MAESTRO: MMT Advanced Echelle Spectrograph

 

Table of Contents

    I. Design Philosophy
 
   II. Optical Predesign
       (a) Heritage
       (b) Assumptions
       (c) Spectral Resolution
       (d) Projected Slit Width
       (e)  Length of Orders
       (f) Echelle Spectrum Format

  III. CCD and Dewar

 
IV.   How blue can we go?

  IV.  Image Quality Requirements

   V. Operations
           (a)  Standard observing
           (b)  Apertures
           (c)  High signal-to-noise
           (d) Calibration lamps
           (e)  Long slit
           (f)  Un-cross-dispersed
           (g)  High stability
           (h) Summary of Operations Requirements

 VI. Summary of System Requirements

 
 VII. Optical Design
        (a) Design Approach
        (b) Design Description
        (c)  Focal Reducer
        (d)  Design Performance
        (e) Bulk Transmission
        (f) Grating
        (g) Anti-Reflection Coatings
           (h) Tolerancing
         (i) Alignment Strategy
         (j) Fabrication
         (h)  Outstanding Issues and Future Work
 
IX. Mechanical Design
      (a) Flexure Requirements
      (b) Thermal Requirements
      (c)  Structural Design
 
X. Computers

XI.  Estimated Cost

 

Appendix I:  Drawings of individual optical elements

Appendix II.   Echelle Formats
    (a) Single UBK7 Prism at MMT
    (b) Single UBK7 Prism at LBT
    (c) Double UBK7 Prism at MMT
    (d) Double Red Prism at MMT
 
 
 
 
 
 

I.  Design philosophy

The following items describe the rationales for the design choices.
 
1. Spectrograph will be at the f/9 Cassegrain station of the MMT.
The earlier design located the spectrograph at a Naysmyth platform, using the f/15 secondary. The Naysmyth platform would have to be built, and the optics which directs the beam to the Naysmyth would have to be fabricated. The acquisition/guider camera would have to be made, as well as calibration lamp functions. By placing the spectrograph at f/9 Cassegrain, we can use the existing secondary mirror, "Top Box" and rotator. The target acquisition and guiding scheme will be the same as that used with the MMT spectrographs). Foltz said that this is a preferable solution from the standpoint of operations, despite the fact that the spectrograph would be removed from the telescope when not in use. One potential disadvantage of the Cass location is that the physical size of the spectrograph is limited; however the adopted Litrow design is compact and well within the weight and size limits.  A description of the MMT optics is given by Fabricant, McLeod and West.
2. Relocation to LBT some day.
The plan is to reloctate MAESTRO to LBT at some point. The other planned echelles for the LBT are optimized to be fed by AO (e.g. AVES), and are near-IR or at least red-optimized.  Details of the mechanical coupling to the telescope, guiding, calibration etc would be worked out later, and would depend on what has been provided by MODS.
 
3. Prism versus grating cross-disperser.
The advantage of a prism cross-disperser is that you can have single exposure coverage which covers more than a factor of 2 in wavelength. For a grating cross-disperser, you are limited to a factor of 2 in wavelength coverage or else you suffer order overlap. By having a very large single-exposure spectral coverage, you can have a point-and-shoot instrument with few moving parts, a very important advantage. A prism is superior in throughput, and spaces the orders more compactly on the detector than a grating.
 
The disadvantage of a prism is that order separation in the red is poor, just where the sky is bright and sky subtraction is difficult.  We therefore will fabricate two sets of prisms, which will be interchangeable.  One set will be optimized for the blue, and be transmissive to the atmospheric cut-off.  The other will be made of a higher index flint and therefore have better order separation in the red, but not be blue transmissive.
4. One-channel vs Two-channel.
A two channel system offers many advantages, including optimization of the optical design, materials and coatings, over a shorter wavelength range. However, the added complexity of having to fabricate channels, and the unattractive dichroic in the middle of the most sensitive part of the optical spectral range, made us chose a one-channel system with a prism cross-disperser.
5. Wavelength Coverage.
There are two science applications we foresee with competing requirements, one (quasars) which requires coverage to the atmospheric cut-off at 320 nm, and the other (stars) which typically requires spectra to the silicon cut-off out at 0.9-1.0 microns. We chose to optimize for the blue, but do not want to preclude the red. Future spectrographs at the LBT will be optimized for use with AO, and should be designed to work to the short wavelength cut-off of InSb. For InSb and red-optimized spectra, the observing sequence is sufficiently different from the optical/CCD observing sequence, that it is not clear that a spectrograph with both near-IR and optical capabilities is desirable.  The wavelength coverage is discussed further below.

Top

II. Optical Predesign

This section describes the optical pre-design of MAESTRO. Equations and notation are taken from Schroeder's Astronomical Optics (Chapters 13-15).
 

(a) Heritage:

The design is a double pass, quasi-Littrow with transmissive camera/collimator.  It is similar to the LCO Dupont echelle spectrograph , Jim McCarthy's echelle at McDonald, and Shectman's echelle for Magellan. The main differences with these spectrographs are
   (1) the availability of a large format detector
   (2) the choice not to have two channels
   (3) the desire to work shortward of 400 nm.
   (4) the desire to have greater order separation, particularly in the red.
 

(b) Assumptions:
 

                                                              or the LBT: D(LBT)= 8.408m, f/15 secondary.
            In a second set of designs being carried out by C. Harmer, we are having 2 sets of 2 prisms each.
           One set will be blue transmissive -- two 35 degree prisms.
           One set will be "flintier" -- i.e. made out of a material with a higher index of refraction than
                UBK7.  This would allow better order separation in the red, at the expense of losing the
                 blue throughput.
 

