TECHNICAL DESCRIPTION OF NAOMI

Internal document number AOW/GEN/AJL/7.3/07/97/Technical Description of NAOMI

Version date: 26 November 1997

CONTENTS

1. Introduction
2. System Overview
3. System-Level Features
4. Key Adaptive Subsystems
5. Optical performance specifications
6. Opto-mechanical system design requirements and goals
7. Alignment and Calibration
8. Global Opto-Mechanical Components
9. Software Architecture
10. Real-Time Control system
11. Observing System control interfaces.
12. User interfaces
13. Observational capabilities and procedures
14. Telescope, GHRIL and supported instruments
15. Software and computing standards; maintenance issues
16. Handling and operational support requirements
17. Lifetime and downtime
18. Maintenance
19. Safety
20. Documentation
21. Future Development
22. References

1. Introduction

This document provides a technical summary of the NAOMI system designed to meet the requirements described in the document Top Level Scientific and Operational Requirements for NAOMI. As each major technical feature or specification is introduced the table format illustrated below is used to describe the following connections: ·

• connections back to the document Top Level Scientific and Operational Requirements for NAOMI, in particular to the driving scientific, instrumentation interface or operational clause which requires the particular technical feature or specification,
• connections to work package specification documents carrying more detailed technical descriptions of the particular feature or amplifications of the specification,
• connections to figures or external documents presenting performance modelling which illustrates the relationship between the driving requirement and the technical specification.

Upgrade routes are available for extending the baseline NAOMI specifications. The maintenance of this upgrade potential is to be regarded as part of the baseline NAOMI specification and subject to the same change control.

This document continues with a brief qualitative overview of the overall system concept and purpose and of the remaining sections of this document and then proceeds to a description of the system's key features and components and their respective functions.

2. System Overview

This section provides a brief qualitative overview of the NAOMI system and of the contents of the remaining sections of this document.

NAOMI is an Adaptive Optics (AO) system to be deployed at the GHRIL (Nasmyth) focus of the William Herschel Telescope (WHT). The acronym NAOMI stands for Natural guide star AO system for Multiple-Purpose Instrumentation. Its purpose is the amelioration of the effects of atmospheric turbulence on GHRIL image quality.

The principle subsystems of NAOMI are illustrated schematically in Figure 1. NAOMI will be a feed-through facility providing improved image quality to near-IR imagers and spectrographs deployed at GHRIL. It will also provide partially-corrected light to instruments at a separate optical science port.

NAOMI will use natural reference stars rather than a laser-beacon for measurement of the instantaneous turbulence-distorted wavefronts. Guide stars of a certain brightness are required in order to provide a given level of compensation and because this compensation is only effective over a limited angular patch, the sky coverage will be constrained. Using low-noise, high quantum-efficiency wavefront sensing and an accurate compensation element, it will be possible, however, to attain the performances outlined in the attached modelling predictions.

A significant part of the project plan for NAOMI is based on the Durham University ELECTRA AO development programme. This will yield early characterisation and testing of key adaptive subsystems: principally the segmented ELECTRA deformable mirror itself . The ELECTRA wavefront sensing and computer systems will also be closely-related to the final NAOMI systems. It will also yield significant components of the real-time, optimisation, visualization and GUI software.

NAOMI may be upgraded in the following ways:

• A full implementation of turbulence conjugation: a method of extending the sky coverage available with natural reference stars (and also of controlling the field-dependence of the adaptively-compensated point spread function).
• Continuing development of more sophisticated observing facilities (e.g. drag-and-drop positioning) into the observatory software and control systems.
• Full laser-beacon compatibility.

The ELECTRA mirror is a linearised, segmented adaptive mirror and will be used with a Shack-Hartmann wavefront sensor. The segmented mirror technology gives an important gain in fitting error at J and H wavebands when compared to a continuous facesheet deformable mirror. This is particularly important for the WHT which can be most competitive at these wavelengths. The linearisation will also give improved closed loop bandwidth for a given sampling frequency when compared to a deformable mirror with hysteresis.

The optical layout is illustrated in Fig.2. It enables a modular build of the opto-mechanical chassis, which in turn allows the operational specifications of removal from GHRIL and testing at a different location to be met. Infrared and optical science port space envelopes are located in outer areas of the layout so that constraints on future AO instrument design are minimized. The use of a double-pass fold mirror in the optical layout has changed from the layout at PDR. The change was imposed by mechanical constraints.

Figure 2: A schematic illustration of the optical layout for NAOMI.

IR throughput is driven by emissivity minimization requirements and will be > 65% to the instrument window, including the telescope. Emissivity due to NAOMI optics alone will be < 20%. Optical throughput depends on the observing mode and is driven by the wavefront sensor (WFS) sensitivity requirement which in turn is driven by the sky coverage requirements. Typically optical throughput to the WFS CCD is 31% over the 0.5-0.8µm wavelength region.

The RTCS is based on a Texas Instrument TMS320C40 ('C40') Digital Signal Processor (DSP) system which is paralellised and expandable. The baseline configuration will be the minimum one which allows the system latency specifications to be met. Visualization tools and hardware will be such that the eye can recognize and follow the detected wavefront shape and the system response using both recorded and averaged live data. The diagnostics and displays will allow the current system performance to be understood so that it can be optimized.

Observational procedures have been devised which allow calibration and observation of observations using infrared imaging and spectroscopic instruments. It will be possible to acquire bright and faint sources, the latter requiring very accurate absolute positions or offset positions from a bright star. Automation of some desirable procedures, particularly target acquisition by drag-and-drop, will not be possible in the baseline implementation.

It is now expected that automatic calibration procedures involving the availability of sub-windowed data from the science camera will be available via links from the up-graded INGRID software.

Interfaces will support key functionality such as jitter mode, tip-tilt off-loading (these terms are defined in the document) and will provide necessary data file header information. The User Interface will allow user-friendly operation of NAOMI itself with links integrating telescope, NAOMI and instrument. Initially it will be a minimal development of ELECTRA's GUI with later optimization based on ELECTRA and NAOMI observing experiences.

The procedural logic of the ELECTRA system (the "sequencer") will be re-used in the baseline implementation of NAOMI. Together with the minimally-reworked ELECTRA GUI processes, the real-time software and the corresponding support and optimisation function processes, this forms RELECTRA - the re-used portion of ELECTRA in NAOMI. Interfaces shall be provided to allow RELECTRA to initiate RGO/ING DRAMA activity and to receive status information. A further interface shall make ELECTRA sequencer status quantities visible to the RGO/ING DRAMA system. In particular it will possible for the RELECTRA procedures to set and examine RGO/ING interlocks. Note that the adoption of DRAMA is not yet secure, but it is being assumed unless an alternative ING standard is agreed. In the event of the latter all references to DRAMA in this document will be changed accordingly.

These will conform with ING requirements. Starlink standards are the guideline, with DRAMA and EPICS adopted for messaging and mechanism control respectively. RELECTRA shall use its own protocols for internal messaging.

Documentation, maintenance, half-arcsecond programme and JOSE are covered briefly.

2.9.1 Assumptions about the Telescope

System models currently assume that imperfections in the telescope alignment can be taken out by the DM. However it is advisable to use as little as possible of the DM stroke in so doing. Therefore the accuracy and stability of telescope alignment and focus are relevant to performance and NAOMI setup. Also the power in telescope aberrations on pupil scales of < 57cm is not correctable by the NAOMI system. Experience with MARTINI and ELECTRA indicates that that these are not likely to be major issues. Provision of the relevant input information would be required from ING if detailed modelling is to be done.

