TAURUS++:   Long-slit and microslit modes with TAURUS-2

Authors: Joss Hawthorn.



Instrument parameters and comparison with LDSS++
Available gratings and order sorting filters
Available detectors
Charge shuffling and available field of view
Spectral dispersion
Available slits for long-slit spectroscopy
Plate-scale comparison with LDSS++
Throughput comparison with LDSS++
PSF comparison with LDSS++
Multi-slit spectroscopy
Mask design and fabrication
Field acquisition and alignment
Operation of TAURUS++
Setting up TAURUS++
Setting up nod & shuffle
Initiating exposures
What to do if the system stalls (shuffle_tidy)
Appendix A: History of TAURUS++
Appendix B: Possible improvements in TAURUS++ performance
Appendix C: The effect of VPH tilts
Appendix D: Theoretical comparison of TAURUS and LDSS  (Zemax)
Appendix E: Calibration of TAURUS++ field aberration and plate scale


TAURUS++ is a low resolution spectrograph which can be used for long-slit or multi-slit observations. The instrument is used at the Cassegrain focus of the AAT. Imaging, long slit, and slit-less spectroscopy modes are also available. TAURUS++ absorbs the functionality of LDSS++ which some observers will be more familiar with. For this reason, we provide a detailed comparison with LDSS++ within this document.

In brief, TAURUS++ is a mode of operation which closely resembles LDSS++. Multi-slits can be laid down over a field of view of about 10' compared with 12' for LDSS++. Operationally, TAURUS++ is much like LDSS++ except that an observer is not required in the Cass cage. The LDSS++ has a 40% advantage in field size for band-limited work (e.g. pre-filtering with a 100A TTF filter), but only a 20% advantage when the full spectral range is used. The main advantage of TAURUS++ is higher throughput across the full spectral band, particularly in the red where the LDSS++ performance plummets. The main advantage of LDSS++, apart from the bigger field, is the better imaging performance over most of the field. TAURUS++ imaging performance is adequate except there is evidence for chromatic aberrations over about 20% of the field. These can be effectively removed with the nod & shuffle operation. Both LDSS++ and TAURUS++ benefit from nod & shuffle for sky subtraction in the red, or during bright of moon. Operationally, nod & shuffle is straightforward and relatively reliable for both instruments.

TAURUS++  is not offered as a common user instrument. It is an expert instrument which means that the proposal must be thoroughly discussed with the Instrument Scientist to assess feasibility before submission and more work will be required of the applicants than usual to prepare for and carry out the observations.


Instrument parameters and comparison with LDSS++.
Here are the important numbers for TAURUS and LDSS used at the AAT f/7.87 focus with the MITLL chips. The LDSS plate scale is known to be correct to the accuracy shown; the TAURUS plate scale is good to better than 0.03%.
Nominal f/7.87 focus mm -54 0
Fcoll mm 560 474
Fcam mm 140 128
Total length mm 824 790
Beam size mm 71 60
Focal plane plate scale arcsec/mm 6.697 6.680
MITLL pixel scale arcsec/pix 0.37 0.37
EEV pixel scale arcsec/pix 0.33 0.33
Field radius arcmin 6.21 4.76
Field radius pix 1007 772
Field curvature
Distortion at field edge 2% 3.5%
Chromaticity weakly chromatic achromatic

Available gratings and order sorting filters.

For the purposes of single slit or multi-slit spectroscopy, Taurus makes use of two VPH gratings with R = 1000 and 3000. Both of these gratings are optimized for the red, so the  MITLL chips (particularly the MITLL3) are highly recommended. These are fixed gratings and cannot be tuned. At R = 1000, the full spectral range (480-1109nm for on-axis object) can be imaged onto the MITLL chips with a sampling of 2.8A/pix. At R = 3000, the dispersed spectral range is approximately 738-928nm (on-axis object) with a sampling of 0.93A/pix. With the MITLL chips, the fringeing is minimal at R = 1000, although it is more noticeable at R = 3000. Most spectroscopy is done in nod & shuffle mode such that fringeing is not really an issue.

The VPH gratings exhibit degradation in blaze efficiency and a shift in spectral range as a function of off-axis angle. More details relating to spectral dispersion are discussed below.

