3. The Mosaic Hardware
3.1. The Mosaic
CCDs
The Mosaic Imager features eight 2048 (serial or pixels/row) x 4096 (parallel or pixels/column) 15 mm pixel CCDs arranged as an 8192 x 8192 pixel detector. The Mosaic-1 CCDs are read out through a single amplifier per chip simultaneously to 8 controller inputs (on 4 Arcon controllers). For Mosaic-2 (at CTIO) 16 amplifiers, 2 per CCD, are working and can be readout simultaneously. Unfortunately, two of the CCDs in Mosaic-1 only have one working amplifier, limiting Mosaic-1 to work in 8 channel mode. The resulting mosaic array is a square about 5 inches on an edge. The gaps between CCDs are about 0.7 mm in the row direction and 0.5 mm in the column direction. [See Figure 3.1.1 showing an image labeled with chip numbers.] The Mosaic Imager is populated with thinned, AR coated SITe CCDs. These chips have only minor flaws that have little effect on their scientific performance, but they do require careful calibration to attain excellent flat-fielded images.
Figure 3.1.1: A Flat-field (R band) map of the CCDs in the Mosaic with their 'extensions' (im1, im2, im3, ?) as used in the IRAF nomenclature (see section 5).
Figure 3.1.2 shows the average QE for the 8 CCDs. Individual CCDs deviate from this curve as shown in Figure 3.1.3.
Figure 3.1.2: The average QE for the 8 SITe CCDs in Mosaic I.
Figure 3.1.3: The QE differences relative to the average for the 8 SITe CCDs in Mosaic I. Also shown are the readnoise (RN) and gain (GN) values for each CCD.
3.2. The Dewar
The Mosaic dewar is a large (6.3 liter) vacuum vessel radiatively coupled to the CCD mount. It is the large round cylindrical object in the center of Figure 3.2.1. The hold time of the dewar is 17 hours. It is filled by an observing technician at the start and end of each night.
Several temperatures within the dewar are monitored and displayed in the Mosaic Graphical User Interface (GUI).The CCDs should be between -83 and -98 C. The dewar tank is cooled to -166 C or cooler. A good way to detect the exhaustion of LN2 is the warming of the 'fill neck' temperature, which is normally near 0 C. If this temperature begins to rise much above zero or any of the temperature boxes turn and stay red, call for assistance to have the dewar filled. Mosaic automatically e-mails mountain personnel if various temperatures rise to levels that indicate a warm-up.
Figure 3.2.1: The Mosaic system mounted at the back of the 0.9-m telescope. The dewar containing the 8 CCDs is the silvery, cylindrical object in the middle. It is surrounded by the filter track, which is housed in the large black oval that extends horizontally across most of the picture.
3.3. The Arcon
Controllers
The eight CCDs are read out through four Arcon controllers. These controllers run at 100 kpix/sec per CCD, yielding a readout time, including all overheads, of~2min34sec for Mosaic-1 (only a single amplifier is used). Data values are stored as 16-bit unsigned integers. For further details about the Arcon controller systems, see the technical paper by Roger Smith (http://www.ctio.noao.edu/instruments/arcon/arcon.html).
The data taking computer (rush or rust) is a 125 Mhz Sun Sparcstation 10 running SunOS. It has sufficient resources to manage the data acquisition, but not much more. The large data volume is handed off to the reduction computer (tan (KP4m) or emerald (W0.9m)) via Fast Ethernet for all analysis and reductions, thereby relieving rush/rust of unnecessary loads. These are fast Linux boxes with >100 Gbytes of disk and >1 Gbyte of RAM.
Figure 3.3.1: A view of the Mosaic system on the 0.9-m telescope indicating 2 of the 4 Arcon controllers. Also the south guide TV housing can be seen, as well as the Mosaic dewar (now appearing in black).