( c) Spectral Resolution

The spectral resolution for a one arcsecond slit is given by R , where


here R = spectral resolution = lambda / 
= the blaze angle of the echelle grating
= the angle of the incident beam onto the grating with respect to the blaze
= + 
       D = the diameter of the telescope
       d1= the size of the collimated beam at the echelle grating.

Given the beam size of 200 mm and the MMT or LBT telescope primary, and assuming a Littrow configuration, the spectral resolution depends on the blaze angle only.

For the MMT,
     R(arcsec) = 12,683 tan
and for the LBT,
    R(arcsec) = 9812 tan

We'd like R=30,000 (10 km/sec) at the MMT with a 1 arcsec slit, so =67 degrees. With this grating at the LBT, the 1 arcsec resolution is 13 km/sec, with a 0.7 arcsec slit 9 km/sec. The maximum resolution with 2 pixel sampling is 3 km/sec, which would require a 0.3 arcsec slit at the MMT or a 0.25 arcsec slit at the LBT.
 
 

( d )  Projected Slit Width


The projected slit width at the detector,  , in microns is

where
r = the anamorphic demagnification
F2 = the camera focal ratio.

In order to get a reasonable plate scale given our 15 micron pixels, we take F2 = 3. In other words, a lens will be needed to take the f/9 or f/15 telescope beam and make it f/3. We assume r=1.0.

Then the scale is 0.154 arcsec/pixel at the MMT and 0.119 arcsec/pixel at the LBT.  In order to get a plate scale better matched to our detector, we'd need a faster camera, which would be hard to fabricate.
 

(e)  Length of the Orders


The groove spacing of the echelle grating determines the length of each order's free spectral range, L.

See Schroeder's Fig. 13.12.

where    f 2    =   F2  x  200 mm = 600 mm
=   the distance between grooves
   =   the central wavelength of the order.
=   blaze angle of the echelle grating.
 

(f )  Spacing between Orders.


Let Ax be the angular dispersion of the cross-disperser, in arcsec/length, then

   =    2    f2     Ax 

where     =   the distance between orders in length units

                            f2 = 600 mm

= the free spectral range

and the factor of 2 comes from the fact that we use the prism in double pass.

For a prism,

where         t = base length of prism
                    a = height of the prism

and

and

for UBK7.     For a 50 degree apex angle, t/a=0.933.
 

(f)   Echelle Format

The cross-dispersed formats were computed for the echelle masters available at Richardson Labs,
a list of which is available here, and other groove spacings which could be ruled.   The trade is to get as much of the spectrum on the 4kx4k CCD while having an acceptable slit length.

We settled on a grating blazed at 63 degrees with 79 grooves/mm.  The results are summarized in the table below and given in detail in Appendix 2.

In all cases we have continuous spectral coverage blueward of about 650nm, and gaps redward of that.

The  optical design described below was carried out by Sarlot, and  assumed a single grating of UBK7
and a 50 degree apex angle.  These give acceptable formats, but with pretty small slits -- we'd have to count on crash hot image quality.

We then thought about 2 prism systems, with two 35 degree apex angle prisms.  These give
better order separation.

The problem is that UV transmissive glasses like UBK7 have low indices of refraction.

Another possibility is to have two sets of prisms:
   - one set is as blue transmissive as possible
   - higher index glass to be used only at wavelengths longward of 400 nm

Once we chose to shorten the single exposure  spectral coverage one has to ask if we are not better off with a grating cross-disperser and a more traditional design.   The two-prism options cover a little more than a factor of 2 in wavelength, but not much -- first order grating cross-dispersers are limited to a factor of 2 in wavelength. However, a grating cross-disperser has less uniform spacing of orders, so we would not be able to fit the full factor of 2 wavelength coverage on a 4k ccd.  Also, even with effectively 8 surfaces, the throughput of the prism cross-disperser is better than a grating.

Thus, we like the 2 prism designs, which will allow red spectroscopy.  A set of red prisms could be made later if they are not within the current budget.
 
 

Cross disperser Telescope Single Exposure Total Wavelength Coverage Slit Length
Single UBK7 50 deg MMT 310 - 1000 nm 7 arcsec
Single UBK7 50 deg LBT 310 - 1000 nm 6 arcsec
Two 35 deg prisms UBK7  MMT 310-700 nm 10 arcsec
Two 35 deg prisms Flint MMT 400-900 nm 12 arcsec

 Top

III. CCD and detector package

Mike Lesser and the University of Arizona Imaging Technology Lab will fabricate the CCD, dewar, and detector electronics.   The detector will be a 4096x4096 pixel CCD with 15 micron pixels.
 
Most observations will be read-noise limited.  The gains in sensitivity as a function of read-noise given the sky at Mt. Hopkins were quantified in an earlier memo.  The run of devices that has been ordered has promised 2 electron read noise, which would be better than what has been achieved elsewhere.

The dewar will be a standard LNe dewar  from Infrared Labs, CD-ND-14.

 
Good cosmetics are also very important since we will be using the entire chip, and with a largely point and shoot design, we do not have  the option to move the spectrum around much to avoid bad traps, rows and other blemishes.
 
A desirable feature is to have a fast read-out mode for calibration frames and standards where superb readnoise is not important.
 
Top

 

IV.  How  far into the UV can we go?

For many science applications, we want to work as far into the UV as possible.  We also want to take advantage of the superb response of the CCD Lab's UV optimized devices.   However,  optical quality gets harder to achieve in a  non-linear way as you require good imaging at shorter and shorter wavelengths.

Here we show the QE curve of the CCD and the atmospheric transmission in the blue, in order to see where
to set the blue limit. We set 330 nm as the bluest wavelength we optimize in the optical design, although
we allow work to 310 nm.

