GENERAL ISSUES

The question posed to us was: "What does the telescope & facility need to provide for instrument control & calibration?" This report gives background information about calibration needs and takes a first cut at assessing what calibration facilities should be provided as part of the telescope vs. what should be part of the instrument. Further iteration with the instrumentation groups is required in order to develop a real plan.

The general issues are:

• Need to provide flat-field calibrations. Often divided into: * pixel-to-pixel requirement for a flat field which is smooth over a few tens of pixels, and * global flat field which accurately models the incoming beam from the telescope over the scales from tens of arcsec to the full FOV.
• Need to provide wavelength calibration sources with line spacings (in wavelength) suitable for a wide range of spectroscopic resolving powers. Exact modelling of the telescope beam is not as critical here as for flat-fielding, but it should be a close enough so that wavelength errors are not introduced by the effects of field-position-dependent vignetting inside a spectrograph.
• Although we should try to build instruments with minimum flexure problems, and the Nasmyth configuration will help in that respect, it is pretty likely that some of the instruments will turn out to have flexure problems. It should always be possible to calibrate our instruments.
• This means that wavelength calibration should always be possible at any telescope position, in order to check for flexure effects in spectrographs.
• Flat-field calibrations are not so strictly needed at any telescope position, but this would be a very convenient feature for use with for multi-slit or multi-fiber instruments, where it is necessary to calibrate the throughput of each slit or fiber. If it takes significant time to put in a slit mask or fiber setup, it could be a real problem to try to calibrate a large number of setups with the telescope in a fixed position in the afternoon, or using the twilight sky.
• If queue observing makes it impossible to forecast which exact instrumental setups will be used during the coming night, it would be reasonable to assume that the instruments will be stable enough so that the setups actually used could be calibrated on the following day or at the start of the following night. [IS THIS A NON-SEQUITOR?]
• Obtaining sufficient surface brightness is usually a problem for flat-fielding in the UV, and for many types of wavelength calibration sources (cf. hollow-cathode arc lamps). This can lead to quite long calibration times, which could easily be a real problem in a telescope where many different instrument setups are used during the same night.
• Many telescopes, including the NOAO 4m's and Gemini, have substantial calibration facilities built into their cassegrain A&G modules. In the case of Gemini, all calibrations will come from one standard unit which can feed any possible instrument. This works because all of the instruments are fed by the same A&G unit. The SOAR instrument package will be distributed at two Nasmyth ports plus possible additional folded cass and the funny under-the-primary (collapsed?) cass port. Thus if we do try to design universal calibration units, we may still need several.

Surface Brightness Problems.

Calibration sources often depend on a diffusing element. The easy way to produce a highly uniform flat-field source is to have a diffuser that radiates uniformly into a large number (like 2*pi) of steradians; then the system is not sensitive to the relative placement of the instrument aperature and the illuminating lamps. An integrating sphere has the right properties. But in such a system we only want to intercept the light over a very small solid angle, so most of the light is lost and hence there is a surface brightness problem with many types of lamps.

It is easy to find flat field lamps which are bright enough to overcome these problems at all wavelength except in the uv. Quartz-iodide driving lamps work well. It is harder to find a range of arc lamps which produce bright lines at sufficiently close wavelength spacings to provide good wavelength calibrations at high resolution and at blue/uv wavelengths. A quartz-iodide continuum lamp feeding light through an inexpensive etalon is a possible souce of bright emission features evenly spaced at any desired wavelength interval.

SOME POSSIBLE APPROACHES

Small units mounted on the individual instruments.

Small units can reproduce on-axis beams with the correct f-ratio, but would run into problems with off-axis beams over large fields. The problem is that you need to place the exit pupil of the calibration system at a position which matches that of the telescope, in order to have the optics of the the instrument illuminated in the same way by the calibration source as by the telescope. Otherwise the rays at off-axis positions in the FOV will enter the instrument at the wrong angle, which can produce significant errors both in flat fielding and wavelength calibrations. In order to overcome this, the cassegrain spectrographs on the NOAO 4m telescopes view a calibration source through a field lens which is essentially in the focal plane of the telescope. This does properly correct the angle of the incoming beam, but the calibration light source still needs to be 1.2m back from the spectrograph slit in order to keep the optical system manageable. The Gemini unit described below is a very good approach, but would still require an optical path length of a few times the 130mm field diameter (7' FOV). Therefore the calibration unit probably will have to be part of the telescope rather than part of the instrument.

Gemini's System.

Gemini is using up one of the potential instrument ports on the A&G unit to house a general-purpose calibration unit that will provide flat-field and wavelength calibration for most of their instruments.