If you need to use the R=3000 grating, special arrangements will need to be made with site (see Appendix B) well ahead of the run.

There are 3 order sorting filters to choose from: GG475, OG515 and OG550. The transmission curves are shown below:

Available detectors.

 The CCD detectors available at the AAT are detailed here.  The plate scales are shown below. In most cases, users will want to use f/8 in order to ensure the largest possible field of view. Typical seeing at the AAT rarely justifies the use of f/15.
AAT f/8  (9.5' max field of view) AAT  f/15 (5' max field of view)
Tek 1Kx1K (24um pix) 0.594 "/pix 0.315 "/pix
MITLL2, MITLL3 2Kx4K (15um pix) 0.37 "/pix 0.20 "/pix
EEV2 2Kx4K (13.5um pix) 0.33 "/pix 0.18 "/pix

The detector transmission curves are shown below:

While it is possible to use Taurus++ with either the EEV or MITLL chips, the gratings are blazed for work beyond 700nm, and therefore the MITLL3 is the recommended detector.

Charge shuffling and available field of view.

Most users make use of nod & shuffle when using gratings with Taurus. The available field of view will depend on whether you are using the EEV or MITLL. The EEV is 10% smaller in linear dimension, 20% smaller in area, as illustrated below. The charge shuffling process occurs in the vertical direction on the CCD. It utilizes one third of the long axis of the CCD (i.e. 1365 pixels) to store each of two images, e.g. on field and off field. The third part of the CCD is lost to the shuffle process, as shown below.

Note that if nod & shuffle is not required, the full 9.5' field of view is available to the user with either detector.

Note from the figure that the convention is to place North at the top of the CCD (as shown on the XMEM display) and East to the left of the CCD. This normally requires that the Instrument PA = 180 degrees which is something that must be checked at the very beginning of the run (see below).

Since the available detector area extends further in the horizontal (east-west) direction, this is the conventional direction for dispersing the grating as shown here. In fact, the grating can be rotated in its holder to any angle on the sky but we strongly recommend sticking with the horizontal direction.

Spectral dispersion.

The VPH gratings show a pronounced shift in the dispersed spectral range falling on the detector with off-axis angle. All values shown below are for the MITLL detector. For the EEV, the range needs to be squashed by 10%. The tables below show the spectral range and mid wavelength (cols 3-5) as a function of slit displacement in pixels (col 4) and off-axis angle (col 5) from the on-axis position. Since the dispersion is east-west, a -ve slit displacement corresponds to an eastward shift of the slit. (In fact, the grating can easily be made to disperse to the east or west by a simple rotation in the grating holder.)

R=1000 grating (this is the grating normally used for Taurus++ work)
displacement (pix)  off-axis angle (deg)  min. wavelength  mid. wavelength  max. wavelength
-800  -5.3  2940  5620  8760 
-600  -4.0  3370  6160  9340 
-400  -2.7  3820  6720  9910 
-200  -1.3  4300  7290  10500 
0.0  4800  7860  11090
200  1.3  5330  8450  11690 
400  2.7  5870  9030  12300 
600  4.0  6440  9620  12950 
800  5.3  7010  10220  13620 

R=3000 grating (this grating is not easy to use and requires prior approval from the Instrument Scientist)
displacement (pix)  off-axis angle (deg)  min. wavelength  mid. wavelength  max. wavelength
-800  -5.3  6660  7570  8540 
-600  -4.0  6830  7750  8730 
-400  -2.7  7010  7930  8910 
-200  -1.3  7190  8120  9100 
0.0  7380  8310  9280
200  1.3  7560  8490  9470 
400  2.7  7750  8680  9660 
600  4.0  7940  8880  9860 
800  5.3  8130  9070  10050 

Available slits for long-slit spectroscopy.