The quantum efficiency (QE) curve for the MAESTRO CCD will be simillar to CCD24 at the 90inch, shown here.

The QE is flat to 0.8500 microns in the red where it starts to fall rapidly.

Although the CCD has high QE all the way down to 0.3200 microns, atmospheric transmission is poor below 0.3400 microns. Some optical image degradation by the spectrograph is tolerable in the blue -- the telescope itself will probably provide further degradation.

Atmospheric Transmission as a function of wavelength in the UV, from Kitt Peak (solid line, from IRAF database) and Mauna Kea (dashed-dot, from Beland, Boulade & Davidge 1988 CFHT Info. Bull. 19, 16).  Transmission at the MMT will probably be between these two curves.
 

 Top

IV. Image Quality Requirements

The median seeing at the old MMT was 0.8 arcsec FWHM.   The distribution of seeing measurements from the seeing monitor is given below.  The seeing monitor measured the seeing in the optical, at a fairly red wavelength, effectively about 700 nm.

Most of the seeing measurements were better than 1 arcsec, and they never measured seeing better that 0.4 arcsec. At the blue end of MAESTRO, we expect the image quality delivered at the slit to be somewhat worse than the seeing measured with the seeing monitor.

The median seeing of 0.8 arcsec FWHM corresponds to a gaussian with

= 0.8  arcsec  /  2.354  =  0.34 arcsec.

Most of the light is within +/2, or 1.36 arcsec.

The range of the seeing, 0.4 - 1.0 arcsec FWHM, corresponds to = 0.17 - 0.425 arcsec, and +/2=0.7-1.7 arcsec, which 4.7-11 pixels at the MMT. Thus the spectrum will be oversampled with 15 micron pixels.
 

The median seeing FWZM of 1.36 arcsec corresponds to 9 pixels at the MMT, and so we'd always bin the CCD down to get 2-3 pixels per resolution element. If we have optics which can deliver images that are good enough to allow the observer to have 2.5 pixel sampling with the 15micron pixels and no binning (i.e. have the maximum spectral resolution), then the full 2gaussian corresponds to 0.385 arcsec or a FWHM of 0.227 arcsec.

Top

V.  Operations

(a ) Standard Observing

The most common observation will be of a point source, with wide spectral coverage.  The image de-rotator
will be on in order to keep the slit aligned with the parallactic angle.

Guiding and acquisition will be accomplished with the existing Topbox, which will view the slit.
The slit is tilted 12.5 degrees with respect to the telescope axis, and the focal plane is 9.0 inches behind the top box mounting plate.

The current guider has two magnification lenses.

(b)  Apertures

 The slit will be machined into an aperture plate, which is then polished and aluminized.   The aperture plate will be mounted in a holder which is removable by the observer with the telescope at stow.

  The choice of aperture plates will include:

     double small holes for testing and focus
     0.7 arcsec x 10 arcsec for standard blue observing
     0.7 arcsec x 12 arcsec for standard red observing
     1.0 arcsec x 10 arcsec for blue observing during poor weather
     1.0 arcsec x 12 arcsec for red observing
     0.3 arcsec x 10 arcsec for high spectral resolution
      collection of 1 arcmin slits for long slit mode
 

(c)  High Signal to Noise

 For some applications, photon counting statistics will not limit the signal-to-noise, but rather one is limited by the ability to beat down the systematics.   For all applications, it is necessary to move the spectrum around on the detector by a few pixels. One way is to move the star on the slit, but that is hard with such a small slit, and you want to move the the spectrum in the dispersion direction as well. If you want to detect a weak line in a certain part of the spectrum, you will want to move that bit of the spectrum around on the detector in order to feel secure about a line detection.
 

Small dither motions of the spectrum on the detector, 15-50 microns, will be done by moving the echelle grating.


 

(d)  Calibration Lamps

It will be necessary to have
   (1) a continuum lamp which goes through the slit
   (2) a line lamp such as thorium argon which goes through the slit and
   (3) ideally, but not absolutely necessarily, a continuum lamp which doesn't go through the slit.
 
The existing continuum lamp in the top box may be too faint for the echelle, and the perennial problem with quartz lamps is that they are so red -- which can be compensated for a little bit with a filter.
In the old top box, hallow cathode lamps like thorium argon required for the echelle were too faint to be fed like the He/Ne/Ar lamps in the top box used by the red and blue channel. So it may be that we need some way to feed the hallow cathode calibration lamp directly to the slit.
In Lesser's lab, there is a diffuse source made of fused optical fibers which would be ideal to illuminate
the CCD for pixel-to-pixel flat-fielding.  It could be inserted in front of the echelle grating mechanically for exposures taken in the afternoon.

(e)  Long Slit

One science application that is of high interest is rotation curves in galaxies, either faint, distant ones (so slit lengths of 20-30 arcsec are adequate) or nearby ones (so you would want > arcminute slits).  Multiobject spectrographs that do this like ESI at Keck would beat us on faint galaxies, but we will have superior spectral resolution and potentially throughput.  What is required is to have the ability to insert an order-isolating filter and the ability to have a stable psf and lsf over a long slit. The lsf/psf can change, we just have to have some way of knowing what they are.
 
This is not a requirement to drive the optical design, but we need to know what the imaging properties are for a single order over a long slit.  An automated mechanism for the filter insertion is not necessary, nor the automated insertion of the filter. An observer would decide to do this and could insert the slit and filter during the day for the night's observing, or in the middle of the night with the telescope at stow.

