SYSTEM DESIGN NOTE

SDN0002.01 - Mechanism Specifications

Prepared by	Date	Approved by	Date	Rev.	Rev Date
Jay Elias	2/3/99	N. Gaughan	2/5/99

1. Introduction

This document is intended to discuss the various requirements on mechanisms. It is intended to be comprehensive, although it may not succeed at this in its initial draft.

2. Overview of Requirements

While each mechanism may have individual requirements, such as the number of positions in the grating turret, there are several generic requirements that all mechanisms must meet, as well as several system-level requirements that must be apportioned among the mechanisms that are affected. These are discussed below.

In the discussion that follows, it is assumed that the configuration permits insertion of a flip mirror after the collimator that diverts light directly to the camera turret, so that the grating and cross-disperser prisms do not need to be moved to acquire objects.

2a. Generic Requirements

The mechanisms must operate at cryogenic temperatures and must be reliable, with periodic maintenance at reasonable intervals, over a 1-year lifetime.

The design of the mechanisms must be such that they do not produce excess background on the detector or noise pick up on the detector read-out during integration or read-out of the detector. See the notes at the end regarding thermal background. Dark current is assumed to be no more than 0.1 electron/sec/pixel, and instrument background should not exceed this value. Instrument background includes leaked thermal radiation as well as scattered light and ghost images.

2b. System Requirements.

Flexure. There are three flexure requirements imposed on the instrument as a whole:

 *The motion of the image of the telescope secondary on the cold
 stop of the fore-optics (Offner relay) may not exceed 1% of the
 stop diameter. Since there are no mechanisms preceding the
 cold stop, this requirement does not affect mechanisms.

 *The motion of the image of the object on the slit, when the
 OIWFS is functioning, should not produce light losses in
 excess of 5% with a 0.1 arcsec slit for integration times of
 one hour. In practice this is interpreted as allowing
 differential flexure of approximately 12 microns between the
 OIWFS detector and the spectrometer slit for any 90-degree
 change in gravity vector. The only GNIRS mechanisms affected by
 this requirement are the filter wheel(s) and slit wheel.
 However, they can only be allocated a portion of the overall error
 budget, because the GNIRS structure, Offner, and OIWFS elements
 must also be included.

 *The motion of the spectrometer slit when re-imaged on the
 detector should be less than 0.1 pixel in 1 hour of observation.
 In practice this is interpreted as allowing differential flexure
 of approximately 10 microns between the slit and the detector
 for any 90-degree change in gravity vector.
 In principle this applies to both the acquisition and
 spectroscopic configurations, although it is formally
 specified (and most important) only for the latter. Note
 that again this budget must include the flexure of fixed optics
 and the support structure in addition to the mechanisms.

Repeatability. There are four repeatability requirements that I would like to impose on the instrument as a whole:

 *Reconfiguration between acquisition mode and spectroscopic
 mode should not produce light losses in excess of 5% with a
 0.1 arcsec slit. In practice this would be interpreted as
 allowing a shift of up to 3 microns perpendicular to the slit
 axis. The two GNIRS mechanisms affected by this are the filter
 wheel(s) and slit wheel.

 I assume errors RSS. If the budget for the slit wheel from
 requirement 3, below, is 1.35 microns, the budget for
 each filter wheel is 1.9 microns. Since the idea is to keep
 the image centered when filters change, the filter wheel error
 includes differential alignment errors. Although the most
 common switch will be between a wide field and a slit, it
 is desirable to have all slits on a common center, in which
 case the slit error includes relative alignment errors.

 *Reconfiguration between acquisition mode and spectroscopic
 mode should not produce an image shift of over 3 microns along
 the slit. The GNIRS mechanisms affected by this are the filter
 wheel(s). There is a weak constraint on the decker wheel if it
 is moved, at about the 1 pixel (30 microns) level.

 The error budgets for the filter wheels are therefore 2.1
 microns each. Again, this includes differential alignment
 errors between filters.

