ING Banner
Home > Astronomy > ISIS > Cryostat alignment and focusing
Text: Normal textBigger tyextBold text

Cryostat alignment and focusing

Kinematic mount

A rigid body has six spatial degrees of freedom - three of translation along the x, y and z axes, and three of rotation (roll, pitch and yaw) about the x, y and z axes. A kinematic mount is a mounting system which allows each degree of freedom to be singularly constrained during alignment. The simplest type of kinematic mount is the cone-groove-flat mount, whose kinematic constraints are depicted schematically in the figure below (from Newport).

In this arrangement three ball-tipped supports rest in a cone and groove and on a flat panel respectively, on a fixed frame. Formally the cone hole should be trihedral-shaped to ensure only three contact points with the ball, but often a conical shape is used for manufacturing simplicity. The cone constrains the three translational degrees of freedom, the groove constrains two rotational degrees of freedom, and the flat constrains the third rotational degree of freedom. In this type of kinematic mount there is some "cross-talk" or coupling between the linear and rotational axes under adjustment.

Capstan adjustment

The ISIS cryostats are attached to the mounting rings on ISIS using kinematic mounts with three adjustable ball-tipped capstans, A, B and C. The A-capstan ball mounts in the cone, the B-capstan ball in the groove and the C-capstan ball on the flat. The following page provides a pictorial explanation of the mounting arrangements for an ING cryostat (not an ISIS cryostat, but the difference is only in the clamp design). Alignment of the cryostat to optimise tip and tilt across the detector, and to get coarse spectral focus, is carried out by adjusting the capstans by amounts recommended by the tilt setup script isis_tilt.

The capstans are clamped with M5 bolts, and it is this arrangement which holds the cryostats in place. All three M5 bolts of a cryostat should never be completely loosened simultaneously; this would cause the cryostat to fall. The M5 bolts have replaced the original hand-tightened knurled brass knobs. By tightening these bolts with a torque screwdriver always to the same torque (3Nm) after each adjustment cycle of the capstans, the stability and repeatability of cryostat alignment has improved significantly.

The torque screwdriver is stored in the exchangeable optics cupboard in a clearly marked box. To use it first insert the M5-size hex bit (marked with green ink) in the screwdriver's socket. The screwdriver can then be used in two ways by (i) rotating it directly using its hand grip as you would a normal screwdriver, and (ii) attaching the ratchet handle to its top end, and twisting the handle back and forth. In both cases when a torque of 3Nm is reached the torque-limiting clutch disengages, ensuring further torque can't be applied. When using the ratchet handle the direction of torque coupling (anti-clockwise for loosening, clockwise for tightening) is set appropriately by toggling the silvered knob on the back of the ratchet handle. Always check that the torque limit is set to 3Nm1 before using the screwdriver to tighten the capstan bolts, and in use keep the screwdriver well-aligned with each M5 bolt's rotation axis to ensure tightening to a consistent torque.

To adjust a capstan first loosen its Allen-bolt lock with an Allen key (3mm, stored in the same box as the torque screwdriver). Next, whilst supporting the cryostat with one hand, loosen the capstan's M5 bolt using the torque screwdriver so that the capstan can be rotated with ease. If applying large adjustments to a capstan, e.g. greater than one quarter turn, also loosen slightly the M5 bolts of the other two capstans to prevent build-up of mechanical tension (small adjustments to a capstan can be effected by loosening only its M5 bolt). The capstan clamps should be re-tightened using the torque screwdriver after each adjustment cycle in the order A, B then C; this constrains the most degrees of freedom soonest. Remember to tighten all Allen-bolt locks again after final adjustment of the capstans.

1This is done by inserting the hex wrench (located in the box) in the socket in the end of the screwdriver's hand grip, and rotating the wrench to position the cursor on the vernier scale at 3Nm. If in doubt seek help.