(f)  Un-cross dispersed mode

Another potential option is to remove the cross-disperser and use the spectrum with completely overlapping orders. Ed Olszewski and Steve Vogt showed with HIRES that they could get really good velocities on faint objects by doing this, to beat read-noise. They put a filter in to get the "11 orders" within which all the velocity info is contained for red giants -- approximately 420-520 nm. They also made slit masks to do multiobjects (on a scale of 10x30 arcsec or so.) This is another mode where MAESTRO will beat Hectoshell with its fibers.

(e)  High stability observations

For applications like asteroseismology, planet searches, and very high s/n spectroscopy, a very stable system is required. With such a wide wavelength coverage MAESTRO will beat Hectoshell for these applications, even if Hectoshell is upgraded later on to have a cross-dispersed mode. So a future upgrade to MAESTRO to consider is to sit it  on the floor in a thermally controlled box, fed by a fiber, possibly with a iodide cell.  This is low priority now, and won't be implemented until demand warrants.  There are no obvious show stoppers to implementing this in the future.


(f)  Summary of Operational Requirements
 

               aperture plate
               filter holder (with space for one filter)
               prism: red, blue, none
               insertion of flat field optical fiber paddle
                 spectrograph shutter

Top

VI. Summary of System Specifications

  • Focal ratio at detector: f/3 (for 6 pixels/arcsec or 0.154arcsec/pixel)
  • Pupil diameter at grating: 200mm
  • Detector size: 4k x 4k with 15um pixels
  • Interchangeable between MMT and LBT (with interchangeable focal reducers)
  • Ability to remove prism and reposition grating
  • Spectral region: 0.330um to 0.7um (minimum)
  • Spectral region: 0.310um to 0.85um (optimal)
  • Prism: Ohara FSL5Y with 50 degree apex angle
  • Grating: 62 degrees blaze angle, 79 grooves per mm
  • Slit length: 7 arcseconds long on sky
  • Imagery: 0.25 to 0.50 arcseconds rms

    • V.  Optical Design

    (a) Design Approach
    The design approach for Maestro was to create a high resolution cross dispersed Echelle spectrometer over a spectral range from the lower limit of the detector to the near infrared. The preferred field for the instrument was defined as 11.4 arcminute inscribed field.
     
    The main difficulty is to find a satisfactory balance between image performance and throughput. Because of desire to go "as blue as possible" most common materials do not transmit. Trying to balance with materials that transmit and can correct color over a wide spectral range is the central challenge of this design. In addition, because of these problems, a minimum number of lenses is required as well as effective anti-reflection coatings.
    Full analysis of the instrument was carried out in Zemax.
    The design presented is a six element double pass near Littrow collimator and camera combined system. In addition, the design includes a field flattener 5 mm in front of the detector and was originally derived from the Shectman f /5 spectrograph. All lenses utilize the Ohara I-line glasses except the center lens of the triplet which is CaF2.

     

    (b) Design Description

    The design consists of six powered elements and an additional focal changer. The lenses are arranged in three groups with various design functions; the triplet is a color corrected collimator, the negative positive pair is a length reducer and the last two lenses reduce aberrations, in particular, field curvature. The overall layout of the design is shown below.

    A close-up image of the detector and field flattener is shown below. Note that the small fold flat that reflects the beam after the slit is slightly in front of the detector. However, the shadow of the fold flat is beyond the free spectral range of the data. The image on the right below is viewing the detector from an on-axis ray coming from the grating to the detector. The red color is the full size of the detector, the blue is the shape of the field flattener (it will most likely be cut square later on) and the small ellipse is the fold flat shadow from the mirror after the slit plane.



     
     
     
     
     

    The design corrects color over the spectral region of 0.338um to 0.812um with the model including the telescope, six element lens system and prism with diffraction grating. The focal reducer was modeled with a three lens system before the slit plane, not as the proposed lens/mirror combination. One weakness of this design is that the fold prism reflecting the slit towards the grating is in front of the detector, thus creating a shadow on the detector and losing light. This can be moved either closer to the detector so that the shadow is minimized or angled completely outside of the rays towards the detector which will reduce image quality.

    Another detail that must be finalized is that the bottom of the grating is slightly touching the prism corner. Moving the grating further away from the prism deteriorates image quality. Tradeoffs for this may be to allow minimal vignetting at the bottom of the grating and cut the grating shorter.

    All lenses are custom and made from either the Ohara I-line catalog or CaF2. The grating is an off-the-shelf Richardson Grating Lab ruling. The full angle between the incoming and outgoing ray on the grating is 3 degrees or 1.5 degrees off Littrow.


    The focal reducer assembly causes a small shadow, shown below.  The red square is the detector, and
    the green and grey circle is the shadow of the focal reducer/prism assembly.


     
     

    (c) Focal Reducer

    The instrument will be used on the MMT as well as the LBT with focal ratios f/9 and f/15, respectively. Therefore Maestro must include a set of focal reducers to change either incoming beam to an instrument f/3 beam. Originally, a small three lens set was designed for this purpose that was positioned before the slit. However, this presented numerous problems, mostly with the external field viewing camera, and we have now decided to use a combination fold flat lens. At this time, only initial design work has been carried out with this technique but the idea looks very promising. Below is a drawing of the f/9 to f/3 dual purpose fold flat and focal changer immediately after the slit plane.
     


    (d) Design Performance

    The following performance estimates include the nominal system with a telescope but do not simulate the atmosphere nor include manufacturing nor alignment errors. In general, the performance is not yet good enough for manufacturing but we believe the overall design should not be changed. Instead, the design needs to be further optimized or perhaps, an additional lens can be added. The first illustration details the spectra on the detector by order with longest wavelength on the left and the shortest on the right. The second image is a representation of the detector with various positions giving the image quality in arcseconds of the wavelength/order combo of the first image. The detector orientation is the same as the earlier detector illustration. The given specifications were for better than 0.5 arcseconds rms.
     