3.4. The CCD
Shutter
The shutter consists of a pair of opposing sliding blades, one of which has rectangular slots open for the guide field. The blades are attached to pneumatically driven cylinders to provide very fast control of the shutter. This design allows the TV guide fields to be shuttered independently of the science field. In the guide mode, the closed shutter still allows the TV guide cameras to see the sky. In the dark mode, these fields are closed as well. The acquisition software controls which mode the shutter remains in between exposures. For object observations, the shutter goes to the guide mode before the exposure begins. For requested observation types of dark, flat, or zero, the shutter goes to the dark mode before the exposure begins (and remains in this mode after the exposure and readout are completed).Note that the TV fields are always open when the shutter is open; the different shutter modes only control the TV fields when the science shutter is closed. If you have been taking darks, flats, or zeros, you may need to set the shutter mode to guide in order to get light to the TV guide camera.
The time for the blades to move completely across the field is 23 msec. The motion of the blades during both opening and closing are in the same direction so that the exposure level is nearly constant over the array. The motion of the shutter blades is along columns. The accuracy of the shutter has been measured to be ~3% in a 1-second exposure (that is, the exposure is really 0.97 seconds; see also Section 8).
3.5. The Filter
Track
The filter track holds 14 filters. For each filter position, there is a filter for the CCD field, and two separate filters for the two TV fields. Separate filters are used so that a narrow bandpass science filter does not constrain the observer to find very bright guide stars. Normally, one would use clear (BK7) filters for the TV, but one can use a red filter to minimize moonlight or match the science filter more accurately. One might want to match the science filter, at least approximately, to minimize a guider drift. Even at the 4-m, residuals after the correction from the ADC are of order 0.1-0.2 arcsec. This, and all filter decisions, must be made ahead of time, as the filters can only be changed during the day by a qualified observing technician.
Adapters exist to allow the use of 4-inch-square (1 adaptor) and 2-inch-square (2 adaptors). At the 0.9-m (f/7.5), these 4-inch filters illuminate approximately 6K X 6K pixels (56% of the total sky area); At the 4-m, 4-inch filters illuminate approximately 5.5K X 5.5K pixels (46% of the total sky area).
The positioning of the filter track is highly repeatable. However, the acceleration of the track can occasionally dislodge dust particles between filter moves, particularly if intervening movements have turned the filter upside down. In all cases, the filter track software moves the track in the direction that minimizes the distance moved to reach the requested filter position.
In addition to the 14-position filter track, there is a manual slide that can hold a single filter of the same size (5.75 inches square).This may be used, for example, to hold a bandpass filter when polarization filters are used in the track. Use of an additional 'hand-insert filter' changes the focus. At the 4-m, changing, inserting, or removing this filter can only be done at the maintenance (Southeast Annex) position.
3.6. Filters For
Mosaic
To fully utilize
the field of view of the 8Kx8K, filters must be 5.75 inches (146 mm) square,
and have 5.43 inches (138 mm) clear aperture. The optimum thickness that
preserves image quality over the entire field of view is 0.47 inches (12.0 mm).
All NOAO mosaic filters adhere to these specifications to maintain a parfocal
condition. Thus, neither the telescope nor the guide TVs should require a focus
change when switching between filters. There is one exception, the CuSO4
U filter (k1001), for which there is a focus offset.
A list of all available filters and their properties can be found at http://www.noao.edu/kpno/mosaic/filters/. Note that in order for post-processing command to display the correct on-the-fly flat the official filter name needs to be specified in the correct parameter set (i.e. wheel1). The official names can be found at http://www.noao.edu/kpno/mosaic/filters/filter_names.
