As noted in Chapter 1.1, hardware design has changed significantly since the PDR. Slitmasks are no longer mounted in frames, simplifying the required slitmask handling/holding facilities, and the number of masks that can be loaded in the instrument has increased from 10 to 13 per barrel.
There have been no major changes in software or procedures since the PDR. Much of the slitmask fabrication process has been tested (see I.3.1). Detailed procedures for slitmask fabrication and handling have been developed (see II.1). We have identified the need for an additional software module (``MillControl'') to interface with the NC mill and the database during slitmask milling and check-in precedures.
Outstanding issues listed in the PDR have been settled. Slitlets will be specified by four corners, which in practice will define a parallelogram. (More complex shapes, such as arcs, can be constructed by multiple slitlets.) The ``mask design'' and method of ingestion into the database has been decided. Mask design now consists of target objects (``Objects''), slits on the sky (``MaskDesign''), a physical description of the mask (``Blueprint''), and a mapping of target objects to slitlets (``SlitObjMap''). These four tables are packaged in a single FITS file that is sent to CARA for database ingestion and mask fabrication.
Plans for both storage and display of DEIMOS images have changed since the PDR. Differences in the manner of storage are described below and detailed in I.3.6 and I.3.7. Differences in the display are described here and detailed in I.4.1 and II.6.
Chapter 7 of the DEIMOS software PDR contains detailed specifications of image storage formats. Some standard practices for handling mosaic images have been emerging from the astronomical community. These practices were not present at the time of the PDR, and some aspects of mosaic image storage have been modified to conform better.
In the original design we proposed that the sections of image from all CCDs in the mosaic could be placed into a single FITS image array. The principal advantage of this scheme is that the images can be displayed by existing astronomical image display clients. The disadvantage is that it interleaves sections of exposed pixels with calibration data.
Emerging consensus in the FITS community dictates that archival storage of raw image data is best done by placing the image sections from each CCD amplifier into a separate FITS HDU. This scheme is in use by the NOAO mosaic detector. (IRAF has already been modified to process FITS files containing multiple IMAGE extensions of such single-amplifier data.)
In the FITS community there is less consensus about the archival storage of ``pipeline-calibrated'' image data. If the DEIMOS CCD controllers are read using dual-amplifier mode then the calibrated data from a single CCD will be stored in a single FITS image array. This greatly facilitates astrometric and photometric analysis.
For the purposes of quick-look analysis the image frames may be written into a single data array which has approximately correct geometry. We do not intend to describe any form of archival storage for such images.
The PDR also specified that table extensions containing slitmask data would be appended to the image. The procedures and data requirements for slitmask fabrication and observation have been determined. The details of these tables can be found in Part I Section 3.6.
Chapter 8 of the DEIMOS software PDR contains detailed requirements for the quick-look image display. There has been no significant change in the requirements.
At the PDR De Clarke demonstrated a modified version of the ESO RTD program as a possible prototype for the DEIMOS guider and quick look image display. Subsequent testing of ESO RTD by Richard Stover revealed some problems that the ESO RTD development team was reluctant to fix.
The current DEIMOS display plan is a two-phase process. Initial laboratory testing of the DEIMOS mosaic will use a slightly modified version of the existing Keck Figdisp. As the NOAO Real Time Display becomes available for initial testing we intend to collaborate with them on the development of features needed for DEIMOS. Details are documented in section I.4.1.
Since the PDR we have reduced the scope of our database effort to one tightly focused on observing and engineering support. We no longer plan to tackle the problem of public accessibility to, or publishing of, DEIMOS science data; that problem is beyond the scope of the instrument project.
We have more faith that STB as installed at K1 and K2 will be adequate for archiving DEIMOS images, and that the database of images headers at CARA will be accessible to us for engineering and observer support. No separate DEIMOS archiving system is currently proposed. Although STB has not been tested with images as large as those DEIMOS will produce, we have no reason to suspect that it will fail due to image size.
We have designed schema for 90 percent of the functionality we expect from the DEIMOS database component (see Part II Chapter 7 for details), and defined the operational role of the database in such a way as to provide a layer of separation from vendor-specific database server interfaces. The detailed schema design is documented in Lick Observatory Technical Report (LOTR) number ??, copies of which will be available at the CDR.
Our plans for highly-developed Web interfaces to DEIMOS data have been scaled back, since we no longer propose to support widespread public access. We will direct our efforts instead to engineering analysis tools and comprehensive instrument performance logging (by means of existing KTL services), friendly online documentation, and semi-automated data reduction.
Our plans for document generation and the practical application of a central database of keyword information have already been realized. Large chunks of the CDR documentation are generated rather than manually drawn or written, and can be regenerated as design changes occur. We have early prototype code for source code generation as well. The database of DEIMOS keywords is about 80 percent complete and is documented in LOTR ???, aka ``The Dictionary".
The only areas which remain ill-defined at this point are the instrument's requirement for calibration data, and the exact functioning of the FCS. We have made headway on function definition here, but no concrete schema design has yet taken place because the hardware and procedures are still unsettled. Some preliminary schema design is documented, but should not be regarded as final or even firm.