     
     
     


     
     
     

    Another method of assessing the performance is looking across the orders. The instrument was modeled with various multi-configurations representing the order/wavelength combos from order 28 to 66 with three wavelengths for each order. The following graph shows the three wavelengths for each order modeled (when three points are vertically aligned.) The graph depicts imagery for each wavelength and shows imagery as compared to the specification. In this manner, a reoptimization can be performed with new weighting to improve the balance over the spectral region for more consistent performance.
     
     


     
     

    (e) Bulk Transmission

    The majority of the materials are manufactured by Ohara in their I-line series and all materials are within present volume capabilities of the manufacturer (with an approximate six month lead time.) The material thickness and transmission data are given below for various wavelength and is therefore the minimum transmission for the design at any longer wavelength and field position. The calculations consider only bulk transmission losses and further light loss is expected from Fresnel losses, grating efficiency, focal reducer, the detector efficiency, etc.
     
     

     
     
    Material
    (inc. double pass)
    Total Thickness in mm
    Transmission

    @ 0.33um
    Transmission

    @ 0.34um
    Transmission

    @ 0.50um
    FPL51Y
    22.0
    .92
    .96
    .998
    BSL7Y
    20.0
    .954
    .98
    .998
    Fused Silica
    98.0
    High
    High
    High
    BSL7Y
    20.0
    .954
    .98
    .998
    CaF2
    229.0
    High
    High
    High
    BAL35Y
    20.0
    .94
    .97
    .998
    FSLFY
    260.0
    .88
    .92
    .97
    BAL35Y
    5.0
    .98
    .99
    1.00
    Full transmission through lenses
    68.8%
    81.4
    96.2%

     
     

    (f) Grating

    The echelle grating will be from the Richardson Grating Lab, catalog number 7343LE-401, master BO54.
    Richardson Labs has sent information on the reflectance, based on measurements for a grating made for
    CFHT.  CFHT has not used the grating in an instrument.

    1.  The grating is rectangular, and measures 220x420x74 mm.
          The ruled area is 204x408 mmxmm.
          The groove length is parallel to the 220 dimension.

    2.  The substrate is Zerodur.

    3.  The wavefront specs are lambda/8 rms in the surface, and  lambda/2 P-V, where lambda = 330-900  nm.

    4.  Roland will determine specs for piston, tip-tilt, both for
        initial alignment and for operation.

    5.  Need to have a motorized micrometer to tip-tilt, move spectrum
        5 pixels at detector, or 60 microns, in 1 pixel (15micron) steps.
     
     
     
     
     

    (g) Anti-Reflection Coatings

    Due to  the double pass nature of the design and the wide spectral range,  high performance AR coatings are mandatory to maximize throughput and minimize ghosts.    The simplest AR coating consists of
    a one quarter-wave thick layer of MgF2  which works at best at one wavelength, and is limited to
    about a factor in two in wavelength.  Sol-gel and multi-layer coatings are possible.  Representative
    transmittance curves are shown below, from Joss Hawthorn's discussion of  broad-band AR coatings  for astronomical instruments.

    Plain sol-gel is hydroscopic and not sturdy enough for observatory use.  Hardened sol-gel from
    Cleveland Coatings is a possibility, but expensive:  90prime had quotes recently of ~$10K per element, for similarly sized pieces.

    However, the gain in overall throughput is significant and it is probably worth spending money on coatings.

    ZC&R has economical BBAR multilayer coatings, but such coatings are hard to remove if the coating is botched, which has happened.

    Tom Ingerson advises that sol-gel + MgF2 is the best choice.  He reports measurements of less than 1% reflectance over 330 to 950 nm on witness plates coated with the CTIO ADC optics.  The problem is that the only person who has done this type of coating is Jim Stilburn at the DAO, and he is not in the business of coating optics anymore.  CTIO has talked about constructing facilities to replicate his technique.  We need to seek other sol-gel + MgF2 coating facilities.

    With 30 optical surfaces, the transmittance as a function of coating loss at one surface is given below.
     
     
    Transmittance per surface Throughput with 30 surfaces
    0.95 21%
    0.96 29%
    0.97 40%
    0.98 55%
    0.99 74%
    0.995 86%


     

    (h) Tolerancing

    A tolerancing analysis has not yet been performed, however, it is already predicted that the surface quality and homogeneity of the glasses will be demanding since the full beam for all fields essentially fills the entire diameter of every surface. The negative positive combo lenses in front show strong beam angle deviations. Depending on the results of the analysis, an independent alignment sensor may be necessary to implement.

     

    (i) Alignment Strategy

    A first attempt at tolerancing will be completed to understand the general sensitivities of the elements. At this point in time an alignment strategy will be created with a full tolerancing analysis performed with the particular alignment strategy modeled.

     

    (j) Fabrication

    In Sarlot's design, without the tolerancing analysis, the known complication in fabrication is the center element of the large triplet. The element is a CaF2 substrate with a 3mm edge thickness, 202mm diameter and 104.5mm center thickness.  Although this piece is quite large, procurement and fabrication is possible although non trivial.   Informal discussions with experienced persons have suggested that an element with such thin edges and  thick middle should be avoided however.

    Harmer has replaced the triplet with two triplets with thinner elements, and is in the process of optimizing the design.