Some of the currently
available filters and approximate count rates (e
-/sec)
for a 20th mag star are:
|
|
|
|
4-m |
0.9-m |
||||
|
Filter |
TV |
Peak T% |
Central Wave |
FWHM |
e-/s |
Central Wave |
FWHM |
e-/s |
|
U |
S8612 |
79.5 |
3577 |
646 |
35 |
3577 |
647 |
2 |
|
B |
BG-38 |
69.3 |
4360 |
990 |
330 |
4360 |
990 |
14 |
|
V |
BK-7 |
88.4 |
5370 |
940 |
340 |
5370 |
940 |
15 |
|
R |
RG-610 |
86.2 |
6440 |
1510 |
410 |
6440 |
1510 |
16 |
|
I |
BK-7 |
93.9 |
8220 |
1930 |
225 |
8220 |
1930 |
9 |
|
Hα |
BK-7 |
94.3 |
6569 |
80 |
|
~6575 |
~80 |
|
|
Hα+4 |
BK-7 |
91.2 |
6611 |
81 |
|
~6615 |
~81 |
|
|
Hα+8 |
BK-7 |
89.5 |
6650 |
81 |
|
~6656 |
~81 |
|
|
Hα+12 |
BK-7 |
86.1 |
6692 |
81 |
|
~6695 |
~81 |
|
|
Hα+16/[SII] |
BK-7 |
90.7 |
6730 |
80 |
|
~6736 |
~80 |
|
|
SDSS g' |
BK-7 |
90.2 |
4813 |
1537 |
|
4813 |
1537 |
|
|
SDSS r' |
BK-7 |
91.8 |
6287 |
1468 |
|
6287 |
1468 |
|
|
SDSS i' |
BK-7 |
94.6 |
7732 |
1548 |
|
7732 |
1548 |
|
|
SDSS z' |
BK-7 |
94.8 |
9400 |
2000 |
|
9400 |
2000 |
|
|
[OIII] #2 |
BK-7 |
75.2 |
5021 |
55 |
|
~5027 |
~53 |
|
|
[OIII]+29#2 |
BK-7 |
90.5 |
5305 |
241 |
|
~5305 |
241 |
|
|
White |
BK-7 |
97.2 |
5600 |
6800 |
|
5600 |
6800 |
|
|
Wash M |
BG-38 |
87.1 |
5100 |
1140 |
|
5100 |
1140 |
|
|
Wash C |
S8612 |
75.4 |
3860 |
1034 |
|
3860 |
1034 |
|
|
DDO 51 |
BK-7 |
85.1 |
5132 |
161 |
|
~5132 |
161 |
|
|
WR CIII |
BK-7 |
68.4 |
4653 |
52 |
|
~4660 |
~50 |
|
|
WR HeII |
BK-7 |
73.3 |
4690 |
51 |
|
~4695 |
~49 |
|
|
WR 475 |
BK-7 |
78.8 |
4750 |
51 |
|
~4755 |
~49 |
|
|
WR CIV |
BK-7 |
72.7 |
~5816 |
46 |
|
5823 |
42 |
|
See Figures 3.6.1 through 3.6.5 for plots of the current filter transmission curves. ASCII Tables that describe the transmissions are available on the Mosaic Web Pages.
The U-band filter is based on the same formulation as our 4" filter set (liquid CuSO4 + UG-1). Because containment of the liquid requires a thickness around the edge that exceeds the nominal Mosaic dead zone, some vignetting is present. At the 4-m, the vignetting introduces a 20% loss of light at the edge, but recovers to zero-loss at 200 pixels from the edge.
Figure 3.6.1: The broad band filter set, including the 'White' filter.
Figure 3.6.2: The SDSS g', r', I', and z' filters, along with Washington C and M (M is the smooth curve slightly redder than g').
Figure 3.6.3: The current set of Hα (plus redshifted) filters. Note that Hα+16 serves as a [SII] filter.
Figure 3.6.4: The blue Wolf-Rayet filters for [CIII], He II, and a continuum at 4750.
Figure 3.6.5: The [OIII] on-band and off-band filters, plus DDO 51.
Figure 3.6.6: The V-band filter installed in the filter track. The 2 TV guider filters are visible to the lower left and upper right of the science filter. An Arcon controller can be seen in the background to the right.
3.7. Operation of
the Guider TVs
Guiding with the Mosaic is accomplished using one of two TV cameras on the north and south sides of the science field. These are intensified fiber-optically coupled CCD cameras ("ICCDs"), and so, they can be damaged if exposed to bright light. The video signal from the selected TV camera is fed to the guider system. The field of view of each camera is about 2.2 arcmin on a side at the 4-m and about 5 arcmin on a side at the 0.9-m.