 *Reconfiguration between acquisition mode and spectroscopic
 mode should not produce motions in excess of 0.1 pixel of the
 slit image on the detector. This affects only the slit and
 flip-in mirror, provided the red camera provides adequate
 K band performance for acquisition [which needs to be
 confirmed].

 This is a motion of 2.7 microns, which is allocated equally to
 each axis at 1.9 microns. Only the flip mirror affects motion
 perpendicular to the dispersion, so it gets 1.9 microns in
 this direction. Along the dispersion the allocation is to
 the slit wheel and flip-in mirror at 1.35 microns each. (There
 is some, limited value to having the acquisition image and the
 spectrum fall on the same row of the detector exactly; this
 would imply alignment of the flip-in mirror tilt to allow
 this.)

 *Reconfiguration between different spectroscopic modes should
 be repeatable at the level of 10 pixels or better on the
 detector. All mechanisms are potentially included in this, but
 since the filter and slit wheels must meet a tighter
 requirement, and since the decker wheel is presumably of a
 design similar to that of the slit wheel, most of the budget
 is allocated to the grating, prism, and camera turrets.

 The total motion is 270 microns. The error budget for the
 filters and slits are 2.7 microns as worst (the flip mirror
 does not contribute here). If we allocate 27 microns to the decker,
 the grating, prism, and camera turrets have a remaining allowable
 error of 268 microns or 155 microns each.

Reconfiguration. There are two requirements on the instrument:

 *Reconfiguration from acquisition to spectroscopic mode or
 within acquisition mode should not take more than 1 minute.
 Even faster would be better! Such reconfigurations involve
 possible motions of the filter wheels (possibly both), slit
 wheel, and flip-in mirror. It would be useful but not
 essential to move the decker wheel as well, so the issue is
 moving up to 5 mechanisms in 1 minute -- whether in parallel or
 sequentially, so long as the reconfiguration is accomplished.
 In may cases the only mechanisms moved would be the slit wheel
 and flip-in mirror. A change in focus should not be required
 except (perhaps) for K-band imaging with the red camera.

 *Full reconfiguration to take place within 3 minutes. This
 means moving every mechanism in that time, in extreme cases.

Errors. In the discussion that follows, it is assumed that errors RSS. This is not necessarily correct in all circumstances. Since the repeatability errors are random, it is fair to assume that they have a characteristic dispersion which does indeed RSS. (If there is a systematic component, then it is by definition NOT repeatability error,
but rather an alignment error.)

Alignment errors, on the other hand, can either be limits on our ability to measure -- in which case they are effectively random – or on our ability to adjust (or a decision not to make finer adjustments). In the latter case the alignment error will be known, and such errors should add linearly. Note that because they are added, two effects may compensate if of opposite sign. Thus one might in principle adjust one mechanism to compensate for an error in another. Since most of the elements in a configuration -- filter, grating, prism, camera – are used in multiple configurations, this is likely not to be possible in practice.

 3. Individual Mechanisms

3a. Filter Wheels

Currently, there are two filter wheels in the design. There is no requirement for two wheels, but there must be at least 14 positions, comprising the following:

7 order sorting filters
cross-dispersion blocker
dark position
K-band acquisition filter
at least 1 custom position
cold stop viewing lens #1
2 focus masks

If there are two wheels, each wheel must have an "open" position. Operationally, it is highly desirable to have more than one custom position for user filters and upgrades; a good number would be >3.

The focus masks are used in conjunction with a filter, so with two wheels they are in one wheel and the order-sorting filters are in the other. With only one filter there is less flexibility as the masks must be set up with a specific filter in the wheel.

Filters are tilted to reduce ghosting; the angle is 2.7 degrees (check). The filter optical thickness is constant in order to maintain focus; this means that the physical thickness will differ from filter to filter. Some filters may be in two parts.

Repeatability and Alignment.