2.9.2 Environmental Factors

Environmental factors including thermal and temperature control, electromagnetic `noise' control and vibrational specifications are being developed between the NAOMI project and the ING.

An astronomical adaptive optics system has certain elements which are indispensable, others which are highly desirable in allowing good observing efficiency and elements which give additional scientific gains but are not essential to allow the system to do worthwhile science. The baseline NAOMI design has the indispensable elements (optics which deliver diffraction limited images, adaptive components and control system which give diffraction limited cores in at least average or better seeing conditions) and some but not all of the highly desirable features which enhance user friendliness and observing efficiency (see Telescope, Instrument and User Interface summary above). NAOMI has no `frills' implemented which are not essential to doing the basic science. However the design is such that many of these could be added efficiently should additional funds become available.

3. System-Level Features

A simple block diagram illustrating the NAOMI system as planned is shown in Fig.1. Before proceeding to the descriptions of the subsystem specifications, which are given in the next section, a number of NAOMI's system-level features are introduced and their requirements traced. In particular, any special optical, control and interface features are described together with the specialization requirements for instruments exploiting NAOMI image quality.

3.2.1 Introduction

The requirement to feed AO-corrected images to unspecified instruments with unknown space envelopes demands a layout in which the science port is in a relatively open area of the bench. The image scales (i.e. camera properties either for a spectrometer or direct imaging device) required to fully sample an image at 1.65µm wavelength (~0".04/pxl) mean that any instrument intended for use with NAOMI will need purpose-designed input optics. A significant change in its input optics configuration will be required for it to be used at an alternative focus of the WHT with an image scale suitable for uncorrected seeing. The baseline system for use with NAOMI has changed in the past 12 months from WHIRCAM (256sq InSb array camera) to INGRID (1024sq HgCdTe array camera being built on a fast track at RGO). Development of software control and interfaces for INGRID is the responsibility of ING. The NAOMI project is working with ING on ensuring the control interface meets NAOMI requirements. A data interface allowing subwindowed instrument data to be requested by and transmitted to RELECTRA will be maintained in order to support automated calibration procedures.

3.2.2 Specification

The input beam to the science port will have an f ratio of f/16.5. For optimum performance the instrument will require a cold stop at an image of the telescope pupil.

Figure 3: The 2.2µm point source sensitivity in 30 minutes for a 3uJy source viewed by NAOMI as designed plotted against the temperature of the system surfaces. An effective seeing of 0".2 has been assumed.

3.2.3 Upgrade Specification

Greater observing efficiency (Clause 6) will be possible if full image analysis facilities are available to the AO system to allow automated calibration and image size minimization to be carried out in a loop mode. This facility is now likely to be provided as part of the baseline through the INGRID control software.

If a conjugation facility is implemented via a full concave lens (as opposed to a lens with a hole passing the science field unchanged) at the first Nasmyth focus, then the instrument will need a cold stop with adjustable position along the optical path and probably adjustable size.

3.3.1 Guide Star selection

The star which is used as a reference object for the WFS can be selected from anywhere with in the 2.9 arcmin unvignetted FOV of the Nasmyth focus (Clause 2). This is done using a small mirror mounted on a thin glass disc at the corrected f/17.05 focus. This mirror, and the rest of the WFS, are moved to the position of a star selected by the user. By making the system (nearly) telecentric at this focus the alignment of the DM and the lenslet array in the WFS is maintained over nearly all off-axis guide star angles. The guide star can be the same as the infrared science object.

3.3.2 Conjugation

3.3.2.1 Introduction

The turbulence in the atmosphere can be concentrated in thin layers above the telescope. In this case there is an advantage, in terms of the isoplanatic angle, in conjugating the correcting surface to the turbulence. The gains that can be achieved are demonstrated in Wilson and Jenkins (1996) and the means by which this can be done is described by Wells (Conjugating AO correction to Turbulence in the WHT AO system design, Proc. OSA topical meeting on Adaptive Optics, ESO Conference and Workshop Proceedings No. 54, 1995.) The increase in isoplanatic angle gives science gains in two related areas: sky cover and PSF variation across the image. There is not a conjugation facility in the baseline NAOMI implementation but the choice of 7.3 sub-apertures/pupil for the system order means that the DM is oversized. It can accommodate conjugation up to 3km for stars up to 51" off-axis should such a facility be introduced. See Fig.4 for an illustration of the footprint on the DM and for a graph indicating the science gains.

3.3.2.2 Specification (upgrade)

3.3.3 The need for ADC's

Over the wavelength range of the WFS the light from the WFS is dispersed by about 2 arcsec at a zenith angle (x) of 45°. This is larger than the images at the WFS detector and the dispersion therefore has to be corrected using an ADC. The NAOMI design uses an ADC within the WFS optics which is optimized over the 500nm to 800nm wavelength range.

For the IR science light the dispersion is greatly reduced. For the 3 IR filters J, H and K the atmospheric dispersions, (at x=45°), are 0.1, 0.05 and 0.015 arcsec respectively. The diffraction limits at these wavelengths and for the WHT are 0.07, 0.1 and 0.13 arcsec. It is clear that at K the dispersion will not degrade the PSF and an ADC is not required. For J and H an ADC in the IR science arm is not in the baseline system and dispersion will start to affect the broad-band PSF's at x >45°

It is very important to note that there will be a change in the apparent separation of the visible object used as a reference for the WFS and the position of the IR object being studied as the Zenith angle changes. This change has to be compensated for during long observations using look-up tables. For exact compensation an effective wavelength, l_eff, (within the WFS) needs to be assigned to the guide star. l_eff will be a function of the stellar spectral type, the overall throughput vs l of the WFS and the accuracy of compensation achieved by the WFS ADC.

3.3.3.1 Specification

3.3.4 Pupil rotation

3.3.4.1 Introduction

The derotator will be used to maintain the required orientation of the science field. Therefore the pupil will rotate. Because the pupil is divided into sub-apertures the vanes may affect the slopes measured by the WFS. This will depend on the area the vanes project onto the pupil and, at any given instant, the angle they project.

3.3.4.2 Specification

3.4.1 Modal control

3.4.1.1 Introduction

Modal control refers to a technique where the instantaneous wavefront is fitted using a set of orthogonal functions. A useful set of functions is one which gives a near-optimal fit to the wavefront over the telescope pupil for a given number of functions fitted. When the signal-to-noise ratio of the wavefront measurements is low, because only a faint reference star is available, then the modes which are fitted can be restricted to those which have the slowest spatial (and therefore temporal) variations. These modes also have the best auto-correlation as the guide star position moves away from the science target and therefore provide an excellent means of tuning the level of guide-axis correction to give an optimal science-axis result (see Wilson and Jenkins, 1996).

Modal control is in distinction to zonal control were the wavefront is fit according to a locally determined optimum. Zonal control is the first level of control which will be implemented during development and will remain available for use in high light level conditions.

3.4.1.2 Specification

NAOMI is required to provide zonal and modal control with adjustable modal gains. These gains may be set on-the-fly without opening the control loops. Adjustment of gains is dealt with in the section on modal optimisation below.

3.4.2 Modal Optimisation

3.4.2.1 Introduction

See the section on Modal Control above. Modal optimisation is the process of automatically adjusting modal gains to cope with observing conditions. This technique has been successfully demonstrated by the Adonis and PUEO AO systems both of which use to good effect an optimisation algorithm derived from that described by Gendron and Lena (1995). Modal optimisation is still recognized as having a strong developmental aspect however; particularly in respect of off-axis or field-averaged optimisation or in the use of optimisation selection functions other than minimal wavefront variance over the pupil.