The field of view of the TVs is fixed with respect to the science field. At the 4-m, the fields are approximately 1440 arcsec north and south of the center of the science field. At the 0.9-m, the fields are approximately 2400 arcsec north and south of the center of the science field. TV focus can be moved remotely; offsets are -0.9 and -1.7 at the 4-m, and -0.1 and -1.9 at the 0.9-m, for the north and south TVs, respectively.
At a given location, suitable guide stars are almost always available without moving the telescope from the desired position. We find that we can guide at the 4-m on stars as faint as V=20 in full moon, and at the 0.9-m to V~17 near full moon.
The TVs and guider are controlled by the telescope operator at the 4-m, but by the observer at the 0.9-meter.The observer needs to first select the N or S TV on the distribution panel (see Figure 3.7.1). For the selected TV, on the ICCD Control Panel:
- Turn the high-voltage potentiometer completely counterclockwise (10 turn pot)
- Toggle the power switch on (on TV screen, pixel defects will appear)
- Neutral density switch should be down (off)
- Push the momentary button to enable high voltage
- Slowly turn the high voltage potentiometer clockwise, monitoring TVs until guide stars appear
- Acquire a guidestar on the computer moss and initiate guiding (consult the 0.9-m telescope user's guide (http://www.noao.edu/0.9m/Tel_Op.html) for details on step 6)
When switching between the two TVs, be sure to turn the high voltage potentiometer counterclockwise and turn off high voltage on the TV no longer in use.
Figure 3.7.1: A schematic drawing of the layout of the TV control panels at the 0.9-m. Only the two leftmost panels in the lower rack are used with the Mosaic TVs. The upper rack is used to select which TV/video signal is seen on the monitor.
3.8. Correctors
The 4-m corrector is a 4-element fused silica (for maximum U-band efficiency) design with additional internal prisms that serve as an atmospheric dispersion corrector (ADC). See Figure 3.8.1 below for the optical layout and refer to Jacoby et al. (1998, SPIE 3355, 721) for a detailed description of the corrector and ADC.
Figure 3.8.1: The 4-m corrector optical layout. All elements are made from fused silica except for the ADC components, which are made from LLF6, UBK7, LLF6, and UBK7 as viewed from left to right. At the right are the science filter, dewar window, and CCD (from left to right). The ADC elements appear in the middle as 4 planar elements (although they are wedged).
The 0.9-m corrector is a simple 2-element fused silica design. There is no ADC at the 0.9-m. See Figure 3.8.2 below for the optical layout.
Figure 3.8.2: The 0.9-m corrector optical layout. Both elements are made from fused silica. At the right are the science filter, dewar window, and CCD (from left to right).
Coatings and Scattered Light: All optics have been coated with very broad-band multi-layer anti-reflection coatings to improve photon collection efficiency and to reduce scattered light. Surface losses are ~10% from below 3500Å to longward of 9500Å at the 4-m and better at the 0.9-m. In addition, all interior structural surfaces have been blackened to minimize scattered light. Tests at the telescope indicate that the new correctors suffer significantly less than the old correctors from scattering. Nevertheless, with the very wide fields being imaged, bright stars are inevitable, producing some ghosting from bright objects in certain fields.
Image quality: The 4-m images are excellent across the entire 35'x35'
field. On good nights, we have documented uniform 0.65' images in R. There is
no measurable focus gradient or PSF variation to within 
~10%.
The 0.9-m telescope is not as well corrected. There is a small focus gradient across the 59'x59' field amounting to 20-30 focus units. Images in the corners of the mosaic degrade somewhat, especially in the lower-left corner (CCD #1). Also, the corners of the field are slightly vignetted (~5-10%) by the internal telescope baffle.