     
     
     
     
     
     
     
     
     

    The field flattener will be used as the dewar window and we must guarantee that this element is able to withstand the pressure differential. The aspect ratio is 14:1. The present window on Aries is 15:1 with the same CaF2 material and has not yet been tested under a vacuum. We will know the results of this test within days, nonetheless, many instruments have survived with similar aspect ratios for the material.


     

    (k) Outstanding Issues and Future Work on the optical design

      1.   Scattered light.   Scattered light is the bane of all echelles, and is a limiting factor in making precise measurements on bright stars.

    2.  Tolerancing needs to be carried out.    Sensitivity to alignment and focus need to be quantified.

    3.  Specs for focus mechanism need to be computed.

    4.  Specs for grating motion mechanical device need to be computed.
     
     

    IX.  Mechanical Design

    (a) Flexure Requirements
       The flexure must be held to less than 0.1 of a resolution element over ~30- 50 minutes, since the data are read-noise limited. Most observers will use the spectrograph within 2 hours of zenith since the slit length is short and atmospheric refraction will make getting the whole spectrum down the slit difficult at large airmasses. However, there will always be some important applications at larger airmasses.


    (b) Thermal Requirements

    Diurnal Variations.   The observer will focus the spectrograph during the afternoon to adjust for the relatively small changes in focal length from night-to-night temperature variations.  The telescope enclosure is kept cooler than the ambient daytime outdoor temperature, since diurnal variations are often large, 20-30 degrees F.  The telescope primary thermal control system has been designed to compensate for changes in temperature of 1 degree C per hour (see Fabricant et al for a discussion).  We'd like to be able to operate all night without refocusing the spectrograph -- so ideally the focus should not change with changes in temperature of 5 -10 degrees C.

    Maximum Variations.  The focus mechanism will have to compensate for mechanical expansion and contraction, for the maximum range of temperatures anticipated throughout the year.  The telescope thermal system has an operating range of -25 to + 30 degrees C, so this is the range we adopt for the maximum and minimum temperatures of operation.

    (c)  Structural Design

    The large optical elements and detector dewar will be supported on a space frame of carbon fiber poles.

    Electronics and other non-critical items will be supported on an outer shell of aluminum.

    Preliminary flexure analysis showed that the end-to-end vertical flexure will be approximately 0.0009 inches, or 23 microns,  from zenith to horizon.    Thus, the innovative use of the space frame promises to exceed performance requirements.

     Top
     


     

    X.  Estimated Cost

    Brad Piccirillo (OSI) has given an estimate of $650,000 for all optics except the grating.  The Ohara iline
    glasses are expensive.  Richardson Labs quoted a cost of $51,000 for a replicate of the grating.
     
     

    Top

    Appendix 1.  Drawings of optical elements

    Element drawings are included as a reference for materials, size and shape.
     
     
     
     




     

     Top

     
     

    Appendix 2.  Echelle Formats

    Echelle formats were investigated for the existing echelle gratings in the Richardson Labs Catalog.  We assume the detector is 4kx4k with 15 micron pixels.

    We settled on the 7343LE-401 grating which has a blaze angle of 63 degrees and 79 grooves per mm.

    For each choice of cross-disperser and telescope, we compute
         R  (arcsec)   defined above in section II.c.

         n = spectral order number
         WC(A) = central wavelength of order in Angstroms
          FSR(A) = Free spectral range in Angstroms
          YC(pix) = central pixel number of each order in the cross-dispersion direction (1-4096 to be on CCD)
           DY(arcsec) = order separation in arcseconds
     

    (a)   Single UBK7 50 degree prism at MMT

    Rtheta =     27750.9 (arcsec)
    One arcsec maps to     97.5600 microns, or     6.50400 pixels on detector

    Order,   n     WC(A)     FSR(A)     FSR(pix)     YC(pix)      DY(arcsec):
      1         70.     3218.        45.97           2034.           105.             16.1
      2         69.     3265.  47.31  2063.   208.                                      15.9
      3         68.     3313.  48.72  2094.   310. 15.6
      4         67.     3362.  50.18  2125.   410. 15.4
      5        66.      3413.  51.71  2157.   509. 15.2
      6        65. 3466.  53.32  2190.   606. 14.9
      7  64. 3520.  55.00  2224.   702. 14.7
      8  63. 3576.  56.76  2260.   796. 14.5
      9  62. 3633.  58.60  2296.   888. 14.3
     10  61. 3693.  60.54  2334.   980. 14.0
     11  60. 3754.  62.57  2373.  1069. 13.8
     12  59. 3818.  64.71  2413.  1158. 13.6
     13  58. 3884.  66.96  2455.  1244. 13.3
     14  57. 3952.  69.33  2498.  1330. 13.1
     15  56. 4023.  71.83  2542.  1414. 12.9
     16  55. 4096.  74.47  2588.  1496. 12.6
     17  54. 4172.  77.25  2636.  1577. 12.4
     18  53. 4250.  80.19  2686.  1656. 12.2
     19  52. 4332.  83.31  2738.  1734. 12.0
     20  51. 4417.  86.61  2791.  1810. 11.7
     21  50. 4505.  90.10  2847.  1885. 11.5
     22  49. 4597.  93.82  2905.  1958. 11.3
     23  48. 4693.  97.77  2966.  2030. 11.0
    24  47. 4793. 101.97  3029.  2100. 10.8
     25  46. 4897. 106.46  3095.  2169. 10.6
     26  45. 5006. 111.24  3164.  2236. 10.3
     27  44. 5120. 116.35  3236.  2302. 10.1
     28  43. 5239. 121.83  3311.  2366.  9.9
     29  42. 5363. 127.70  3390.  2429.  9.7
     30  41. 5494. 134.00  3472.  2490.  9.4
     31  40. 5632. 140.79  3559.  2550.  9.2
     32  39. 5776. 148.10  3650.  2609.  9.0
     33  38. 5928. 156.00  3746.  2665.  8.7
     34  37. 6088. 164.54  3848.  2721.  8.5
     35  36. 6257. 173.81  3955.  2775.  8.3
     36  35. 6436. 183.89  4068.  2827.  8.0
     37  34. 6625. 194.86  4187.  2878.  7.8
     38  33. 6826. 206.85  4314.  2927.  7.6
     39  32. 7039. 219.98  4449.  2975.  7.4
     40  31. 7267. 234.40  4592.  3021.  7.1
     41  30. 7509. 250.29  4746.  3066.  6.9
     42  29. 7768. 267.85  4909.  3110.  6.7
     43  28. 8045. 287.32  5084.  3152.  6.4
     44  27. 8343. 309.00  5273.  3192.  6.2
     45  26. 8664. 333.23  5476.  3231.  6.0
     46  25. 9010. 360.42  5695.  3268.  5.7
     47  24. 9386. 391.08  5932.  3304.  5.5
     48  23. 9794. 425.83  6190.  3339.  5.3
     