The priorities for NAOMI are the provision of the baseline Gendron-type optimisation and the hooks to allow further extensions. The WHT is well-placed in this respect as the JOSE measurements will provide tests for optimisation techniques and on the frequency of mode gain updates required. It should be note that Adonis and PUEO differ in this respect: the former has an off-line open-loop method and the latter can calculate updates concurrently with closed-loop operation.

3.4.2.2 Specification

NAOMI will provide the ability to update modal gains on-the-fly whilst the adaptive control loops are closed.

Concurrent calculation of loop gains will take place on a NAOMI workstation. The initial algorithm to be implemented will be as Gendron and Lena (1995).

3.4.2.3 Upgrade

Future optimisation algorithms requiring higher processing capacity may be implemented on additional processors in the real-time control computer rather than on the host workstation.

3.4.3 Non-sidereal tracking

3.4.3.1 Introduction

Many solar-system objects will be bright enough for self-referencing or to act as references for other sources. The system will be designed to use a non-sidereal object for wavefront sensing and observe a non-sidereal science object at a maximum differential rate of 0.007 arcsec/sec. It will also be able to observe sidereal science objects and perform wavefront sensing with a non-sidereal object. For the latter case the maximum rate will be 0.04 arcsec/sec. In both modes a tracking accuracy of ± 0.02 arcsecond or better will be a design goal. Note that, given current funding constraints, the non-sidereal tracking capability will be incorporated at low priority.

In contrast, the jitter slew rate used for activities such as image mosaicing is specified to be 4"/sec. Jittering will not maintain AO quality position registration during the 'slew' but will repeat to better than 0".01.

3.4.3.2 Specification

3.5.1 Telescope interface

3.5.1.1 Introduction

An interface to the TCS will be implemented, allowing the integrated tilt error to be off-loaded to the telescope autoguiding system. This will adjust the telescope position via a slow feedback loop to keep the integrated tilt error to zero. Focus will also be offloadable to the telescope in order to provide the capability of keeping the WHT wavefront focus peak-to-valley error <0.2 µm. The telescope position and derotator settings will also be available to be attached to the saved data files output by the instrument in use. The interface will be handled by the DRAMA messaging system (but note caveat re DRAMA).

3.5.1.2 Specification

3.5.1.3 Upgrade Specification

The software architecture has been designed such that the long term URD requirements can be implemented within the existing architecture. Thus software upgrades could further increase capabilities for the AO system procedures to control and co-ordinate all DRAMA-available functionalitity of the telescope and of other instruments AND for AO system functionality to be available to external process such as instrument control and TAG tasks.

3.5.2 Observing interface

3.5.2.1 Introduction

The detailed long term goals for a GUI are described in the Software User Requirements Document. For the baseline system, a GUI will be available which enables user friendly control of the AO system with only minimal interface to instruments or telescope. The top-level control will be derived from the ELECTRA GUI with minor add-ons found to be high priority from Electra-1 (first light cophasing test of ELECTRA) experience. Control of the NAOMI optical bench will be via a revision of the (editable) ELECTRA optical bench GUI. The project philosophy is that GUI tools are a rapidly changing area of software technology and therefore a full system GUI will be one of the last items to be written and cannot be specified at the present time. It is also envisaged that there will be developments of GUI standards at ING for running the WHT and its instruments. The specifications of these are not yet available, which again drives the decision to provide a minimal GUI based on ELECTRA for the baseline system.

An integrated telescope control - instrument control - NAOMI control GUI would be part of a software upgrade.

3.5.2.2 Specification

3.5.2.3 Upgrade Specification

The long term goal is to have a fully integrated NAOMI, Telescope and Instrument interface mitigated by DRAMA.

4. Key Adaptive Subsystems

The overview of implementation commences with the principal adaptive subsytems:

• Wavefront Sensor
• Tip-tilt Mirror
• Deformable Mirror
• Processing Subsytem

The specifications for each subsytem are introduced and attributed to the driving system-level specifications. References are given to the detailed specification and subsystem workscope documentation. Closely-related ancillary (non-adaptive) features are described briefly for completeness.

4.2.1 Introduction

The wavefront sensor (WFS) will use the Shack-Hartmann configuration. In this configuration the input pupil is divided into several subapertures by an array of small lenses. Each lens focuses the light from a guide star on a section of a detector array, e.g. 4 x 4 pixels, and the focused spot's centroid shift is determined by a processor. The average phase gradient for each subaperture is determined by dividing the centroid shift of the focused spot by the effective focal length of the lens. Because the Shack-Hartmann sensor does not provide a direct phase measurement of the turbulence-degraded wavefront, the phase gradients are transmitted to a reconstructor which calculates the phase of the wavefront. Note that both the wavefront processor and reconstructor are part of the RTCS. The conjugate (or reverse) wavefront (with its overall tip/tilt removed) is then applied to the deformable mirror. The tip/tilt information is sent to the tip/tilt mirror discussed in Section 4.3.

The WFS will usually operate with 7.3 subapertures across the WHT pupil. This number defines the system order and it is driven by the science clause 1. System modelling shows that Clause 1 can be satisfied using from 7 to 8 subapertures across the pupil diameter (see Figures 10 and 11). The segments of the ELECTRA deformable mirror will be optically mapped onto the WFS subapertures. Occasionally the system will operate with a smaller number of subapertures (see section 4.2.2.6).

The main components of the WFS are a pickoff mirror with a field stop, a collimating lens, an atmospheric dispersion corrector (see Section 3.3), interchangeable lenslet arrays (with = 10 x 10 lenslets for the largest array), a relay optic and a CCD. The relay optic images the array of focused spots at the appropriate pixel scale on the CCD. All of these components will be mounted on a remotely controllable stage to allow operation with a guide star anywhere within the 2.9 arcminute field. This stage will also allow jittering to be supported in accordance with Clause 4. The WFS will also have its own calibration source.

4.2.2 Specifications and Description

4.2.2.1 Wavelength range

The WFS will operate over a spectral range from 0.4 µm to 1.0 µm.

4.2.2.2 Phase gradient accuracy

The phase-gradient measurement accuracy along any axis will be equal to or better than 0.018 waves rms (where wavelength = 2.2 µm ) over each subaperture when operating with >= 1500 photons per subaperture per measurement incident upon the CCD when using no more than 4 x 4 pixels/subaperture. This performance includes the effects of sensor noise, photon noise and other sources of error. It is derived from system modelling for the Clause 1 conditions, i.e. system performance with bright stars, and it has been included in the Clause 1 error budget.

When operating with faint stars under the conditions for Clause 2, the phase-gradient accuracy is expected to be equal to or better than 0.14 waves rms (where wavelength = 2.2 µm ) over each subaperture when operating in the quad-cell mode with >= 40 photons per subaperture per measurement incident upon the CCD. This lower level of performance takes into account the high photon noise present at such low light levels. This performance has been included in the Clause 2 error budget.

4.2.2.3 Read noise/ read latency

Two CCDs will be used in a binned operating mode in order to get as close as possible to the read noise and latency requirements. This approach was first used in a WFS by Lincoln Laboratory and it is described by Barclay (The SWAT Wavefront Sensor, Lincoln Laboratory Journal, 5,122 (1992)). The WFS CCDs will have two readout rates that will be electronically switchable without recabling. There is no need to switch between frames. Under some conditions, e.g. good seeing with low wind speeds, one will be able to use a longer read latency and thus operate the CCDs with lower readout noise. At 100 kilopixels/second/port the CCDs will have <= 4.5 noise electrons/pixel (the original specification was for <= 3noise electrons/pixel) and at 500 kpixel/second/port the readout noise will be <= 7 noise electrons/pixel. CCD readout noise and latency have been included in the modelling and the error budgets for Clause 1 and Clause 2.