Cryostat and detector geometry

When the telescope is parked at zenith the dispersion direction in each arm on the detectors is horizontal, and the spatial direction lies in a vertical plane. In this telescope orientation the A, B and C capstans are located respectively at ten-thirty, six and one-thirty by analogy with a clock dial, when viewing the cryostat from a vertical stance.

This understanding of the cryostat-CCD geometry aids in defining changes to coefficients in the rotation and tilt setup scripts when a new detector is being commissioned. For example, when commissioning the red-arm backup detector, Red+2, laboratory images taken with it were flipped along its short axis compared to images taken with the Red+ detector. It follows trivially from this that the sign of the rotation coefficient is different for the two detectors. Furthermore, since the short axes of the detectors correspond to the spatial direction, and adjustments of the B capstan will tilt the cryostat in the spatial direction only, the sign of the coefficient of the B capstan in the tilt setup script must also be different between these two detectors.

Cryostat alignment - rotation, tilt, and focus

The cryostat must be aligned so that
  • The optimal focus is within the anastigmatic range (±2500μ from the autocollimation position) of fine focus adjustment by the collimator
  • The detector is aligned with either the slit (spatial) or dispersion axes. Normally the rows of the CCD are aligned with the spatial direction, i.e. arc lines, and therefore sky lines, are aligned along the detector rows
  • The detector is not tilted with respect to the focal plane, i.e. the focus is uniform over the full illuminated extent of the detector
  • Optimal spectral resolution should not be degraded by more than "a few percent" by any residual misalignment
It's important to check cryostat alignment with the dichroic to be used on-sky; different dichroics will introduce specific image rotations, especially in the blue arm. Similarly, the grating should be set to the central wavelength to be used on-sky. Inserting a filter will introduce a focus offset that must be accounted for if done post-setup.

Alignment of rotation is effected using arc exposures (to align arc lines with the CCD rows), and alignment of detector tilt with respect to the focal plane and focus on the detector is carried out using a Hartmann test. IRAF scripts are provided which analyse arcs and Hartmann pairs of arcs, and advise quantitatively on specific adjustments of the rotation micrometer, capstans and collimator focus to determine the optimum cryostat alignment.


To determine the rotational alignment of the cryostat an arc exposure is taken and the IRAF script,, extracts arc spectra in two windows widely separated spatially on the CCD and cross-correlates them to determine any relative pixel shift which would be indicative of rotation. This shift is converted into a detector rotation offset angle, and the corresponding micrometer offset in mm is calculated. A typical output of the rotation script is:

pixel shifts between top and bottom   :       0.44 
corresponding angle in degrees        :       0.0267
corresponding micrometer offset in mm :      -0.088 (anti clockwise rotation)

The pixel shift between top and bottom refers to the spatial direction (not up/down on the image display), and being measured directly on the arc image, is the fundamental metric of rotation error. The rotation and micrometer offsets involve conversion scale factors and can on occasion be in error, e.g. if the cryostat is rotated significantly with respect to its nominal position. For a precisely-aligned cryostat you should aim to have the pixel shifts between top and bottom <0.5 pixel, preferably <0.25 pixel. This corresponds to a rotation angle of about one minute of arc over the full spatial window. Note that arc lines are curved by up to 2 pixels centre-to-edge of the standard window for the high-resolution gratings.

The spectral and spatial directions are not orthogonal in either arm of ISIS; rotation of the cryostat to align the spatial direction with the detector rows, which of course simplifies sky subtraction and is the default setup, means that the spectral direction is then not precisely aligned with the detector columns. This misalignment is ∼0.3-0.4 degree for all gratings with the exceptions of R158B (∼0.2 degree) and R158R and R316R (∼0.5 degree).

Good alignment of the dispersion direction with the detector columns helps trace the spectra of very faint point targets without significant degradation of signal-to-noise by the extraction process. Therefore observers may on rare occasions request instead to have the spectral direction well aligned with the detector columns. This is achieved by taking a continuum lamp spectrum with the narrow 0.3" dekker deployed (standard dekker slide, position 2), and running the rotation script in mode 2 (tungsten lamp).