Translation of the filter does not produce motion of the image on the slit, so only tilts matter in this context. The optical thickness of the filters is subject to negotiation with the vendor, but for the worst case (3 mm at infinite index) the tilt sensitivity is 3 microns/mrad. In addition, for a rotation about the optical axis there is a motion along the slit of 0.14 microns/mrad. If we allocate 1.9 microns to tilts about both X and Y, then the error produced by rotation about the optical axis must be <0.9 microns. This gives us:

rotation about axis parallel to slit:   <0.6 mrad
rotation about axis perpendicular to slit: <0.6 mrad
rotation about optical axis:   <6 mrad

These errors must in turn be allocated between filter-to-filter alignment and mechanism repeatability. If we allocate them equally, they come out as follows:

alignment about axis parallel to slit:  <0.37 mrad
alignment about axis perpendicular to slit: <0.37 mrad
alignment in rotation about optical axis: <3.7 mrad
filter wheel tilt errors parallel to slit: <0.37 mrad
filter wheel tilt errors perpendicular to slit: <0.37 mrad
filter wheel rotation errors:  <3.7 mrad

Filter translation errors do affect overall transmission in that filters are not absolutely uniform. If we assume that filters are uniform to 5% (fairly conservative) and that we want to flat-field to 1%, then the translation error cannot exceed 1/5 of the "blur" at the filter. If the closest filter wheel ends up 80 mm from the slit, the blur is approximately 5 mm and the allowable translation error is 1 mm. For the allowable rotation error listed above (3.7 mrad) the translation error is <1 mm for wheel radii <270 mm.

If the wheel is driven by counting steps, one full motor step must be less than 3.7 mrad. A standard stepper motor has 5 mrad/full step, so a gear reduction of 5:1 or greater ensures that stepper digitization doesn't contribute to the alignment error.

3b. Slit Wheel

The slit wheel currently contains 10 positions, comprising

7 slits
"open" position
2 IFU module positions

The original (CoDR) specifications called for 8 slits; one was sacrificed for the second IFU position. The 8th slit would be useful but should not drive redesign.

Repeatability and Alignment.

It is important to understand that the tightest requirements on the slits are in the perpendicular direction; since the slit ends are  defined by the decker wheel (and are not critical anyhow), the tolerances in this direction are much looser. This suggests a design in which the slits are tangents rather than radial.

In addition to the alignment and repeatability tolerances described above, the slits must all be parfocal. We adopt a maximum allowable defocus error of 35 microns (check this). This is the value for differences between slits as well, for consistency. Note that absolute focus does not need to be to this precision, since the OIWFS focus motion can be
used to correct for this as part of instrumental alignment. The absolute focus tolerance is therefore about 1 mm.

The slits must also be aligned on the detector. Although the detector itself can be rotated to align with the slits, it can only be adjusted as part of alignment, and will therefore be aligned to a value corresponding to the average slit position. Thus, it cannot be used to correct for slit-to-slit alignment errors, but can correct for systematic effects, so the absolute slit alignment needs to be done only relatively coarsely. The appropriate alignment procedure is therefore to do a "best effort" absolute alignment on one slit and then use that as a fiducial to align all the other slits
in the wheel.

The slits have 6 degrees of freedom:

tilt about the slit axis: immaterial for reasonable values
tilt perpendicular to the slit axis: affects focus at slit ends
rotation about optical axis: affects alignment on detector
displacement along slit axis: immaterial for reasonable values
displacement perpendicular to slit axis: affects relative position on array
displacement along optical axis (in/out of focal plane): affects focus

If we take one slit (presumably the narrowest) as fiducial and align all other slits relative to it, in the sense of comparing positions when each slit is "in the beam", we see that focus is affected by 2 degrees of freedom and slit alignment is also affected by 2 degrees of freedom. If the allowable errors are further divided between alignment and repeatability, provisionally on an equal basis, one gets:

Alignment along dispersion. The error budget for this from the breakdown above is 1.35 microns. This refers to average position and does not include rotation about the optical axis (=alignment of slit with columns on the array). For the latter, the maximum allowable end to end error is 0.1 pixel. If we assume that the array is accurately aligned to the array, then the other slits can have the full 0.1 pixel error, again divided equally between alignment and repeatability. (Note that when the fiducial slit is used in practice, it will show repeatability error but not alignment error.)