4.2.2.4 Pickoff

A small (1.5mm x 1.5mm diameter) pickoff mirror will direct light from the guide star into the WFS. The pickoff will be an integral part of the WFS and it may be moved anywhere within the 2.9 arcminute field to acquire a guide star. Light passing the pickoff will be directed to the optical science port. In addition to providing an acquisition function the pickoff, together with the main WFS assembly, will support a dithering mode. For a dither range of 5 arcseconds the repeatability will be <= 0.01 arcsecond. When the dither range is increased to 18 arcseconds, the repeatability will be <= 0.025 arcsecond or better. Non-sidereal tracking will also be possible (see Section 3.4.3.1 above).

4.2.2.5 Field stop

A field stop will be provided to allow the use of guide stars in crowded fields. The field stop size which is set by the pickoff mirror dimensions will be approximately 5 x 5 arcseconds.

4.2.2.6 Spatial descoping

Under some conditions, e.g. low light levels with modal optimisation, a performance advantage may be gained by using fewer subapertures across the pupil diameter; this is known as spatially descoping. The WFS will have the capability to change lenslets remotely to operate with only 3.68 subapertures across the pupil diameter. This facility is driven by Clause 2.

4.2.2.7 Auto change of lenslet focal length

A shorter focal length lenslet array will be provided to handle moderate to strong turbulence conditions (r₀ < 13 cm). As the atmospheric turbulence increases, the spot excursions also increase. Changing to a shorter focal length keeps the spots on the CCDs. This facility is driven by Clause 2 and Clause 3 together.

WFS Optics

4.3.1 Introduction

As previously stated in Section 4.2, the WFS (or a tip/tilt sensor in a future upgrade) will provide information on overall (or common mode) tilt. This is the tilt present over the entire WHT pupil and it will be corrected by the tip/tilt mirror. There are four sources of tip/tilt error as listed below.

1. Atmospheric turbulence
2. Pointing jitter of the WHT
3. Component vibrations
4. WHT long-term pointing drift

The tip/tilt mirror (a.k.a. the fast steering mirror or FSM) will use the WFS tip/tilt data to primarily correct for the first three sources. The tip/tilt mirror significantly reduces the stroke requirements that would otherwise be placed on the deformable mirror. Large low-frequency tip/tilt errors will be passed on to the TCS to avoid an excessive range requirement for the tip/tilt mirror. The mirror will also serve as a collimating optic, i.e. an off-axis paraboloid, in the common-path optics. This dual function reduces the number of optical surfaces resulting in higher transmission and lower emissivity.

4.3.2 Specifications

The mirror surface will cover an angular range of =<1 mrad over two orthogonal axes. This range is equivalent to about 5.6 arcseconds in WHT object space. The frequency response of the mirror shall extend to at least 250 Hz with an amplitude of at least 50 microrad at this frequency. (Current JOSE data from Richard Wilson indicate that 250Hz will be safe - this will be confirmed for a wider range of conditions within clause 3) The error budgets for Clause 1 and 2 allow for residual rms closed-loop tilt jitters of 0.016 arcsecond and 0.022 arcsecond respectively in WHT object space.

The mirror's clear aperture of about 115 mm diameter in the current design is large enough to allow conjugation over a 102" field for a dominant turbulent layer at 3 km above the telescope. This matches the effective over-sizing of the DM introduced by having 7.3 sub-apertures across the pupil. However turbulent-layer conjugation is an upgrade and it will not be employed in the baseline system. 100mm sized mirrors are supported by two potential commercial suppliers of the tip-tilt mirror; consultations are taking place with suppliers concerning the effect on specification of using a slightly larger mirror.

4.4.1 Introduction

The deformable-mirror subsystem will correct for the higher-order, i.e. tilt-removed, wavefront errors. Modelling shows that either continuous facesheet or segmented deformable mirrors can satisfy the requirements of Clauses 1, 2 and 3. The baseline system will operate with the 76-segment Electra deformable mirror and its drivers. This mirror has 10 segments across its maximum dimensions and approximately rounded corners. The WHT pupil, as imaged at the mirror, will cover 7.3 segments. The additional segments provide for a future upgrade to a turbulent-layer conjugation capability. Each segment has dimensions of 7.6 x 7.6 mm². The segments' optical surfaces are coated with aluminium. Three actuators are used to move each segment thus providing tip-tilt and piston control. The mirror will be provided with electronics to drive the actuators; these electronics are usually referred to as the drivers.

4.4.2 Specifications

The stroke of each actuator is 6 µm. Modelling indicates that this stroke is adequate to handle the strongest turbulence specified by Clause 3 (r₀ = 8 cm). Strain gauges and electronics will be provided to measure and correct the hysteresis present in the actuators. The hysteresis will be reduced to =< 0.7% (0.2% currently expected) as required by the results of system modelling with a propagation code. Hysteresis affects both the mirror fitting error (defined below) and the servo system response. The mirror settling time, i.e. the time required for the mirror surface to form a specified shape, will be < 400 µs. An allotment for mirror settling time as part of the system latency has been included in the error budgets for Clause 1 and Clause 2.

The deformable-mirror fitting error coefficient is a simple way of defining how well the mirror surface matches the conjugate of the turbulence-degraded wavefront. The following equation defines the coefficient.

s² = k ( d / r_o )5/3

where s² = variance of the residual wavefront error (radian²)

d = actuator spacing (57 cm projected at WHT pupil)

r_o = atmospheric turbulence coherence length (>= 8 cm)

Clause 1 is the driving science requirement for the deformable-mirror fitting error. It is the dominant source of error when operating with bright stars. The error budgets for Clauses 1 and 2 were based on a fitting error coefficient of 0.4 which is characteristic of a well-designed continuous-facesheet mirror. A segmented deformable mirror has a smaller fitting error coefficient (about 0.18) and thus it provides better performance (see Figure 12). This performance improvement was taken into account in recent propagation code studies.

The NAOMI computer and software systems are illustrated schematically in figure 1. The overall software architecture is described in section 9.

4.5.1 Real-time computer

The real-time computer will perform the three stages of processing required to transform WFS pixel data into mirror figuring commands:

1. The NAOMI real-time computer will accept 'horizontal' and 'vertical' one-dimensional pixel data respectively from each of two 4-port WFS cameras at at least the maximum peak readout rate and digitization accuracy indicated in the WFS subsystem error budget and specification. These are 500 kpixel/sec/port and 12 bits/pixel respectively. A peak rather than average readout rate is specified because the WFS will operate in a frame-transfer mode. The first processing stage of the real-time computer will reduce these pixels to subaperture wavefront slope values and will provide for switching its slope estimation algorithm between multi-pixel and quad-cell mode on-the-fly (between one frame and another). (Note that the on-the-fly feature is a goal that may not be implemented initially.) The supported WFS readout formats and required phase gradient accuracy are described in the WFS subsystem specification. These are: 6x6 pixels [acquisition-only], 4x4 pixels [WF slope measurement accuracy: 0.018 waves rms at 2.2 microns for clause 1 conditions] and 2x2 pixels [WF slope measurement accuracy: 0.14 waves rms at 2.2 microns for clause 2 conditions]. The algorithm switching will be synchronized with corresponding changes in the WFS camera readout format.
2. The slopes will then be used to derive either a complete set of deformable mirror drive commands directly via a single matrix multiply (zonal control) or a set of instantaneous modal coefficients (modal control).
3. The modal coefficients (if modal control is being used) will then be transformed into mirror drive data. The mirror drive data will then be output via DACs to the mirror power amplifiers.