However precise alignment of the spectrum trace with detector columns won't be maintained on-sky because ISIS doesn't have an ADC, and so atmospheric differential refraction will conspire to misalign the spectrum trace as a function of observed zenith distance, spectral coverage and slit orientation with respect to vertical. For example, at zenith distance 50-degrees light of wavelength 3500Å is dispersed by ∼1.4" in the vertical direction with respect to light of wavelength 5000Å. This is equivalent to 7 EEV12 pixels in the spatial direction of the detector when the slit's oriented vertically.

Tilt and coarse focus - Hartmann test

In a Hartmann test masks block off alternate halves of the pupil and two arc lamp exposures are taken, one with the left Hartmann shutter closed (image S1), the other with the right Hartmann shutter closed (image S2). The order in which this is done isn't important. If the spectrograph is perfectly focussed on the detector, the image of an arc line remains static when either Hartmann mask is deployed. However, if the instrument is out of focus, the image of an arc line will shift on the detector (Figure 1), and the relative shift introduced by the Hartmann masks is a measure of the level of defocus.

The IRAF tilt script,, then takes these two images as input, along with the positions of three arc lines uniformly distributed along the dispersion axis of the detector. This script extracts spectra in nine small windows separated both spectrally and spatially within each image, cross-correlates these spectra between the two Hartmann exposures, and returns the pixel shifts (S1-S2) for each of the nine positions across the illuminated region of the detector. These shifts are then used to compute offsets for the capstan positions by which the cryostat tilt or coarse focus can be adjusted, and the script provides "expert" advice on which adjustment to make. A typical output of the tilt script is:

collecting relevant data from keywords...
extracting spectra in nine sub-windows...
subtracting median...
correlating spectra...

sub-window centre x,y coordinates are :
(990,3298) - (2067,3298) - (3342,3298)
(990,2023) - (2067,2023) - (3342,2023)
(990,946)  - (2067,946)  - (3342,946)

Hartmann shifts are in pixels :
      0.42         0.30         0.07
      0.36         0.40         0.23
      0.29         0.40         0.25

NOTE: This chip reads out with dispersion along the Y-axis.
The top-bottom shifts reported still refer to the spatial direction,
and the left-right shifts to the spectral direction, i.e. they refer
to the real orientation, NOT how it appears on your image display.

average    Hartmann shift :       0.30 pixels
top-bottom Hartmann shift :       0.17 pixels (    0.0573 deg)
left-right Hartmann shift :      -0.05 pixels (   -0.0068 deg)

to correct for measured Hartmann shifts turn capstans as follows
capstan offsets, positive is clockwise, in units of full turns :

overall     A: -0.05   B: -0.05   C: -0.05
top-bottom  A:  0.00   B: -0.29   C:  0.00
left-right  A: -0.03   B:  0.00   C:  0.03

EXPERT ADVICE: correct for overall Hartmann shift

Note: Cosmic rays or faint arc lines may induce inconsistent results.
      Tighten capstans in the following order: A, B, C
      Changing the capstans will introduce a focus shift.

      Report problems to instrument specialist.

Again, the pixel shifts are measured on the Hartmann arc pair and are the fundamental metrics of tilt and focus errors; the capstan offsets are computed using a scale factor conversion and, as in the case of rotation, may be misleading on occasion. The average Hartmann shift is the average of the nine individual shifts and is a measure of overall focus, and the top-bottom and left-right Hartman shifts are the differences in the 3 Hartmann shifts measured at the spatial and spectral extremes of the detector, and therefore are measures of detector tilt in those axes. If some of the measured pixel shifts are "absurdly" large, you've likely selected a low SNR or even a saturated spectral line, or possibly a doublet (the script will issue a warning).

