFKAM Web Page:
- http://www.astronomy.ohio-state.edu/MDM/TIFKAM/
This site includes descriptions of the instrument (with pictures), and
access to copies of all of the documentation.
Prospero Support Web Page:
- http://www.astronomy.ohio-state.edu/~prospero
This web site includes a searchable online manual for the Prospero
package, and copies of all of the documentation. A help line and other
services will be included in the future.