The elapsed time for all these operations will be =<250 microseconds between the end of the WFS frame readout and the end of the deformable mirror write operation (but not the DM mechanical settling period). Any relaxation of the overall latency specification (WFS readout + computation + DM settling) will in the first instance be allocated to CCD readout (with a potential read noise gain - see the WFS subsystem specification) rather than to computation.

The real-time computer will provide command and status interfaces which will permit the wavefront correction method (zonal or modal), WFS pixellation and control matrices to be altered and examined. These operations should not cause the control loops to open (in conditions which were otherwise stable).

The real-time computer will make all of its operational data available to an external visualization system where it may be recorded and/or analyzed and subsequently presented. The retrieval of these data for external analysis will in no way compromise the performance of the real-time control computer.

The implementation for the above is described in section 10.

4.5.2 Mechanism control

ING standard methods will be used for the control of all standard mechanized functions (Clause 19).

The baseline is EPICS and VxWorks on a VME-based MVME-167.

Engineering and operational control and status methods will conform to ING standards.

4.5.3 User Interface

A procedural (scripting) interface and simple graphical user interface will be provided for the control of mechanism, real-time, optimisation and visualization functions. Basic visualization functions will be provided.

5. Optical performance specifications

The optical design of NAOMI aims to fulfil, as nearly as possible, the science clauses in the document Top Level Scientific and Operational Requirements for NAOMI. The givens for the design are

• The WHT optics - diameter, f-ratio, and the position of exit pupil
• The position of the Nasmyth focus 16mm from the edge of the GHRIL optical table
• The GHRIL table - size (1.35 x 2.5 m) and space constraints especially the derotator
• The unvignetted FOV - 2.9 arcmin
• The pixel scale at the science camera - 0.04 arcsec/pixel so as to nearly fully sample the diffraction core of the J PSF.
• The need to minimize the number of warm surfaces in the IR science path to maximize the IR sensitivity. The reductions in background gained by using smaller pixels are easily lost if the emissivity of the AO system itself is too high. See section 5.3.
• The need to maximize throughput to the WFS. The sky coverage of a natural guide star system such as NAOMI is limited by the availability of stars. It is important therefore to make sure that as much visible light as possible reaches the WFS detector.

5.2.1 IR Science

The operating wavelength range of instruments fed by NAOMI is limited at long wavelengths by the emissivity of the telescope and instrument. Measurements longwards of the K band will have a lower sensitivity because the increased background more than offsets the gain from using smaller pixels. The short wavelength limit of the IR science path is set by the reflectivity of the dichroic which in turn is affected by the need to maximize the visible throughput to the WFS. The reflectivity-transmission curve for the dichroic design being used in modelling is shown in Fig.13. Although not shown in this figure, the mean J band reflectivity is >80% and K-band reflectivity >94%.

5.2.2 WFS

The wavelength range of the WFS is fixed by the QE of the CCD's, the transmission curve of the dichroic, atmospheric dispersion and the fact that stars at the faint limit of those that can be used as reference objects will be predominantly red (G and later). Taking these factors into account the wavelength range for the WFS is fixed at 0.5 to 1 µm. Only the central ~1 arcmin of the dichroic substrate will be coated. This matches the field of a 1024sq array sampling at 0".04/pxl. The throughput to the WFS for guide stars more than 1 arcmin off-axis is thereby significantly increased. For historical reasons, observation through the coated and non-coated sections are referred to as Mode 1 and Mode 2 observations respectively.

5.2.3 Optical science port

For wavelengths shorter than 1 µm (Science clause 3) a new dichroic with high reflectivity at 0.8 µm could be used. This has the advantage of a FOV with no vignetting due to the WFS pickoff but, because of the lower throughput to the WFS, the limiting magnitude on the stars that can be used as a reference will be brighter. Alternatively the dichroic can be removed in which case all the visible light, apart from an area of ~5 x 5 arcsec used for the WFS pickoff, is available at the optical science port.

5.3.1 Introduction

All broad band IR (J filter and longer wavelengths) will be background limited after several tens of seconds on-chip integration, either by OH airglow (J,H) or thermal emission (K,L, M). Narrow band (=< 4%) filters at wavelengths < 2.4µm will be detector noise limited for most observations because for on-chip integrations of longer than about 5 minutes array and/or sky stability has usually become an issue. This could change as arrays develop. S/N ratio once on-chip integration times have reached background limited conditions varies as the square root of the total integration time. Keeping the emissivity low is important for two main reasons: (a) it improves S/N ratio and (b) the array wells have limited depth so lower background allows either longer on-chip integration times or gives higher dynamic range. At the longer thermal wavelengths the array can saturate even with the shortest possible integration time unless the emissivity is reasonably low. A high IR throughput is a by-product of a low emissivity goal and cannot sensibly be specified independently.

5.3.2 Specification

5.4.1 Introduction

The optical throughput driver is therefore to get as many photons to the WFS as possible (Clause 2) and also to meet the optical science port throughput requirement (Clause 9). This is jointly met by minimizing the number of surfaces and making an optimum choice of coatings. The balance of dichroic coating between maximum optical transmission and maximum IR reflectance (minimum emissivity) presents a direct conflict. The NAOMI system philosophy is that sky cover is very important while a change in the emissivity of one surface from 3% to 6% makes negligible difference to the S/N ratio obtainable at IR wavelengths given the overall system emissivity. Therefore Clause 2 becomes the main driver for the optical throughput of the dichroic. The dichroic coating currently assumed (see section 5.2.1) has the 50% transmission/reflectance cross-over at ~0.8µm.

5.4.2 Specification

This table shows the integrated throughput (in percent) from 0.5 to 1µm at the WFS for guide stars of different spectral type. The model includes the reflectivity of all telescope and NAOMI mirrors, all the air-glass interfaces, the expected QE curve for the EEV CCD and the dichroic transmission curve shown in Fig.13. It then gives the expected number of detected photons for a 16th magnitude star. Viewed together with the sky cover figures this demonstrates that the proposed system meets Clause 2 of the specifications.

Photons/sub-aperture/25 msec, spectral bandwidth 0.5-1 µm, detected by the WFS CCD for stars of these spectral types with m_v=16

5.4.3 Upgrade Specification

The field delivered to the acquisition TV is the full unvignetted field the telescope passes through the derotator, but with some obstruction of the pupil by the WFS pick-off mechanism. With an upgraded acquisition camera it will be possible to carry out some simultaneous IR (approx. >1µm) and optical (approx. <0.8µm) scientific experiments with an AO corrected image.

The FOV of the IR science port is fixed by the size of the dichroic mirror. The baseline design now assumes the FOV is 1 arcmin diameter which matches to a 1024² instrument with 0.04 arcsec/pixel (science Clause 7)

The system is diffraction limited at J and longer wavelengths, in the absence of atmospheric and telescope aberrations, over this 1 arcmin FOV.

At the larger off-axis angles used by the reference stars there are significant aberrations introduced by the telescope and AO optics. These aberrations will produce offsets in the WFS which have to be subtracted before the modal decomposition is carried out. The stability requirement (Section 5.6) means that these non common path errors should not change during an observation.