So we get:

alignment perpendicular to slit axis:  <0.7 micron
repeatability perpendicular to slit axis: <0.7 micron
rotation alignment about optical axis: <0.07 mrad
rotation repeatability about opt. axis: <0.07 mrad

For a slit 60 mm in length, 0.07 mrad corresponds to 4.2 microns difference end-to-end. If the slits are mounted tangentially on a wheel of 200 mm radius, a rotation error of 0.07 mrad corresponds to 14 microns at the edge of the wheel.

If the slits are mounted tangentially, the main source of repeatability error in the alignment perpendicular to the slit axis is play in the wheel axis (bearings). Note that if it is not possible to align the slits to 0.7 micron in the wheel, it should be possible to determine the relative alignment empirically and then offset the telescope when slits are changed. This would be an additional software task. In that case, the repeatability perpendicular to the slit axis
becomes <1.35 microns and the alignment perpendicular to the slit axis now must meet the 10 pixel requirement, or 300 microns.

Focus. If the allowable focus error is 35 microns (check), the allowable error at the slit center is 25 microns and the allowable end-to-end difference is 50 microns for a maximum error at the ends of 35 microns, assuming these RSS. (That is, that our ability to measure is limited. Known and measurable alignment errors must add linearly.)
Again, these are divided provisionally equally between alignment and repeatability.

35 microns end to end on a 60 mm slit is 0.6 mrad. For a wheel 400mm diameter 0.6 mrad corresponds to 240 microns at the edge of the wheel, which suggests that the slit center focus repeatability of 18 microns in fact defines much better tilt repeatability; some reallocation of errors is probably justified. Without doing that, we get:

slit alignment along optical axis: <18 microns
slit repeatability in focus  <18 microns
slit tilt perpendicular to slit: <0.6 mrad
slit tilt repeatability:  <0.6 mrad

A reallocation based on the discussion of repeatability above and the fact that both tilt and center focus are differential measures gives the following:

slit alignment along optical axis: <22 microns
slit repeatability in focus:  <22 microns
slit tilt perpendicular to slit: <0.5 mrad
slit tilt repeatability:  <0.06 mrad

The discussion above assumes that the slits are absolutely uniform: that there are no irregularities in their edges. If there are, any motion along the slit axis will project these irregularities at a different place on the array. For the rotation repeatability  given above, the resulting positional error (14 microns) corresponds to about 1/2 pixel on the array. This means that to flat field to 1%, the pixel-scale irregularities must be <2% of the slit width for the narrowest slit, or ~1.2 microns. Greater irregularity implies a tighter tolerance on the wheel rotation repeatability.

Again, if the wheel position is defined by stepper motor position, we want 1 step to be <<0.07 mrad. A gear reduction of 200:1 gives 0.025 mrad/step. This then requires operating the stepper at 200 rpm or faster (40,000 Hz) in order to meet the mechanism speed requirements.

3c. Decker Wheel

The decker wheel contains deckers (British "dekkers") defining the slit length. There are:

5 deckers
2 IFU modules
Lyot stop viewing lens #2

The deckers should be as close to the focal plane as possible in order to minimize the partially vignetted region at their ends. If they are located 6 mm behind the focal plane, the vignetted region is approximately 13 pixels.

Repeatability and Alignment

Since the deckers are changed only as part of major reconfigurations, they only contribute to the "major alignment" error budget; their share is set as 155 microns (see above). If the mechanism is essentially the same as the slit wheel, the repeatability in this direction (rotation) is ~14 microns (200 mm radius) and so an alignment error of <150 microns is acceptable.