Note that with the telescope parked at zenith the dispersion direction in each arm is horizontal. As mentioned in the output of the tilt script, the "top-bottom" pixel shifts refer to the spatial direction (i.e. "up-down" at the cryostat, not on the image display). For point targets, tilt in the dispersion direction ("left-right") is generally more important than in the spatial direction.

Fine-tuning the spectrograph focus

When the measured tilt and overall Hartmann shifts are <0.25 pixel or so, the IRAF focus script, isis_focus with the last Hartmann pair of arcs as input, can be used to calculate adjustments to the collimator to fine-tune the spectrograph focus.

An offset of the collimator focus should not take it outside of its anastigmatic range, otherwise de-collimation of the beam incident on the grating would cause the best focus on spectral lines to degrade the spatial resolution (and signal-to-noise). If the adjusted collimator focus would be astigmatic, then you must instead move the three capstans (each in the same direction as indicated by the tilt script, possibly by more than one turn). Take a new set of Hartmann exposures and run the tilt script again. Check the pixel shifts for the nine windows and compare with the previous values; you can easily see if the recommended capstan offsets were in the right sense.

A typical output of the focus script is:

collecting relevant data from keywords...
subtracting median...
correlating spectra...

Hartmann shifts derived in units of pixels (top, center, bottom) :
      0.37 ,         0.39 ,         0.23
average Hartmann shift :       0.33
corresponding collimator offsets :
-458 ,   -486 ,   -293
recommended collimator offset (average) : -412

Note: Cosmic rays or faint arc lines may induce inconsistent results.
      The final collimator position has to be within the
      appropriate range, if not, change capstans !

      Report problems to instrument specialist.

To adjust the collimator according to the recommendation of the focus script, the recommended offset is applied to the current focus setting. Applying offsets of ∼200 microns or less brings a negligible improvement in spectral resolution.


To align and focus the cryostats, the following procedure is recommended.

  • first, ensure the collimators are near the middle of their anastigmatic ranges, taking into account offsets due to optical elements in the light path.
  • second, correct for the overall Hartmann shift, i.e. ensure the spectrograph focus is reasonably good. This may only require a collimator focus adjustment, so check the focus with the IRAF focus script; the collimator must remain anastigmatic throughout. Detector tilt should be broadly OK too.
  • third, correct for cryostat rotation. For a precisely-aligned cryostat aim to have the pixel shifts between top and bottom <0.5 pixel, preferably <0.25 pixel, which corresponds to a rotation of ∼1 arc-minute over the full spatial window.
  • fourth, iteratively correct for left-right or top-bottom tilt, or overall (coarse focus) shifts, following the "expert" advice from the script. In many instances collimator focus offsets (up to several 100 microns corresponding to up to ∼1 pixel of Hartmann shift) can be applied in lieu of overall adjustment of the capstans. After adjusting the capstan(s) or collimator take another pair of Hartmann arcs, and check the nine pixel shifts to be sure the adjustments are converging, and repeat this sequence until the tilt Hartmann shifts are <0.25 pixel or so.
  • fifth, when the top-bottom and left-right pixel shifts are within tolerance, fine tune the focus by adjusting the collimator.
  • sixth, check that the tilt corrections haven't induced a change in rotation. If they have, correct the rotation and then re-check the tilts and focus, ensuring the collimator remains anastigmatic.
  • finally, perform a "sanity check" on the setup to ensure there are no remaining configuration issues. Take an arc, and for three lines well separated in wavelength, measure their centroids and FWHMs at three positions separated spatiallty (e.g. at x∼w/4, ∼w/2 and ∼3w/4, where w is the spatial extent of the detector window). The FWHMs should be commensurate with the projected slit size and should not vary by more that about ∼0.1 pixel, and the centroids measured for a given line should not vary by more than ∼1 pixel (bear in mind arc lines are curved by up to 2 pixels centre-to-edge of the standard window for the high-resolution gratings). Ensure that the central wavelength and spectral range of this arc conform to the grating setting by comparing it with an arc-line atlas.
Of course ideally we'd have (S1-S2)=0 over the entire detector, but in practice some small Hartmann pixel shifts remain. It was asserted above that shifts of <0.25 pixel are desirable; this tolerance can be justified in terms of the induced degradation of optimum spectral resolution.