The system is being designed such that the performance specified in Clauses 1, 2 and 3 is achieved for integrations up to 1 hour, without recalibration, provided the telescope alignment and focus stability is adequate for this. This requirement is driven by Clause 5.

6. Opto-mechanical system design requirements and goals

The constraints and requirements on the optical design of NAOMI are given in section 5.

A diagram of the layout is shown in figure 2. This shows the principle components and their relationship to the derotator and the GHRIL optical table. A brief 'walk through' description of the system is given in and a more detailed list of all the components is given in section 8.1.

The IR science port has a FOV limited initially by the dichroic beamsplitter. In principle it can be up to the full 2.9 arcmin available at the Nasmyth focus, but the telescope/instrument PSF will not be diffraction limited beyond about 30 arcsec from the optical axis at 1.2 µm (J-band). The following numbers are nominal values. Actual designs of IR instruments will need to use the as-built values and a full ray-trace.

The f/ratio is 16.5.

The plate scale is 335 µm/arcsec

The image of the telescope pupil is at a distance of 1100mm beyond the focus and has a diameter of 66.7 mm.

The footprint available for IR instrumentation is as shown in Figure 2.

The optical science port has a FOV of 2.9 arcmin.

The f/ratio is 17.05

The plate scale is 346 µm/arcsec

The image of the telescope pupil is at infinity. i.e. the system at this focus is telecentric.

The footprint available for optical instrumentation is as show in Figure 2.

6.3.1 Acquisition Camera

An acquisition camera will be provided at the optical science port. Initially a low cost CCD video camera will be used. The camera will have a nominal pixel scale of 0.45 arcsecond/pixel. The video camera should be able to detect stars with V<14. The choice of camera will be driven by cost and availability. It will cover the full 2.9 arcminute field in at least one axis. A possible alternative to the video is the use of an ING autoguider CCD. This possibility is currently being investigated.

As part of a system upgrade, it may be replaced by a copy of the Gemini acquisition and HRWFS camera, most probably using the 1024 x 1024 pixel EEV47 CCD. This camera will operate with a smaller pixel scale (0.17 arcsecond/pixel),view much fainter stars and allow detailed inspection of a limited area (= 256 x 256 pixels) at = 10 frames/second.

7. Alignment and Calibration

Two calibration units are required at different locations for alignment and calibration purposes. Provision will be made to insert remotely various sources at the f/11 Nasmyth focus at selected locations within the field. A second point source will be located off axis at the corrected f/17.05 focus to provide a diffraction-limited input wavefront for WFS calibration. It will be positioned just outside the 2.9 arcminute field but within the range of the WFS pickoff stage.

7.1.1 Specifications for the Nasmyth Calibration Unit (NCU)

The NCU will perform several functions:

1. Provide radiation over at least the 0.5µm to 2.2µm spectral region for use by the WFS, the acquisition camera and science instrumentation, e.g. to boresight these components; to calibrate the common-path and non-common-path wavefront errors. Point sources that are diffraction limited in the visible and at K-band respectively will be provided at or close to the optical axis.
2. Provide an on-axis non-diffraction-limited source subtending approximately 1 arcsecond to simulate a time-averaged source viewed under strong turbulence conditions. In some operating modes, e.g. quad cell, the Hartmann spot size affects the WFS transfer curve and thus calibrations should be performed with source sizes representative of anticipated operating conditions.
3. Simulate the WHT exit pupil in a simple manner to check that calibrations made without the central obscuration are valid.Checks will be made at a single field position, probably on-axis, and thus the pupil simulator does not have to be placed at a pupil image.A mask simulating the WHT pupil configuration will be placed in the collimated beam between the DM and OAP2.
4. Map the AO field distortion over the full field for astrometry purposes. The repeatability of the mapping operation will be =< 0.01 arcsecond.
5. Map the optical system aberrations at selected locations over the field of view. Sufficient locations will be selected to adequately characterise the variation of aberration with field position.
6. Introduce small (=< 1.2 arcsecond, frequency <150Hz) motions of an on-axis point source for functional checks of the dynamic response of the AO control system.
7. Uniformly illuminate the f/11 beam so that when a pupil of the system is imaged on to the WFS detector each pixel receives the same signal. This requirement follows from the need to flat-field the WFS detector.
8. Provide an optical feed to a pre-correction camera which is needed to observe the visible image quality before the AO correction is applied.
9. Provide an upgrade capability for an extended ( 40 arcsecond diameter) flat-field source for IR and optical science instrument calibration.
10. Manually introduce a He-Ne laser beam for initial alignment of the NAOMI optical train and for the rapid diagnosis of major misalignment problems.
11. Provide a means of generating a known static aberration for AO system calibration.
12. Allow for the future installation of a turbulence generator for use during laboratory tests.
13. Allow the future installation of a sodium lamp as an upgrade to laser guide star operation.

7.1.2 Specifications for WFS local source

The maximum source brightness will be equivalent to at least a magnitude-8 star representative of Clause 1 conditions and the brightness may be decreased by either a factor of 10 or 100, as desired, to simulate fainter stars. The source spectral bandwidth will cover 0.4 µm to 1 µm.

8. Global Opto-Mechanical Components

The following table lists all the opto-mechanical components of NAOMI.

9. Software Architecture

Fig.14 indicates schematically how the software for the baseline NAOMI system is to be developed using the maximal ELECTRA software re-use policy.

The ELECTRA system software has a number of presentation layer (GUI) processes communicating via DTM (NCSA Data Transfer Mechanism) with application logic located either on the C40s (RTCS) or the host (OPT) or both (VIS). The majority of the presentation layer exists.

The ELECTRA presentation layer also includes an optical bench metaphor GUI. This is an editable tool which has been initially configured for the ELECTRA EPICS mechanisms.

A minimal intervening layer is introduced between presentation and applications.

The maximal re-use of ELECTRA software involves retaining a large proportion of the ELECTRA presentation and application layers, modified as determined by ING, for NAOMI. The ELECTRA sequencer will be retained along with its communication protocols. A communication route shall be provided by which the sequencer shall be able to initiate activity in the RGO/ING central intelligence system and receive status information. A further mechanism shall allow the RGO/ING central intelligence to monitor status objects in the ELECTRA sequencer. The ELECTRA optical bench metaphor GUI will be revised for NAOMI mechanisms.

Sections 10 - 12 describe the software and computing subsystems in more detail.

10. Real-Time Control system

The implementation for the real-time control system is based on an array of TI TMS320C4x parallel Digital Signal Processors (DSP's). Each C4x has the following features:

• Multiply-accumulate optimisation: the processor can accomplish a floating point multiply, accumulate and 2 data moves in 1 cycle (40ns on a 50MHz processor)
• local and global (shared) system buses to avoid contention. When the latter is mapped to, say, a 32-bit VME bus backplane it will provide for the "FIFO" data to be DMA-transferred to a workstation at up to 20 MB/sec.
• 6 x 20MB/s communications ports with DMA engines on-chip. These communicate with other C4xs. This provides for scalability and should allow zero or minimal contention monitoring of the wavefront and reconstruction data
• JTAG scan chain providing an independent parallel debugging path with multiple independent hardware breakpoints.

The specifications in section 4 can be met by a ring architecture of 12 C4x processors, 2 of which are connected to 4 WFS ports each and one of which is connected to the DM drivers. All these internal and external data interfaces go via the C4x communications ports. An additional C40 with a larger memory buffer (32Mbytes) will be connected to the ring and will perform pre-processing of visualization data. The WFS and DM data input and out data rates will permit pipelining of ring data transport operations and all eight ring processors will have access to all slope and mode data.