Relatively large errors in the other 4 degrees of freedom are allowed, at least 10x those required from the slit wheel.

3d. Flip Mirror

The flip mirror (if provided) has a single active position and a position out of the beam. For the "in" position, there are only 2 degrees of freedom where small shifts affect performance, namely tilt of the mirror about the two axes in the plane of the mirror. Repeatability is discussed above. Absolute alignment is less critical, and the precise value of the alignment error is debatable. It clearly must be good to about 10 pixels, and an error of ~1/2 pixel along the slit would be nice. This latter values has been applied below but should be revisited if the specification proves unduly demanding.

A tilt of 0.1 microrad of the flip mirror produces a displacement of the slit image on the long camera of ~0.3 micron; the displacements with the short camera are 3x less so the long camera drives the performance requirements.

Therefore:

tilt repeatability about axis parallel to dispersion: <0.0006 mrad
tilt repeatability about axis perpendicular to disp'n: <0.00045 mrad
tilt alignment about axis parallel to dispersion: <0.0045 mrad
tilt alignment about axis perpendicular to dispersion: <0.09 mrad

The repeatability errors of ~0.5 microradians are VERY demanding. If the mirror in its "in" position is defined by 3 points defining the 2 axes spaced ~200 mm apart, the location of the mirror at each point must repeat to an accuracy of ~0.07 micron.

If this requirement cannot be meet, there is a mode of operation that permits the instrument to function.

In the case where the repeatability requirements are met, the acquisition process would occur by first observing with the mirror in and an open position in the slit wheel. The object would be identified at the telescope would then be offset so as to place the object at the location where the desired slit would fall. One might then confirm the positioning by moving the slit into position, but it would not be strictly necessary.

If the flip in mirror were less repeatable, then one would measure the object position in the wide field, then move the desired slit and measure its position on the detector, and then calculate the required telescope offset. Note that one could probably still predict the location of the spectrum on the array using the location of the slit ends in the acquisition and spectroscopic modes.

For this second operational mode, the allowable tilt errors become roughly 0.09 mrad, or errors of ~13 microns at the mirrors defining points.

3e. Cross-Dispersion Turret

The cross-dispersion turret carries a mirror for long-slit modes, a cross-dispersion prism for each camera (two total) and a polarization analyzer (=Wollaston prism or polarization analyzer). The Wollaston prism is used with both cameras.

As indicated above, the cross-dispersion turret is allowed to produce up to 155 microns displacement of the slit image on the detector, in each axis. If we allocate this between alignment error and repeatability error, we get 110 microns for each. As calculated in 3d, this is equivalent to a tilt error of 0.037 mrad.

If the turret is a mechanism rotating about an axis 200 mm in length and has a radius of 100 mm (these numbers are meant to be illustrative), an error of 0.037 mrad in rotation corresponds to 3.7 microns at the radius of the turret, and to approximately 5 microns at the ends of the turret axis if the errors are random, and 3.7 microns if they are correlated.

Because the turret is in a collimated beam, motions in other degrees of freedom are less significant. The rotation of the prism about the optical axis does have an effect on the location of the orders on the array, but alignment to a few mrad should be good enough. Since repeatability must be similar to the tilt in other directions, its effects will be negligible.

If stepper motor position is used to define location, we need 1 step to be <<0.037 mrad. A 200:1 reduction gives 0.025 mrad/step, while 500:1 gives 0.010 mrad/step. In the latter case, running the motor at 200 rpm gives a maximum positioning time of 2.5 minutes, which is still acceptable for this mechanism.

3f. Grating Turret

The grating turret is used to both select one of three gratings and to set the tilt of individual gratings. If a flip-in mirror is present, there does not need to be a mirror on the grating turret.

The gratings must be mounted at tilts matched to their individual blaze angles.