The impact of a given Hartmann shift in pixels on spectral resolution depends on pixel size and on the projected slit size on the detector. The top panel of Figure 2 shows the measured spectral resolution degradation for a 2-pixel slit of a TEK detector, with 24 micron pixels, as a function of average Hartmann shift (S1-S2). The degradation in resolution of course increases rapidly as the Hartmann shift increases, but is still only 5% when the Hartmann shift (S1-S2)=0.5 TEK pixel.

The pixel size of the EEV12 detector on the blue arm is 13.5 microns, and a slit width of approximately 1 arcsec projects to ~4 pixels FWHM on the detector. In this case, measurements show that average Hartmann pixel shifts of 1, 0.5 and 0.25 EEV12 pixel degrade the 4-pixel resolution by 5%, 2% and <1% respectively (Figure 2, bottom panel). Resolution degradation by a 0.25-pixel average Hartmann shift therefore is comparable to the resolution change induced by the mechanical setting accuracy of the slit width (∼0.02"). Also, experience shows that tightening the capstan clamps by-hand can induce Hartmann shifts of up 0.15 EEV12 pixel. Resolution degradation at levels corresponding to Hartmann shifts <0.25 pixel is therefore affected, and likely dominated, by mechanical "noise".

The recommendation for the EEV12 and RedPlus (15 micron pixels) detectors to adjust tilt and fine tune the spectrograph focus with the collimator until the average, top-bottom and left-right Hartmann pixel shifts are each as low as 0.25 pixel is fully commensurate with the requirement of negligible degradation of optimum spectral resolution.

The acceptable tolerances for rotation, tilt and focus will depend on the nature of the science programme being carried out; those given above are for the best setup beyond which there is no significant advantage. But, for example, on service nights compromises may have to be made due to time constraints. Potential concessions should be judged against the science programme, e.g. if the full spatial extent of the window will not be used the rotation tolerance can be relaxed, and for a point target good spectral tilt is more critical than good spatial tilt. Nonetheless one should always aim for Hartmann shifts and rotation to be less that ∼1 pixel.

Spatial focus

At this stage the slit is focused on and aligned with the detectors in each arm. The final step in setting up is to focus the telescope on-sky, fine-tuning the longitudinal position of the secondary mirror so that stellar images are focused on the slit. This is done by acquiring a star with V- or r-magnitude approximately 9-11 on the slit and taking a series of integrations over a range of focus values with the focusrun script. It's preferable to do this in the red arm and at low air mass where the seeing is better; this allows greater sensitivity to the effect of changes in focus on the spatial profile so that the optimum focus is better determined. Integrations should be of 10-15 seconds to ensure that fluctuations in the seeing are averaged out, i.e. its lower spatial frequencies are well sampled.

If the focus determined in this way (in mm) is Fmeasured then the focus applied to the telescope is

Fapplied = Fmeasured + ΔF(temperature) + ΔF(elevation) + ΔF(filter) mm

where the ΔF terms are corrections for telescope truss temperature, telescope elevation and filter focus offset. The temperature and elevation corrections are applied automatically by the TCS throughout the night to ensure the telescope remains focused on the slit as its environment changes. Example values (in mm, taken from the FITS headers of image r2345903 on 20160412) are

Fmeasured = 97.85
ΔF(temperature) = 0.099
ΔF(elevation) = 0.055
ΔF(filter) = 0.00

The measured focus for ISIS is typically 97.85±0.05 mm, and the filter focus offset, denoted by dF in the TCS display, should be zero (but, see cautionary note below). Note that the FWHM of the spatial profile at optimum telescope focus is greater in the blue arm than in the red arm due to (i) the λ-0.2 dependence of seeing on wavelength, λ, and (ii) broadening by charge diffusion in the blue arm. This difference can exceed 0.3-arcsec at widely-spaced wavelengths.