Booting and command and status interfaces will be via C4x comm-ports using a connectionless protocol to a DTM (NCSA data transport mechanism) process on a Sun SPARC host which will also support debugging via an XDS/ JTAG interface.

An implementation of the above on Loughborough Sound Images and Sun SPARC station 10 is available for the ELECTRA development and testing stage of the project. NAOMI will be provided with a corresponding implementation on Transtech C4x hardware and an embedded Sun SPARC host acceptable to ING.

Procedural (scripting) capabilities will be provided by the ELECTRA sequencer which will be equipped with communications links allowing it to initiate and monitor activity in the ING system. A further link provides a means whereby the RGO/ING central intelligence can monitor status objects within the ELECTRA sequencer.

10.1.1 Upgrade

The system is readily upgradable by addition of further processors.

10.2.1 Wavefront Sensor input

The real-time control system must have a high bandwidth data interface to the wavefront sensor camera. In the case of 4x4 pixels per subaperture, better than 8-bit digitization (12 is currently specified) and no pre-processing within the WFS camera, then an average of 8MB/sec must be accommodated. This would then deliver a readout latency of 250µs after frame transfer.

The implementation will flow the EEV-39 CCD wavefront data into two processors of the twelve processor ring (one per CCD). Ring data distribution can be fully pipelined with data input at the above rate.

10.2.2 Deformable Mirror and FSM output

DM output to the ELECTRA mirror and FSM will be via the C40 comm port to DM DAC card interface developed at Durham.

10.2.3 TTS Upgrade

The real-time processing system can be upgraded to accept tip-tilt sensor data input via either C40 comm port or VME bus. The data can be processed to provide FSM, DM and TTS signals.

The optimisation processing system will be implemented on the host workstation and will have access to all the wavefront and reconstruction data without contention. The initial implementation provided will be that of Gendron and Lena (1995). The requirement for this facility is described in section 4.5.

The optimisation system will also be capable of providing performance estimation to the science data acquisition system for possible inclusion in science data headers.

10.3.1 Upgrade

The real-time optimisation facility may be upgraded to perform future optimisation algorithms on additional processor attached to the C40 ring rather than on the host workstation.

ISO Open/GL and Motif/X11 based software will be available for visualization on any Unix workstation specified by ING which supports these facilities (at the time of writing this is believed to include all commercially-available new Unix Workstations). Visualization tools are available for displaying live or recorded data selectable from the real-time data flow as moving surfaces, images or traces.

This software is primarily concerned with development and fault-finding rather than observational use.

11. Observing System control interfaces.

Scientific requirements A (image quality) and D (dithering) require a link to the telescope control system to allow NAOMI to offset the telescope as DC tip-tilt components are detected.

11.1.1 TCS offset link specification

A method will be provided for writing AO data into science instrument file headers.

11.1.2 Science header data link specification

12. User interfaces

The specification for user interfaces is driven by Clause 6 and Clause 15.

User interfaces will be provided from the ELECTRA programme which will control all the real-time visualization and optimisation functions. Software will be provided to flow the operational command and status via the ELECTRA procedural system (the sequencer) and thence if necessary to the ING DRAMA environment. The ELECTRA user interface will be modified to cope with the new RGO and ROE mechanisms and will also access these via the DRAMA-based procedural layer.

A full upgrade path shall be maintained whereby all user interfaces may be re-implemented to future ING GUI standards.

13. Observational capabilities and procedures

Observational capabilities of the system design have been tested against observers requirements for infrared imaging and spectroscopy. Procedures will exist to allow observation of optically bright and optically faint/invisible sources provided they can be detected with the infrared instrument, infrared bright and infrared faint/invisible sources provided they can be detected by the WFS. There is no guaranteed acquisition method for sources which are faint at both optical and IR wavelengths. For infrared spectroscopy it is assumed that an internal camera for viewing the slit field is incorporated in the instrument, if an offset position from a close (< ~1 arcmin.) bright source to within the pick-off repeatability accuracy is not available. The same accuracy will be required for imaging if spatial registration against other images is needed which cannot be determined post-observing. See also section 3.4.3 on non-sidereal rates.

In the baseline implementation the position registration will be defined by the relative position of the WFS pick-off to the optical axis (or WFS pick-off datum). Acquisition will be by the pre-correction camera.

Automation of acquisition procedures (e.g. using `drag-and-drop' on visible sources to position either the telescope or WFS stages) will be possible with an upgraded implementation of the acquisition software and acquisition camera.

13.1.1 Specification

All the following procedures will be possible but some will not be optimized initially.

• system startup
• system configuration
• WFS calibration
• optical alignment
• determine rotation centre and calibrate TCS pointing ( may be manual input in baseline system)
• acquire setup star
• set telescope focus
• determine seeing parameters
• set up target parameters
• acquire science object
• align science target/guide stars
• determine closed loop parameters
• make science instrument exposure, including dither operation if necessary
• write data file with necessary headers (position, time, key NAOMI settings)

Greater observing efficiency (Clause 6) will be possible if full image analysis facilities are available to the AO system to allow automated calibration and image size minimization to be carried out in a loop mode. Improving the functions implemented and extending the number of them is expected to occur as a continuing process driven by what is learned from the initial delivered system. The purchase of an upgraded acquisition camera and more sophisticated software to implement point-and-drag acquisition would significantly improve user efficiency.

14. Telescope, GHRIL and supported instruments

This section details the relevant working assumptions which the NAOMI specifications make about the telescope and its instruments.

Final end-to-end performance predictions will require the following telescope data:

1. Formal values for the amplitude of WHT aberrations up to Zernike radial degree three. Interim values have been available for system design. Errors on these scales are correctable by NAOMI and so the issue is one of DM stroke.
2. The phase variance on spatial scales corresponding to less than 57cm on the primary (i.e. errors which NAOMI cannot correct).

The improvement of intrinsic WHT seeing to free-atmosphere levels is regarded as outside the scope of the NAOMI project. The exception to this is Clause 21 which constrains introduction of additional seeing by the NAOMI systems themselves.

NAOMI performance predictions strongly support efforts to reduce local seeing.

The taking and reduction of JOSE data is a high priority requirement for the NAOMI project. It is the primary means by which more effective optimisation algorithms can be developed [Clause 2, Clause 6] and by which the accuracy of AO performance modelling can continue to be improved. Many nights of JOSE have have now been obtained and analysed. The results are being used to refine certain specifications such as those for the FSM.

GHRIL issues are partly covered by section 16 (see also clause 17). The following NAOMI requirements with respect to the GHRIL are noted here.

1. Global (GHRIL-wide) EMC screening strategy
2. Assurance testing on GHRIL earthing
3. GHRIL global dust control
4. Thermal insulation of GHRIL optics room from GHRIL control room
5. GHRIL heat removal policy and general minimization of heat sources in the GHRIL area

Information on any specific instruments to be supported should be supplied by ING.

15. Software and computing standards; maintenance issues

See the requirement in Clause 14.

The software for ELECTRA which is taken into NAOMI is being developed according to the distributed ELECTRA standards. These include a coding standard, source code control standards, and the use of DTM, Motif and Open/GL.

Coding standards for NAOMI AND ELECTRA will be adjusted and revised to the provided ING standards.

An revision of all software to future ING standards may be provided.