The range of tilt for each grating ranges from just under 14 degrees for the high-dispersion grating to ~2 degrees for the low-dispersion grating.

In principle, grating tilt should be adjusted about the grating vertex. However, if the grating is made oversize and the radius vector is roughly normal to the grating surface, tilts about another axis will produce only small displacements of the beam on the camera aperture and will not affect image quality.

The alignment and repeatability errors for the grating turret are similar to those for the cross-dispersion turret. If the axes are perpendicular, and if errors in one direction are easier to minimize than in the other, then the error budget can be re-allocated. However, the gain would not greater than sqrt(2) in the best case.

It is desirable but not critical to have the spectra exactly orthogonal to the slits. In order to achieve to 0.1 pixel end to end, the gratings must be aligned and repeatable in tilt about the optical axis to a combined error of 0.1 mrad. Spectral extraction routines can easily cope with a tilt of a few pixels over the width of the array, which amounts to a few mrad error.

Since the desired repeatability error about the axis of rotation is about 0.037 mrad, (see above), the alignment spectrum alignment error is dominated by initial alignment and not by repeatability.

The stepper motor requirements for this mechanism are essentially the same as for the cross-dispersion turret.

3g. Camera Turret

The camera turret contains 4 cameras, barring a design revision. It may also have an intermediate position to allow pumping down the optical volume through holes in the turret.

The alignment of individual lenses will cause some displacement of the spectrum on the array. If the errors are allocated as above, the total alignment error of the camera must be <110 microns, distributed among the elements, and the positioning error would also be <110 microns in each axis. This corresponds to a somewhat greater error at the outside of the turret. If the radius to the camera aperture center is 150 mm the allowable error is 0.7 mrad in rotation.

It is clear that this requirement is considerably easier to meet than the grating turret or cross-dispersion turret requirements, which suggests that the camera turret error budget might be reduced in order to help out the other two mechanisms.

The camera turret must also maintain focus. The alignment is not critical, because focus shifts can be determined and corrected with the focus drive. Repeatability is, however, essential. The allowable focus error is set by the short cameras and is 12 microns according to Ming's error budget.

If stepper motor step is used to define camera turret position, less gearing down is required by positional tolerances than for the grating drive or grating turret. A reduction as low as 50:1 is adequate even if the error budget is tightened, and the reduction may well be driven by mechanical requirements.

Effect of external mirror.

The calculations for the camera turret given above assume that the optics are wholly contained within the turret. If they are not, then the positioning is more sensitive.

Specifically, for the long cameras it is proposed to have the beam reflected from a flat behind the initial doublet though and angle of close to 90 degrees, out to a flat, and then back and down do the detector. There is a field-flattener at this point for the blue long camera. If we assume that the external flat is approximately at the half-way point of the camera focal length, and neglect any power in the field-flattener, then a rotation of the turret of 0.1 mrad will produce a motion of 150 microns on the array. This effect dominates the effects due to motion of the lens centers, so the allowable error reduces to ~0.07 mrad.

This rotational tolerance is only slightly more forgiving than that for the grating and cross-dispersion turrets, and so one would tend to allocate the errors more nearly equally.

3h. Focus Drive

The focus drive is intended to set detector focus to correct for errors in camera focus and partially correct for errors in collimator focus. Since the error budget for short camera focus is ~12 microns, the focus drive should be reproducible to less than this -- 6 microns or better -- and should also not introduce tilts on the array of similar size.

It should also not rotate the array about the optical axis by >0.1 mrad over the working range of travel. Displacements are assumed to be small, but should be evaluated within the overall reconfiguration budget (i.e. 270 microns at the detector). Note that this requirement is less stringent that for an instrument such as NIRI where there is refocus for every filter change, so that it is desirable to minimize translation in the focal plane to <<1 pixel.
 

---

---

If you have any questions or suggestions regarding this website, please contact Melissa Bowersock.

NOAO Intranet Services

NOAO Copyright  Statement