For conventional long-slit spectroscopy, we have a range of slits. The properties are summarised in the table below.
material reference
slit width  slit length N-S holes struts offset
brass SLIT_B1 1" 12.2' yes 2 0
steel SLIT_S2 1.33" 7.2' yes 4 0
brass SLIT_B3 1.67" 12.2' yes 2 0
steel SLIT_S4 2" 7.2' yes 4 0
brass SLIT_B5 5" 12.2' yes 2 0
brass SLIT_B6_OFF 1" 12.2' yes 2 500
brass SLIT_B7_OFF 1.67" 12.2' yes 2 500
brass SLIT_B8_OFF 5" 12.2' yes 2 500
brass SLIT_B9_OFF 10" 12.2' yes 2 500

The slits are stored in a portable box in the filter room. These are clearly labelled with the reference number shown in column 2.

All slits have supporting struts spaced along the slit length. Here is a postscript plot of one of the 7.'2 on-axis slits: in the figure, the actual slit + supporting struts are shown in black; the green rectangles illustrate the extent of the dispersed spectrum at the detector. Four of the slits have been offset by 500 pix. From the table above, this means that the spectral response is shifted 500A to the blue or to the red, depending on whether the slit offset is to the left or right in the mask holder. Of course, you can insert the mask into the holder in either configuration. Please ensure that the slit is placed N-S (up-down) in the mask holder. All masks have the necessary holes to facilitate this (see column 4 above). The two steel masks will be delivered in mid Dec 2003.

If the slits described above are not optimal for your observing run, please contact the Instrument Scientist as far in advance of your observing run as possible (4-6 weeks at least). (Please note that our rapid turnaround slit manufacturers at Mt Stromlo can only make slit widths of 200um, 250um, 300um, and so forth. They cannot cut narrower slits than 200um.)

Here are the resolutions you can expect with the R=1000 and R=3000 gratings for different slit sizes. The effective resolutions (at a wavelength of 7500A) and are accurate to about 5% and apply to either the MITLL or EEV CCD.
grating  dispersion
slit   width  effective resolution
R=1000  2.8 1.0" 1000
1.33" 750
1.67" 600
R=3000 0.93 1.0" 3000
1.33" 2300
1.67" 1800

Plate-scale comparison with LDSS++.

There is an important difference in plate scale between LDSS and TAURUS. The former is the longer instrument and was originally raised upwards to allow the IPCS to fit into the Cass cage. The LDSS focus position is 54mm higher than for TAURUS. To understand the impact on the plate scale, see the optical layout of the AAT. The f/3.3 primary focuses to a point at 12.7m radius. The f/7.87 secondary is 4.26m ahead of the prime focus. The f/3.3 to f/7.87 transition results in a magnification of m = 2.4. Moving the LDSS focus at Cass by -54mm requires the secondary to be moved upwards by 54/(m*m+1) = 8mm. This reduces the object distance and enlarges the plate scale by 8/4260 = 0.19%. Conversely, the offset LDSS focus reduces the image distance which reduces the plate scale by (54-8)/10320 = 0.45%. The value 10320mm is the distance between the secondary mirror and the instrument focus. The overall effect is that the plate scale of TAURUS is 0.26%  (=0.45%-0.19%) smaller than for LDSS. In other words, the LDSS images are shifted inwards by 0.26% (demagnified) compared to TAURUS.  (It's worth noting that Peter Gillingham derived a plate scale error of 0.25% from a comparison of the Zemax files, and Will Saunders derived 0.22% from TAURUS  observations of guide stars through LDSS masks.)

Throughput comparison with LDSS++.

The total system efficiency curves for Taurus spectroscopy at R=1000 and R=3000 are shown below. The R=1000 curve for Taurus indicates better red performance than LDSS, even after allowing for the different detectors used. The expected peak in the Taurus++ response is  P% =  mirrors * TAURUS * vph * MITLL3 = 65 * 70 * 80 * 90 = 33%. Note that the response at 1um was better than expected.  CCD responses at these wavelengths are difficult to measure and it may be that the quoted values for MITLL3 are a little low beyond 900nm. (This is a well known problem.) However, the R=1000 response needs to be checked again since a correction was made for the poor seeing.

The R=3000 response, shown below, was calibrated under ideal photometric conditions.

It is important to note that the peak efficiency of the grating falls with off-axis angle, i.e. for a slit displaced from the centre of the field. A study of this was made by Ivan Baldry for the R=1000 grating. For an on-axis object, the peak efficiency of the VPH grating alone is 85%, falling to 80% at 1.5 deg, and 65% at 3 deg. From theoretical considerations, the effect should go as cos(theta) where theta is the off-axis angle (see Appendix C).