     

    (b) Single UBK7 50 degree prism at LBT
     

    R =     21443.7
     One arcsec maps to     126.255 microns, or     8.41700 pixels on detector

     Order, WC(A), FSR(A), FSR(pix),YC(pix), DY(arcsec):
      1  70. 3218.  45.97  2034.   105. 12.4
      2  69. 3265.  47.31  2063.   208. 12.3
      3  68. 3313.  48.72  2094.   310. 12.1
      4  67. 3362.  50.18  2125.   410. 11.9
      5  66. 3413.  51.71  2157.   509. 11.7
      6  65. 3466.  53.32  2190.   606. 11.6
      7  64. 3520.  55.00  2224.   702. 11.4
      8  63. 3576.  56.76  2260.   796. 11.2
      9  62. 3633.  58.60  2296.   888. 11.0
     10  61. 3693.  60.54  2334.   980. 10.8
     11  60. 3754.  62.57  2373.  1069. 10.7
     12  59. 3818.  64.71  2413.  1158. 10.5
     13  58. 3884.  66.96  2455.  1244. 10.3
     14  57. 3952.  69.33  2498.  1330. 10.1
     15  56. 4023.  71.83  2542.  1414. 10.0
     16  55. 4096.  74.47  2588.  1496.  9.8
     17  54. 4172.  77.25  2636.  1577.  9.6
     18  53. 4250.  80.19  2686.  1656.  9.4
     19  52. 4332.  83.31  2738.  1734.  9.2
     20  51. 4417.  86.61  2791.  1810.  9.1
     21  50. 4505.  90.10  2847.  1885.  8.9
     22  49. 4597.  93.82  2905.  1958.  8.7
     23  48. 4693.  97.77  2966.  2030.  8.5
     24  47. 4793. 101.97  3029.  2100.  8.4
     25  46. 4897. 106.46  3095.  2169.  8.2
     26  45. 5006. 111.24  3164.  2236.  8.0
     27  44. 5120. 116.35  3236.  2302.  7.8
     28  43. 5239. 121.83  3311.  2366.  7.6
     29  42. 5363. 127.70  3390.  2429.  7.5
     30  41. 5494. 134.00  3472.  2490.  7.3
     31  40. 5632. 140.79  3559.  2550.  7.1
     32  39. 5776. 148.10  3650.  2609.  6.9
     33  38. 5928. 156.00  3746.  2665.  6.8
     34  37. 6088. 164.54  3848.  2721.  6.6
     35  36. 6257. 173.81  3955.  2775.  6.4
     36  35. 6436. 183.89  4068.  2827.  6.2
     37  34. 6625. 194.86  4187.  2878.  6.0
     38  33. 6826. 206.85  4314.  2927.  5.9
     39  32. 7039. 219.98  4449.  2975.  5.7
     40  31. 7267. 234.40  4592.  3021.  5.5
     41  30. 7509. 250.29  4746.  3066.  5.3
     42  29. 7768. 267.85  4909.  3110.  5.2
     43  28. 8045. 287.32  5084.  3152.  5.0
     44  27. 8343. 309.00  5273.  3192.  4.8
     45  26. 8664. 333.23  5476.  3231.  4.6
     46  25. 9010. 360.42  5695.  3268.  4.4
     47  24. 9386. 391.08  5932.  3304.  4.3
     48  23. 9794. 425.83  6190.  3339.  4.1

    (c)  Two 35 degree UBK7 prisms, MMT

    Rtheta =     27750.9
     One arcsec maps to     97.5600 microns, or     6.50400 pixels on detector

