• Gravity invariant rotation to minimize flexure.

• A high throughput and low wavefront error, to optimise the instrument sensitivity. To achieve this, challenging performance requirements have been put on all optical surfaces. Given the wavefront error expected from the AO systems, and the difficulty of correcting non-common path aberrations across the whole field, the optical performance budget sets the goal of reaching only 120 nm rms wavefront error between the cryostat entrance window and detectors.

• An array of 3 × 3 H4RG detectors, which are read at a pixel rate of ~200 kHz to allow the most sensitive exposures. A low resolution imager offers a 4 mas pixel scale over a 50.5″ × 50.5″ field of view, to fully sample the diffraction limited PSF from 1.5 μm to 2.4 μm; the wide field effectively provides a major multiplex advantage. A high resolution imager provides a 1.5 mas pixel scale over a 19″ × 19″ field of view, to fully sample the diffraction limited PSF from 0.8 μm to 1.5 μm; the fine sampling at long wavelengths provides the capability needed for PSF de-blending in very crowded fields.

• Large filter wheels, able to hold >30 filters, for broad-band and narrow-band imaging as required by the science cases. The filter suite includes neutral density filters for bright targets, and spectroscopic filters that could in principle also be used for imaging. The large 14 cm filter diameter, and the requirement to have high throughput (exceeding 95% for the broad-band filters) as well as steep bandpass edges, will mean the manufacturing is not straightforward. Given that for typical observatories, more than ~90% of observations are performed with only ~10 filters, if a balance is needed between performance, cost, and number of filters, the preference will be for fewer high quality filters. The number of available slots, however, would remain the same.

• An atmospheric dispersion corrector (ADC), which is always in the beam path. The chromatic dispersion is large enough with respect to the diffraction limit, that, over nearly the whole sky, the gain in sensitivity by correcting it and keeping the PSF compact outweighs the loss of throughput (due to the ADC’s 8 optical surfaces) even for isolated point sources.

• Dithering is possible within 0.3″ of any pointing (with a small time overhead of ~2 s), to address detector systematics; and within ~10″ of the initial pointing (with an overhead up to ~15 s) to derive sky background. Larger dithers may be possible depending on the NGS configuration used by MCAO. Offsetting by several arcminutes to sky (without AO correction) is also possible with a somewhat longer ~30 s overhead.

• Secondary guiding is the use of reference sources in the science frames to measure slow drifts of position, rotation, and plate scale (on timescales of seconds to minutes). Optionally, these corrections can be applied during observations. Information about the reference sources is available to the pipeline, which can use them for fine matching of individual exposures before combining them.

• Differential tracking, needed for observing solar system targets that have a non-sidereal speed up to 100″ hr^–1, will be possible with both SCAO and MCAO.

• The preparation tool (PreCADO¹⁸) is essential for configuring observations. Among its many tasks are: the assessment of how the NGS asterism restricts the dither pattern, and whether some pointings might be allowed where only 2 rather than all 3 NGS are accessible; the identification and handling of ‘spare’ NGS in case any of the primary set cannot be used; notification about sources likely to saturate for the selected exposure time; and estimation of the observing efficiency.

• The data pipeline will provide standard processing of the exposures, combining frames from a single observing block to create a photometrically calibrated image, together with an estimate of the noise. The alignment of the individual exposures will be done to a few milliarcsecond precision. While photometric calibration will ideally be done using sources in the science frames, the distortion correction will rely as little as possible on those. Instead the aim is primarily to use corrections that are pre-calibrated or derived from models.

• Mechanical and thermal flexure are minimized by: (i) the gravity invariant instrument orientation; (ii) an optical path in which all mirrors are fixed (for the high resolution imager, the ADC and filters are the only movable optics; the only additional movable optics in the low resolution imager are two flat fold mirrors); and (iii) controlling the temperature of the cold opto-mechanics to ~0.1 K over a 1-hr observing block.

• Calibration masks will be used to measure instrument distortion. These will include the distortion both in MICADO and in the warm optical relay, as a function of rotation (for field versus pupil angle); and distortions from the ADC as a function of rotation angle (for zenith distance). Stabilisation of the temperature in the cryostat will ensure that calibrations are applicable over sufficiently long timescales.

• In contrast to standard imaging, observational constraints may need to be imposed. These could include: (i) fewer filters may be offered (for example, one broad and one narrow filter for each of the J, H, and K-bands), to avoid excessive calibration load; (ii) night-time calibrations may be attached to the science observations (e.g. a distortion measurement at the start and/or end of an observing block); (iii) specific dither patterns may be used to allow on-the-fly calibration. (iv) the airmass range may be restricted (not too large, since atmopsheric effects become severe; but not too small, since field rotation increases).