PSF comparison with LDSS++.

The TAURUS++ field of view exhibits a complex aberration pattern over about a third of the field (see Appendix E). This was quite unexpected from our detailed Zemax analysis (see Appendix D). TAURUS++ should behave better than LDSS++ over most of the field but this is not found to be the case. TAURUS++ appears to exhibit spherochromatic aberration which is hardly apparent in the blue, but highly visible beyond 750nm. There appears to be a misaligned optical element, most likely in the camera.

As of July 2001, we do not have the necessary science data over the full field to determine how the system throughput is affected by the aberration. This is expected to be at least as high as a 15% drop in throughput when compared to the on-axis performance. Nod & shuffle sky subtraction is needed to ensure that this is done properly, but this is the normal mode of operation in any event beyond 750nm.

Multi-slit spectroscopy
Mask design and fabrication.
There is a critical procedure for mask manufacture which must be adhered to by all prospective observers. This procedure must be completed at least 6 weeks before the scheduled observing run. Here are the steps:
1.  The observer must make an astrometric catalogue of targets and fiducial stars with an internal accuracy to better than one third of the slit/hole size. It is essential that the observer get this part right before proceeding further.  The fiducial stars need to be brighter than  15th or 16th mag in any band; you will need 4 or more stars widely spaced over the field.

Our new procedure for making masks is found to be more reliable and simpler than before. You will need to generate an ASCII file with a very simple format as shown in (2) below.  We find GAIA under Starlink software to be an adequate facility for doing rigorous astrometry and mask preparation -- see here for an illustrated guide. For any image with a properly formatted FITS header, the observer can overlay pointers from a wide selection of astrometric standard catalogues. All images from astronomical cameras exhibit astrometric distortion. GAIA allows you to compute the distortion map to a tenth of a pixel or better.

2.  The observer needs to generate a list of linear offsets (in arcseconds) from a field centre. Your final mask file of positions should look like:

# 1 ID
154-559  05:35:15.421 -05:26:00.09 -137.69 -153.09 0
017-635  05:35:01.667 -05:26:36.28   67.70 -189.28 0
127-1947 05:35:12.610 -05:19:45.71  -95.74  221.29 0
042-017  05:35:04.150 -05:20:15.68   30.61  191.32 0
016-636  05:35:01.550 -05:26:36.57   69.44 -189.57  1
035-609  05:35:03.419 -05:26:10.12   41.54 -163.12  1
052-603  05:35:05.174 -05:26:03.76   15.33 -156.76  1
055-555  05:35:05.441 -05:25:55.93   11.34 -148.93  1

This is a format that GAIA can read directly as we demonstrate here. Note the offsets are N & E, with N from the field centre being +ve, and W from the field centre being +ve. The position angle PA of the field must be zero degrees. The last column denote fiducial ( 0 ), and target ( 1 ). The X_IMAGE, Y_IMAGE coordinates are used for mask making. The observers must ensure that they bring to site a log of accurate field centres for each mask and finding charts/images for each mask to facilitate telescope alignment.

3.   Once you have made the ASCII target list in the above format, download this tar file  and follow the short README file in order to compile the MaskMake program. You should check that the Makefile points correctly to your local PGPLOT libraries. This code allows for a single object per row, and avoids spectral overlaps. The input file which is piped to the code allows the user to specify microslits vs. microholes, and their specific properties. The complete list of options is given in the test.input file in the tar file:

test.ASC              ! ASCII filename (GAIA format as shown above)
10.0                    ! Diameter of fiducial hole in arcseconds (you will need at least four of these widely spaced; please ensure holes are no smaller than 10")
2.0                      ! Diameter of microhole in arcseconds (this assumes you've set the last parameter to 0; note median seeing at the AAT is 1.5")
1.0                      ! Slit width in arcseconds (all slit parameters assume you've set the last parameter to 1)
5.                        ! Slit length in arcseconds
1.5                      ! Offset in arcseconds along slit for nod & shuffle (this is normally preferred so that object falls within slit in both the on and off positions)
-1.0                    ! Offset direction ( -ve towards top of slit)
0.5                      ! Minimum spacing in arcseconds between slits to avoid overlaps
1                         ! Use holes ( 0 ) or slits ( 1 )

While observers are allowed to modify the code for special needs (e.g. densely packed narrowband spectra), the AAO takes no responsibility for resultant errors in the mask files. You are strongly advised to keep the fiducial holes as large as possible, say 10". Please consult the Instrument Scientist if there are special reasons why the fiducial holes need to be made smaller.