     Order, WC(A), FSR(A), FSR(pix),YC(pix), DY(arcsec):
      1  70. 3218.  45.97  2034.   147. 22.5
      2  69. 3265.  47.31  2063.   291. 22.2
      3  68. 3313.  48.72  2094.   433. 21.9
      4  67. 3362.  50.18  2125.   574. 21.6
      5  66. 3413.  51.71  2157.   712. 21.2
      6  65. 3466.  53.32  2190.   848. 20.9
      7  64. 3520.  55.00  2224.   982. 20.6
      8  63. 3576.  56.76  2260.  1114. 20.3
      9  62. 3633.  58.60  2296.  1244. 20.0
     10  61. 3693.  60.54  2334.  1372. 19.6
     11  60. 3754.  62.57  2373.  1497. 19.3
     12  59. 3818.  64.71  2413.  1621. 19.0
     13  58. 3884.  66.96  2455.  1742. 18.7
     14  57. 3952.  69.33  2498.  1862. 18.4
     15  56. 4023.  71.83  2542.  1979. 18.0
     16  55. 4096.  74.47  2588.  2094. 17.7
     17  54. 4172.  77.25  2636.  2207. 17.4
     18  53. 4250.  80.19  2686.  2318. 17.1
     19  52. 4332.  83.31  2738.  2427. 16.7
     20  51. 4417.  86.61  2791.  2534. 16.4
     21  50. 4505.  90.10  2847.  2639. 16.1
     22  49. 4597.  93.82  2905.  2741. 15.8
     23  48. 4693.  97.77  2966.  2842. 15.5
     24  47. 4793. 101.97  3029.  2940. 15.1
     25  46. 4897. 106.46  3095.  3036. 14.8
     26  45. 5006. 111.24  3164.  3131. 14.5
     27  44. 5120. 116.35  3236.  3223. 14.2
     28  43. 5239. 121.83  3311.  3313. 13.8
     29  42. 5363. 127.70  3390.  3401. 13.5
     30  41. 5494. 134.00  3472.  3487. 13.2
     31  40. 5632. 140.79  3559.  3570. 12.9
     32  39. 5776. 148.10  3650.  3652. 12.6
     33  38. 5928. 156.00  3746.  3732. 12.2
     34  37. 6088. 164.54  3848.  3809. 11.9
     35  36. 6257. 173.81  3955.  3885. 11.6
     36  35. 6436. 183.89  4068.  3958. 11.3
     37  34. 6625. 194.86  4187.  4029. 10.9
     38  33. 6826. 206.85  4314.  4098. 10.6
     39  32. 7039. 219.98  4449.  4165. 10.3
     40  31. 7267. 234.40  4592.  4230. 10.0
     41  30. 7509. 250.29  4746.  4293.  9.7
     42  29. 7768. 267.85  4909.  4354.  9.3
     43  28. 8045. 287.32  5084.  4412.  9.0
     44  27. 8343. 309.00  5273.  4469.  8.7
     45  26. 8664. 333.23  5476.  4523.  8.4
     46  25. 9010. 360.42  5695.  4576.  8.0
     47  24. 9386. 391.08  5932.  4626.  7.7
     48  23. 9794. 425.83  6190.  4674.  7.4

    (d) Two 35 degree prisms at the MMT, made out of material with 1.2x the index of UBK7

    Rtheta =     27750.9
     One arcsec maps to     97.5600 microns, or     6.50400 pixels on detector

     Order, WC(A), FSR(A), FSR(pix),YC(pix), DY(arcsec):
      1  70. 3218.  45.97  2034.   226. 34.8
      2  69. 3265.  47.31  2063.   449. 34.3
      3  68. 3313.  48.72  2094.   669. 33.8
      4  67. 3362.  50.18  2125.   885. 33.3
      5  66. 3413.  51.71  2157.  1098. 32.8
      6  65. 3466.  53.32  2190.  1309. 32.3
      7  64. 3520.  55.00  2224.  1515. 31.8
      8  63. 3576.  56.76  2260.  1719. 31.3
      9  62. 3633.  58.60  2296.  1919. 30.8
     10  61. 3693.  60.54  2334.  2116. 30.3
     11  60. 3754.  62.57  2373.  2310. 29.8
     12  59. 3818.  64.71  2413.  2501. 29.3
     13  58. 3884.  66.96  2455.  2688. 28.8
     14  57. 3952.  69.33  2498.  2872. 28.3
     15  56. 4023.  71.83  2542.  3053. 27.8
    16  55. 4096.  74.47  2588.  3231. 27.3
     17  54. 4172.  77.25  2636.  3405. 26.8
     18  53. 4250.  80.19  2686.  3577. 26.3
     19  52. 4332.  83.31  2738.  3745. 25.8
     20  51. 4417.  86.61  2791.  3909. 25.3
     21  50. 4505.  90.10  2847.  4071. 24.8
     22  49. 4597.  93.82  2905.  4229. 24.3
     23  48. 4693.  97.77  2966.  4384. 23.8
     24  47. 4793. 101.97  3029.  4536. 23.3
     25  46. 4897. 106.46  3095.  4685. 22.9
     26  45. 5006. 111.24  3164.  4830. 22.4
     27  44. 5120. 116.35  3236.  4972. 21.9
     28  43. 5239. 121.83  3311.  5111. 21.4
     29  42. 5363. 127.70  3390.  5247. 20.9
     30  41. 5494. 134.00  3472.  5379. 20.4
     31  40. 5632. 140.79  3559.  5509. 19.9
     32  39. 5776. 148.10  3650.  5635. 19.4
     33  38. 5928. 156.00  3746.  5757. 18.9
     34  37. 6088. 164.54  3848.  5877. 18.4
     35  36. 6257. 173.81  3955.  5993. 17.9
     36  35. 6436. 183.89  4068.  6106. 17.4
     37  34. 6625. 194.86  4187.  6216. 16.9
     38  33. 6826. 206.85  4314.  6323. 16.4
     39  32. 7039. 219.98  4449.  6426. 15.9
     40  31. 7267. 234.40  4592.  6526. 15.4
     41  30. 7509. 250.29  4746.  6623. 14.9
     42  29. 7768. 267.85  4909.  6717. 14.4
     43  28. 8045. 287.32  5084.  6807. 13.9
     44  27. 8343. 309.00  5273.  6895. 13.4
     45  26. 8664. 333.23  5476.  6979. 12.9
     46  25. 9010. 360.42  5695.  7059. 12.4
     47  24. 9386. 391.08  5932.  7137. 11.9
     48  23. 9794. 425.83  6190.  7211. 11.4

    Top