Summary of requirements for the Gemini Calibration Unit (GCU):

• Provide flat-field and wavelength calibration from UV to Near-IR (0.3-5 microns), for HROS, GMOS, GNIRS and GNIRI.
• Flatness of flat-field source:
• Final flat-fielding accuracy of instruments needs to be:
• 1% over 7' dia. FOV * 0.2% over 100"
• 0.1% over 20"
• Gemini will combine GCU output with sky flats [NIGHTLY, I GUESS?] to get this accuracy.
• ==> Calibration unit should be flat to 10% (peak-to-valley) over Gemini's 7' dia. FOV and 1% over 3' FOV.
• ==> Two-dimensional shape of GCU flat field should be stable to 0.3% over 12 hours.
• Exposure times: 5 s generally, but 60s needed in UV.
• Flexure: pupil must be centered to within 1% of its diameter and this centering should not change with telescope position by more than 1%.
• Wavelength stability of arc lines: centroids of lines not to shift by more than 0.1 pixel (although this is actually covered by requirement 2 above).
• Odds and ends: color balance filers to get flatish continuum spectrum, ND filters to avoid saturation.

Gemini believes that they will meet these requirements, using an illumination and pupil imaging system which is mounted about 2m back from the focal plane. The illumination system partially overcomes the surface brightness problem by using a novel combination of a reflecting sphere and reflective diffusing screen to obtain a calculated 40% throughput. The system is sketched in Figs 1-3.

Fig 1. Schematic of Gemini Calibration Unit mounted on A&G unit.

Fig 2. Optical layout of Gemini cal Unit. Light path to AO unit shown here is no longer planned.

Fig 3. Details of Gemini illumination system.

Some initial requirements that Gemini have chosen to drop, presumably because they were unable to meet them, are:

• GCU only works out to NIR rather than MIR.
• No longer have color balance filters to correct continuum shape to that of the sky.
• Calibration beam no longer has to go through AO system. Will do supplemental calibration on sky to determine AO flat-field response.
• Requirement that GCU replicate telescope vignetting was removed (presumably this is why sky flats are now required?).
• GCU flatness requirement over 7' field relaxed from 3% to 10%.
• Temporal stability of lamps: requirement relaxed from 0.1% to 5% with 1% goal.

Great White Spot on Dome

Flat-field sources consisting of diffusing screen mounted on the interior of the dome are used by many telescopes. These typically give 1-2% flat-field accuracy, and are regarded to be marginally accurate enough for general-purpose photometry in the RED? but not in the BLUE?, and to not be as good as the twilight sky. As compared to various types of projector flats, they have the advantage that they include all of the telescope optics. But they have the disadvantages of having to be used in a dark dome, and of low surface brightness (common to anything using a diffusing screen). Never-the-less, unless something better is provided on the telescope, we recommend the inclusion on SOAR of a great white spot, with careful attention to the illumination scheme, as cheap insurance for occasions when it is impossible to obtain sky flats.

Intermediate White Spot

An intermediate screen at perhaps the general position of the tertiary mirror, illuminated by lamps mounted on the telescope. This would be a scaled down version of a great white spot. It's advantage over a white spot mounted on the dome is that it could be used at any telescope position. It would share with a dome-mounted white spot the problem of having to be used in a dark dome, and any surface-brightness problems arising from solid-angle considerations.

A system of this sort is being built for use with the Hydra spectrograph on the Blanco Telescope. The screen will be inside the cylindrical baffle which sticks up from the center of the primary mirror, and the illuminating lamps will shine upwards from a ring near at the lower end of that baffle. It is hoped that it will also be a good solution for calibrating the other instruments used at the cassegrain focus.

Flat-field screen at the telescope pupil.

This is really the correct place to have a simple white spot in which no additional field lenses are used at the instrument. If the telescope's pupil is at the primary mirror, this could conceivably be the primary mirror cover. The pluses and minuses would be the same as for a smaller intermediate screen, except that the time to move this larger screen in and out of the beam is likely to also be a problem.

Lamps on mounted on spider.

These could have optics directing their beams into the focal plane, to overcome surface brightness problems. They would provide a low-order approximation to a properly filled pupil. They would have to be used in a dark dome. This was proposed, but not implemented, at CTIO for use with the fiber-fed Argus spectrograph used at prime focus. In that application the fibers would mix thebeam so much that a more completely filled pupil was not necessary. Instrument groups should comment on the usefullness of this approach in the context of SOAR's long-term needs.

RECOMMENDATIONS.

In the long run, the most straight forward approach seems to be to build something similar to the Gemini calibration unit as part of each AO/Guider Module. This would allow the maximum flexibility when introducing new instruments and at least minimize duplication of calibration modules (one per AO/Guider module rather than one per instrument).

We need to estimate the cost of building such a calibration module and the feasibility of putting it into the AO/Guider (along with the tip-tilt feeder and guider). An early step would be to iterate on the Gemini design to see if we can produce something suitable for SOAR, once we know details for SOAR such as the maximum instrumental FOV and the optical design of the telescope.