• A reference frame is needed to set the global plate scale and low order distortion terms. The Gaia (and perhaps Euclid) reference frame will be an option for this in some cases – but not all, because many of the fields observed by MICADO will be either crowded or obscured. It may be possible to omit a reference frame if the purpose of the observations is to measure only proper motions, rather than positions, of the stars.

• The pipeline will process and combine the frames from a single observing block, in such a way as to yield a photometrically and astrometrically calibrated data product. Correcting variations, or evolution of, low order distortions between individual frames using astrophysical sources in the data themselves is central to the astrometric data processing. This will be possible, because the pointings where the most precise astrometry is required are likely to be stellar fields.

• This is one of the main drivers for a SCAO system, which is able to provide a high Strehl ratio on axis using a bright NGS (while, naturally, one also aims to achieve good performance on fainter stars too). A pyramid WFS is a natural choice because it can reach higher Strehl ratios than a Shack-Hartmann WFS for the same star magnitude; and it also yields better performance in terms of contrast close around the star because of the lower speckle noise. ³⁵

• Phase masks will be available in both the focal plane and pupil plane. The former will include a small inner working angle phase mask³⁶ (akin to an annular groove phase mask, or vortex, but optimised for the ELT pupil) and a classical Lyot coronagraph, ³⁷ both to be used with a Lyot stop in the cold pupil. The latter will be a grating vector apodised phase plate.^{38, 39} Neutral density filters are also required, to avoid saturation during acquisition and calibration exposures. Sparse aperture masks⁴⁰ will also be available.

• Since the field of interest is close around the parent star, it is only necessary to read out the central detector. An option to increase the frame rate by reducing the number of rows read out will be available.

• Pupil imaging is needed in order to ensure there is a detailed record of how the pupil appears for each observing block. This is expected to be a standard calibration mode in MICADO, since the ELT pupil changes from night to night, due to ‘missing’ segments, and the different individual segment transmissions (resulting from the segment cleaning schedule).

• The control software will provide an option for pupil stabilisation (de-rotation), to enable angular differential imaging (ADI). Since the performance of the focal plane mask depends critically on its alignment with the star, the QACITS algorithm⁴¹ will be used to provide re-centering feedback at ~0.1 Hz rate during observations.

• The pipeline is only required to process individual frames. The reason is that the way in which multiple exposures are combined, or used, depends on the science goals, and the optimal procedure is expected to differ between programmes. In addition, processing techniques are developing very rapidly, so that a fixed pipeline procedure is an unattractive option.

• The implementation of the spectroscopic mode in MICADO is optically smart, enabled by a single fixed cross-dispersing grating module that can be moved into the beam path by the main selection mechanism.⁸ Order sorting filters are used to switch between wavelength ranges.

• Various slits are available in the focal plane mask. The narrowest, 16 mas wide, is optimised for point sources. Two slit lengths are available: an off-centre 15″ slit covers the 1.45–2.45 μm (HK-band) and 1.15–1.35 μm (J-band) wavelength ranges; a 3″ slit is used for the 0.83–1.45 μm (IzJ-band) range. An additional set of slits, at slightly shifted positions, provides some flexibility to ensure that scientifically important spectral ranges do not fall across the gaps between detectors.

• Operationally, because the ADC is located in the cold pupil after the focal plane mask, the recommended alignment of the slit will be along the parallactic angle; with the orientation updated for each exposure. The user, however, is not required to follow this scheme.

• The spectral resolution is around R ~ 20000 (15kms^–1) for point sources, and R ~ 10000 when integrated across the slit. The difference arises because the slit is wider than the diffraction limit of the telescope, a necessary outcome of tolerancing and stability issues. The key effect is that the spectral resolution depends on the spatial profile of the science target (and narrower than the background sky lines); and hence an image of the target in the slit is taken as part of the acquisition sequence, when centering the object.

• A metrology system is used to maintain optimal alignment of the science target in the slit. The dominant issue here is the SCAO dichroic, which reflects the visible light to the WFS and transmits the near-infrared light. This is a large optic in a large mechanism; and even though the NGS module is mounted stiffly to the cryostat, even very small drifts in its position could lead to a shift of the science target on the instrument focal plane that corresponds to a significant fraction of the 60 μm (16 mas) slit width. The aim of the metrology system is to minimize these.

• The pipeline will rectify, calibrate, and combine the 2D spectral segments; and, for bright sources, extract a 1D spectrum. A key design choice, to avoid too tight specifications on the mechanism repeatabilities, is that the pipeline should be able to handle small rotations and translations between the day-time calibrations and night-time science observations.