4.   The generated .dxf files should then be forwarded to the Instrument Scientist who will make contact with the mask cutters at Mount Stromlo Observatory.  Due to intensive workload, the cutting of masks must be scheduled a month ahead of the scheduled run. The mask designs must be with the instrument scientist at least 6 weeks before the scheduled run to ensure that the masks are acceptable. Failure to meet this deadline will result in the cancellation of your run.

Field acquisition and alignment.

Acquiring a field is an art form which requires an expert observer to be present. The 10" holes for the fiducial stars ensures that the guide stars can always be acquired within the holes. The AAT can acquire a field to within 1" once it has been calibrated with an APOFF operation.

But the guide stars must be centered on the holes to high accuracy to provide accurate alignment of the microslits with the target objects. This requires software (currently under development) which fits to the fiducial stars within their circular apertures, and returns a suitable offset and rotation, if required.

We cannot stress enough the importance of the fiducial guide stars. These should be well spaced over the full field to provide an adequate lever arm for guiding.

Operation of TAURUS++.

Setting up TAURUS++.

           Here is a list of tasks that you need to get through up to and including your first science exposure.

  •  We strongly recommend that you work with an Instrument Position Angle of 180 deg. This puts N up and E left on the XMEM. Any of the detectors can be rotated to ensure that the charge shuffling is north-south on the detector. But do check whether your masks were designed for a PA of 180 or 0 deg, or some odd angle.
  •  Check that either the R=1000 or R=3000 VPH is loaded into the ETALON wheel. The R=3000 VPH requires expert guidance since we need to lock the pupil wheel to the motion of the etalon wheel. In order to use it, you must have requested the R=3000 grating in your observing proposal.
  •  Check that there is a load file under OBSERVER on the VMS system with your favourite wheel  positions set. The FILTER wheel is for the order sorters, the APERTURE wheel is for the masks, the PUPIL should always be position 8, and the ETALON wheel is for the VPH. The load file also sets the CAMERA focus so edit in the correct value once this is determined (see next item). See for example disk$user:[observer.icl_load]vph_jan02.icl. This is loaded from the TAURUS control window after hitting "." on the numeric keypad. At the prompt, type "load disk$user:[observer.icl_load]file.icl"
  •  Check the camera focus as we do for TTF (see here) using the matrix mask.
  •  Get the order sorter(s) loaded and mark on this postscript plot.

  • Appendix A.  History of TAURUS++.

    After LDSS++ was successfully commissioned at the AAT in 1998, a proposal was put forward by JBH & KGB to absorb the LDSS++ operations into TAURUS. This was accompanied by an extensive science case.  It has long been realized that the two instruments share important similarities and that there is a well-defined hybrid which would allow both instrument functions to operate within a single machine. While the necessary funds were unavailable (A$120K), the push to combine both instruments set the stage for the ATLAS project at the AAT, and OSIRIS on the Grantecan 10m telescope.

    During 2000, we undertook extensive work to establish whether the existing TAURUS can still be used in microslit mode since the advantages of a merged machine remain. These are: (i) TAURUS++ would be a world first for the AAO, combining tunable spectroscopic imaging or multi-slit spectroscopy on any given observing night; (ii) consolidation of the instrument suite since LDSS would be decommissioned; (iii) automated filter wheels which bypasses the need for a second observer riding in the Cass cage;  (iv)  improved instrument response at all wavelengths particularly beyond 800nm.

    In summary, TAURUS++ outperforms LDSS++ at all wavelengths in terms of throughput and ease of operation. However, there appears to be an aberration (due to a misaligned camera element) which affects a third of the TAURUS field, which is particularly prominent at the reddest wavelengths. We await science data to establish what impact the aberration is having on the various science programmes. TAURUS is in need of minor refurbishment which in fact is long overdue. We propose that the optical realignment be coordinated with solgel application to all optical elements. This will further enhance the optical performance of TAURUS++.