Image Scale: The 4-m scale is slightly variable (6.3%) due to pincushion distortions, from 0.261" per pixel at the center (f/3.1) to 0.245" per pixel (f/3.3) at the corner of the field.
The 0.9-m scale is 0.425" per pixel. The spatial variation is small, with the scale decreasing to ~0.420" per pixel at the corner of the field.
Ghost Pupil: When using narrow-band interference, very blue (e.g. U) or red (e.g. I, z') band filters at the 4-m, a faint image of the telescope pupil falls on the CCD and has a diameter of about 10 arcmin. Depending on the bandpass and construction of the filter, this reflection typically manifests itself at <1% for broad band to 2-4 %(narrow-band [OIII], Ha, U, I, z') level above the background. It arises from an internal reflection off the front surface of the rear element of the corrector despite the use of an extremely good anti-reflection (AR) coating. Our investigation suggests that similar 4-element correctors currently in use should exhibit a similar effect, and tests performed by Alistair Walker with the CTIO 4-m confirm this analysis.
Although the ghost pupil can subjectively appear severe when viewed at high contrast (for narrow-band filters), photometric accuracy is preserved when this additive term is removed during the reductions. One can avoid the affected region on the Mosaic array by moving the telescope slightly if there is any concern about the reduction process [see Sec. 4].
For a description of how to correct for pupil ghost images in your observations you can read the discussion in the NOAO Deep Wide-Field Survey MOSAIC Reduction Cookbook (http://www.noao.edu/noao/noaodeep/ReductionOpt/frames.html).
3.9. Atmospheric Dispersion Corrector (at the 4-m
only)
The Earth's atmosphere disperses the light from stars significantly when observing away from zenith. The effect is greatest and similar at U and B where the stellar image is stretched, for example, ~0.5" at a zenith distance of 45º (1.4 airmasses), and 0.9" at 60º (2 airmasses). The ADC prisms can be configured via a rotation to counter this effect nearly completely, thereby greatly reducing the elongation of the image introduced by the atmosphere.
Modes of Operations: There are 3 modes that the ADC system can be used[1]. Null mode, where the ADCs make no correction; Track mode, where corrections are periodically updated; and, Preset mode, where the corrections are preset at the beginning of an exposure to be correct for the middle of the exposure, but otherwise not moved. We recommend using Track mode to improve the image quality with a minimum of attention to the ADC operations.
ADC Filter Mode: In Modes 2 and 3 (Track and Preset), the positions of the prisms are dependent on the observing bandpass because the optimum corrections have a color-dependent functionality. Thus, in using the ADC, it is important to select the proper filter mode in the ADC GUI on rush.
While the ADC is not necessary to compensate for atmospheric dispersion when using narrow-band filters, the guide cameras have separate broad-band filters [see sec 2.7] that will see the effects of atmospheric dispersion. Differential refraction between the guider and the science CCDs will change as a function of telescope position, and will cause blurring for long exposures at high zenith distances if the ADC is set to Null Mode. In this case, the observer should select an ADC Filter Mode that encompasses both the narrow-band filter and the TV camera filter bandpasses and select ADC mode 2 or 3 (track or preset).
The ADC GUI: Control of the ADC prisms and modes is selected via the ADC GUI (see Figure 3.9.1). Status of the ADC prisms also is displayed in the GUI window. The primary user parameters in the GUI that concern the observer are the ADC Mode (Null, Track, Preset) and ADC Filter Mode (U, B, V, R, I).
Figure 3.9.1: The ADC GUI.
When you change the filter you are using with the ADC in use you will be prompted to confirm that you want to change the operating mode of the ADC. There is no beep, or other sound warning that the system is waiting for your confirmation. There is a pop-up window that asks you to pick the appropriate mode. If you would rather have the system automatically decide which is the correct mode, you can unclick the selection button in the GUI (see figure 3.9.1). The default mode for each filter is listed at http://www.noao.edu/kpno/mosaic/filters/filter_names. Note that the default of start-up mode of operation is for the system to prompt the observer to confirm the ADC mode before starting the first observation with a new filter.