It's instructive to quantify the "depth of focus", that is, the telescope focus error that can be tolerated. A change of δF mm in the position of the secondary moves the Cassegrain focal plane by an amount that depends on the curvature of the secondary and its distance from the focal plane. ING TN No. 9 reports that the ratio of focal plane movement to secondary mirror movement is 20, that is, a movement of the secondary by δF mm moves the focal plane by 20×δF mm. This translation of the focal plane causes a spatial point spread (defocus) of 20×δF/f# mm, where f#=10.94 is the focal ratio of the beam incident on the slit. The plate scale at the slit is 206265/(D×f#), where D=4200 mm is the WHT's diameter, and therefore the spatial point spread at the slit, δθ (in arcsec), caused by the focus error δF mm is

δθ = 20×206265 ×δF/(D×f#2)

that is,

δθ (arcsec) ∼ 8×δF (mm)

Therefore, for example, a focus error of 0.10 mm causes a defocus of 0.8 arcsec, and this level of focus error would dominate the spatial profile for roughly half the time assuming median image quality (seeing and intrinsic PSF) of ∼0.8  arcsec, and ignoring the wavelength and zenith-distance dependencies of seeing. On the other hand, a focus error of 0.05 mm, corresponding to a spatial point spread of 0.4 arcsec, would dominate the spatial profile for only about 15% of the time, given the distribution of site seeing. In practice the focus can be measured to ∼±0.025 mm in median seeing and better.

Cautionary Note: When ACAM is in use a focus offset for the optical element deployed is applied automatically by the TCS to the secondary mirror. This means ACAM observers can change its configuration without the need to modify manually the telescope focus to compensate for the physical thickness of the filter or grism in use.

Under some circumstances this focus offset can be either inherited or applied to the telescope when ISIS is being used. Depending on the optical element, the focus offset wrongly applied can cause the telescope to be very significantly defocused; for example, the focus offset for the frequently used ACAM grism is -0.16 mm, equivalent to a defocus when using ISIS of 1.3 arcsec. A focus error of this magnitude would increase slit losses with a 1 arcsec slit from about 15% to some 40% in median seeing (0.8 arcsec), and would dominate the spatial profile for more than ∼85% of the time, given the distribution of site seeing.

The circumstances under which a focus offset is wrongly applied by the TCS are:

(i) During an ISIS run a daytime reconfiguration of ACAM such as deploying the grism will cause the corresponding focus offset to be inherited when ISIS observations resume on-sky. If a focus run is made the focus offset wrongly applied will be compensated for in the measured focus, Fmeasured. But, if a focus run is not carried out the telescope will remain defocused by an amount commensurate with the focus offset.

(ii) If during ISIS observations ACAM is reconfigured in anticipation of later observations, the focus offset of the optic configured will be applied immediately to the telescope, and the ISIS observations instantly defocused.

To mitigate against this anomalous behaviour the ISIS observer should (i) ensure at the beginning of the night that the dF term in the TCS display is not set to a non-zero number, and if it is should request the OSA to remove it; (ii) never reconfigure ACAM during ISIS observing, even if it's planned to use ACAM later in the night.

Finally, if observations are made with ACAM and the TCS subsequently switched to ISIS mode, any ACAM focus offset applied will be (correctly) removed. However, if subsequently ACAM is reselected in the TCS to be used in the same configuration the corresponding focus offset is not automatically reapplied. To reapply it either the action (gear-wheels) icon in the ACAM GUI for the corresponding filter wheel must be activated, or the optic redeployed explicitly from the ICS command line. If this isn't done and an ACAM focus run is not made then subsequent ACAM observations will be defocused.

Top | Back

Contact:  (ISIS Instrument Specialist)
Last modified: 06 July 2020