16. Handling and operational support requirements

16.1.1 Introduction

As can be seen from Clause 13 these requirements are largely driven by standards set by ING. The optical chassis design allows a modular implementation mechanically, with electronics also in a small number of racks. The handling concept is that the racks and opto-mechanical modules can be removed mostly by hand (two people) but a small hoist or additional hydraulic attachment is likely to be required to install and remove the WFS. Removal and installation is expected to take several hours (less than one day) with this model. Installation on a similar optical bench elsewhere to allow some pre-use alignment, fault tracing and maintenance will be possible but will not have full GHRIL-level facilities. The GHRIL control room will be cooled so that electronics racks will not need individual cooling in addition to normal fans. Very little heat is dissipated on the GHRIL bench itself and NAOMI components there will not be cooled. These issues are currently covered by a Strawman Support Engineering proposal, given below. However, the concepts may need to be revised as changing budget conditions influence funds available for infrastructure support.

16.1.2 Specification

16.1.1 Dust control.

The proposed system is to have covers over modules; the number and shape chosen to fit final layout but with a goal of not more than four. The GHRIL room itself will have a positive pressure environment with filtered air input.

16.1.2 Thermal control of GHRIL optics room components

This is control of heat generated by bench components and nearby electronics (see Clause 21).

The proposed system is local cooling of all heat sources, either by water or air; heat taken to global GHRIL environment heat removal system. All electronics heat sources not associated with motors and drivers which must be on the optical bench should be above or away from the bench.

16.1.3 Thermal control of GHRIL control room electronics

This is control of computing and any other equipment required to be in vicinity of the GHRIL room especially the GHRIL control room (see Clause 21).

Ideal system

1. good insulation between the two GHRIL room and other heat source;
2. local cooling of other heat sources with heat removed either by the GHRIL environment system or the equivalent for the appropriate area.

The proposed system is:

1. establish good insulation between the GHRIL optics and control rooms as soon as possible.
2. Install a glycol cooling system which cools the entire GHRIL control room and removes heat to outside the dome.

1. a global heat removal system from the GHRIL area is certain to be required, especially if the concept of an enclosed GHRIL room is maintained.
2. the GHRIL area must not be allowed to warm up because
1. an excess temperature in GHRIL over that of the dome air of more than one or two degrees is likely to induce significant seeing degradation at the temperature interface and
2. the infrared thermal background increases with temperature and affects progressively shorter wavelengths.

16.1.4 Thermal control of optical component surfaces

This refers to control of optical component surfaces (see clause 17).

The Ideal would be that optical surfaces could be cooled to ambient (dome) air temperature and stable to < 1°C (although the concept of cooling to close to dew point would reduce thermal emission even more, it would also induce local convection currents which would be a worse problem than the thermal emission).

The proposed solution is to constrain local heat sources and to remove heat from GHRIL control generally but not to control directly the optical surface temperatures. The Reason is that given the number of surfaces which cannot be cooled (telescope) or it would not be wise to cool (DM, FSM), the differential gain from cooling the remaining two surfaces before the IR instrument would be negligible.

16.1.5 EMC control and protection

See Clause 18.

(this section to be developed further after current investigation by the project engineer into EMC suppression and rejection are completed).

The Ideal is (i) to reduce EM emissions in the GHRIL environment (Faraday cage?) and (ii) to have integrated system electronics grounding design and to take all standard precautions such as keeping signal and power cables well separated throughout system.

It is Proposed: (i) to measure the properties of the current EM environment at GHRIL; (ii) use E0 and E1 opportunities to learn more about the EM environment; (iii) to have integrated system electronics grounding design and to take all standard precautions such as keeping signal and power cables well separated throughout system; and (iv) to assign a group as responsible for EMC overall, including full system grounding.

The E0 run indicated that whilst there was certainly a worse EM environment than Durham there were no deleterious effects on any component (including the DM).

16.1.6 System Testing and Installation.

Ideal: Fully integrated stand-alone GHRIL bench with `umbilical' attachment to environment control; can be lifted as a unit after removing umbilical, placed on a handling trolley, taken to separate lab and operated fully with (e.g.) its calibration unit for alignment and testing.

Proposed: Modular build; accurately re-locatable modules on the optical bench and ~ five electronics racks; power and other supplies connected through well-designed umbilicals. All on-bench units except the WFS are removable by hand. The WFS is a single 70Kg unit and will need a hoist for lifting; simplest possible handling trolley to allow safe transport of components; another optical bench to re-assemble modules for limited testing; this bench will have some attachments for NAOMI electronics, power and signal I/O but will not fully simulate the GHRIL facility.

Reason: Opto-mechanical modules have now been defined; some work needs to be done on estimation of electronics racks requirements. (This solution is likely to be the cheapest which meets the La Palma requirements for instruments).

17. Lifetime and downtime

See clause 16.

The adaptive optics components and associated electronics will be designed for a minimum of 3000 hours of operation with a goal of 10,000 hours. Parts that will fail before that time will be identified , together with the cost and delivery of replacement parts.

As a goal no more than 5% of the available observing time should be lost to AO facility failures.

18. Maintenance

System maintenance will be performed in accordance with ING procedures (see, for example, Clause 13)

19. Safety

See Clause 13.

Potential safety hazards will be identified and the appropriate measures will be taken to protect personnel, e.g. warning notices, covers with interlocks. Handling procedures and simple lifting aids, e.g. eye bolts, will be provided for heavy items. Only a simple handling trolley will be needed.

20. Documentation

See Clause 14.

Documentation will be include a Users' Manual, full system engineering diagrams as built, maintenance procedures and trouble shooting guidelines. The WP specifications and the WP Cover document define these requirements in more detail. The documentation approach will be to recognize that the system must be supported by staff who have good appropriate mechanical, electronics and software engineering skills but who did not build the system.

21. Future Development

Key development areas are projected to be:

• improved TCS and Instrument Control System integration, giving increased observing efficiency
• more sophisticated real-time control algorithms, allowing better optimisation between system performance and conditions.
• turbulent layer conjugation facility (capability designed in)
• a laser-upgrade (capability designed in)

22. References

1. R.W.Wilson and C.R.Jenkins, "Adaptive Optics for Astronomy: Theoretical Performance and Limitations", MNRAS 278, p39 (1996)

2. E. Gendron and P. Lena "Astronomical Adaptive Optics 2. Results of an Optimised Modal Control", Astron. Astrophys Supp., 111, p153 (1995).

23. Modelling Assumptions

The performance of NAOMI has been modelled using a number of techniques:

1. Using established scaling laws (RAH; RMM)

2. By analytical methods (CRJ, RWW)

3. Using numerical techniques (APD)

The common assumptions have been Kolmogorov turbulence and a constrained level of aberration in the WHT (i.e. the models are not end-to-end).

Input data on timescale and C_n² turbulence profile and strength parameters have been limited up to the present but JOSE statistics are now being used as they become available.

List of Figures.

• Figure 1: A NAOMI system block diagram.
• Figure 2: A schematic illustration of the optical layout for NAOMI.
• Figure 3: The 2.2&#181;m point source sensitivity for NAOMI.
• Figure 4: The NAOMI system footprint on the deformable mirror.
• Figure 5: The gains in Strehl ratio for a given off-axis angle.
• Figure 6: Comparison of on-axis NAOMI performance.
• Figure 7: Angular decorrelation for simple zonal correction.
• Figure 8: Angular decorrelation for modal correction.
• Figure 9: All-sky and galactic latitude guide star probabilities
• Figure 10: The effect of system order on Strehl ratio (H-band)
• Figure 11: The effect of system order on Strehl ratio (K-band)
• Figure 12: Fitting error gains from linearised segmented adaptive mirror technology.
• Figure 13: A dichroic mirror transmission/reflectance curve
• Figure 14: Schematic illustration of NAOMI software architecture