Upon the recommendation of the PDR panel, we have dropped consideration of the Guider. However, we would like to list some desired features for the guider which would aid DEIMOS observations and efficiency:
There have been no changes to the slitmask alignment procedure. This procedure has had an additional year of field-testing with LRIS, with no reported problems.
The specifications for the DEIMOS mosaic CCDs are shown in Table 2.1 The current status of CCD fabrication and availability was extensively discussed at the Detector/Controller CDR on May 20, 1997 (copies of those documents are available). The consensus at the review was that DEIMOS would ultimately be able to obtain CCDs that met or exceeded all the specifications in Table 2.1, but not on the originally planned timetable. We are currently discussing the advisability of fabricating an interim detector, and the results of that discussion will be presented at the Software CDR.
Figure 2.1: Current Schedule
Production of the DEIMOS mosaic CCD detector is currently the pacing item on the critical path. Figure 2.1 shows a success-oriented schedule involving a single detector (i.e., no interim detector). Instrument delivery is already delayed four months beyond our original schedule despite the success-oriented approach. This schedule calls for delivery of the main mosaic CCDs by January 1, 1998. None of the development paths outlined below can be depended upon to deliver CCDs on such a tight schedule, hence our consideration of an interim detector.
The following is a brief summary of the status of three potential delivery paths for the DEIMOS mosaic CCDs.
CARA entered a consortium with five other observatories to co-fund the development at MIT Lincoln Laboratories of a 2K 4K 3-side buttable sensor meeting the specifications given in Table 2.1. Thinned and packaged CCDs from the first round are just now being tested. The sensors are of two types -- a normal 20- m thick epitaxial device and a thicker 40- m thick ``deep depletion" device built on high-resistivity silicon. The latter has the advantage of pushing the onset of fringing from 7000 Å out to 8000 Å. Red QE is also increased, being over 80% from 6000 Å\ to 9000 Å. The 20- m epitaxial devices from the same round also appear to be excellent performers for their type.
DEIMOS expects to receive about six usable devices from this first development round. However, before sinking all or part of our CCD budget into a second round, several technical problems would need to be solved:
(1) The MITLL production line is switching from 4-in to 6-in wafers, with the result that the new mask design brings the corners of the CCDs to within 4 mm of the wafer edge. This necessitates a whole new method of thinning and may introduce defects into the portions of the CCDs near the edge.
(2) A new blue-sensitive two-layer anti-reflection coating needs to be developed to increase QE in the blue. This might require coating by an independent contractor such as Mike Lesser, complicating the production process.
(3) MITLL uses a laser-annealing process with boron implant to passivate the back surface. The laser heats the back surface unevenly, leaving a ``brick-like" pattern of QE variations with peak-to-valley amplitude of 15% at 4000 Å. Since the amplitude of this pattern is likely to be steeply color-dependent, flat-fielding direct images with objects of widely varying color could prove impossible in B, and perhaps also in V. This is unacceptable and must be reduced in amplitude by at least a factor of three, which might require major alterations to the laser.
(4) High-resistivity 40- m thick devices demand much higher-quality silicon wafers and clean-room conditions. More testing of high-resistivity wafers is needed before a large-scale run of these devices could be embarked on.
Testing of these items is underway, but results will not be known until mid-September at the earliest. Should tests be successful, another 8 months at minimum would be needed to deliver DEIMOS CCDs, pushing the delivery of CCDs to mid-May and the final delivery date 5 months beyond that shown in the success-oriented schedule in Figure 2.1.
We recently tested a very good SITe 2K 4K CCD. It meets our requirements except that QE in the blue and visual is low (40% at 4000 Å, 60% at 6000 Å). SITe has accepted orders for several tens of astronomical CCDs but is having trouble meeting production schedules. It is not yet delivering devices routinely. We have so far been unable to get them to quote a firm delivery date on DEIMOS CCDs.
EEV has two 3-side buttable formats. One has 13.5- m pixels in a 2048 4500 format. This size would provide the same spectral coverage as our 15- m px DEIMOS specification but would cover 10% less slit length. EEV is taking orders for delivery of this device by the end of 1997 but none in the full size has as yet been tested by American astronomers. Smaller devices with the same amplifier have been tested by Geary at CfA and are said to meet the DEIMOS specifications in Table 2.1. Adoption of the 13.5- m format would make it awkward to mix these devices with those from other vendors, locking us into one supplier.
EEV are also developing a standard 2K 4K 15- m-px device for ESO. Its development is being pursued by several European observatories.
The lead time for either kind of device is said to be 12 months, and thus is comparable to the most optimistic delivery time for MIT Lincoln devices.
The salient conclusions from the Detector/Controller CDR are that:
(1) we should continue to develop software to operate an 8K 8K focal plane with specifications as originally planned,
(2) the delivery schedule of the final CCDs is highly uncertain and likely to be delayed at least 6 months beyond the success-oriented schedule in Figure 2.1, and
(3) it may be desirable to fabricate an interim detector to keep the project moving along. Such an interim detector could be constructed of thick CCDs from our in-house Orbit development program at Lick, perhaps including some engineering-grade devices from the first round of CCDs from MIT Lincoln.