    Appendix B.    Possible improvements in TAURUS++ performance.

    Increasing the spectroscopic resolution of TAURUS++.

    On the matter of instrument consolidation, RAS asked to what extent TAURUS++ could take on some of the workload of RGO spectrograph. JGR has run detailed calculations to show that the range of allowed resolving powers (beyond what we get with the red vph) is as follows:
    Resolving power VPH prism apex angle VPH lines / mm
    1000 15 460
    1500 20 630
    2000 25 800
    2500 30 970
    3000 35 1150
    3500 40 1334

    All of these could be readily manufactured, and would fit within the pupil space. The expected performance of TAURUS++ is significantly better than the RGO spectrograph for R < 3000 over the full optical range. One advantage of the RGO spectrograph is that the grating can be tilted for optimal coverage over the detector area.

    Ivan Baldry has now designed an R=3000 grating for use with TAURUS++ (see the figure below). This grating is so unwieldy that special arrangements are called for in order to use it (e.g. the pupil wheel needs to be disabled).


    Improved anti-reflection coatings.

    Battles are won and lost at the air-glass interface. The existing TAURUS optics are coated with the humblest of dielectrics, MgF2. As it turns out, this is fortuitous as it can be treated directly with solgel to produce just about the best broadband AR coating around. See the figure below.

    The combination of solgel + MgF2 raises the transmittance per air-glass surface from 0.984 (peak) to 0.998 (peak) with the AR differential being close to 2% in the red. For 14 air-glass surfaces, this amounts to a 30% improvement in the instrumental transmission at 1um, although more like 20-25% improvement within the R and I bands. Note that an improvement in transmission with AR coatings often produces the added benefit of suppressing the scattered light within the system. Conclusion: TAURUS++ would benefit substantially from the application of solgels. The newly coated TAURUS optics would ensure a twofold throughput advantage compared to Keck/LRIS in the R band, with an even bigger gain in the I band.

    Appendix C.    The effect of VPH tilts.

    Note that for an on-axis object, the straight through wavelength for the R=1000 grating (determined by the bonded prisms) is 786.3nm, and this is not affected by tilting of the VPH. But we can shift the peak transmission (on-axis) to more like 700nm by tilting the VPH within its cell by 1.5 degrees. This is likely to be of limited use for microslit work because, as Gordon Robertson points out in his SPIE paper, as you go off axis, the transmission peak shifts progressively further to the blue in one direction, and to the red in the opposite direction. The edge of the illuminated TAURUS-2 field is about 5.4 degrees in physical angle from the field centre, i.e. 800 pixels. (Note that if we could get to the corner of a 2K detector, this would be more like 10 degrees!). This is only a problem in the direction of the dispersion, not across the dispersion.

    For extended objects of multi-slit work, flux calibration with either TAURUS++ or LDSS++ requires standards to be observed at several positions in radius along the dispersion direction.  See the official AAO VPH web site  for more details on AAO VPHs.

    Appendix D.     Theoretical comparison of TAURUS and LDSS.

    The original design parameters are available for both TAURUS and LDSS. A detailed comparison under Zemax shows that TAURUS is not performing to spec as demonstrated above. Here are some of the findings from the Zemax analysis.

    Both TAURUS and LDSS cameras operate at f/2. The LDSS optics are more telephoto than TAURUS so that the LDSS collimator length is 440 mm, and the camera length more like 310mm. The clear LDSS collimated cavity is 72 mm which is somewhat smaller than TAURUS. In terms of imaging properties,  both TAURUS and LDSS were optimized for 27um pixel detectors. In theory, TAURUS has the better performance at all wavelengths over the full field. The numbers are tabulated below. LDSS PSF numbers are about 15% worse in linear dimension.
    Field angle (radius) arcmin 0.0 4.4 6.2
    PSF FWHM microns 24.0 28.8 33.2
    PSF FWHM arcsec 0.62 0.74 0.86

    The big difference between LDSS and TAURUS arises from the AR coatings. TAURUS uses MgF2 so therefore has a slow decline in performance well into the near IR. However, LDSS uses a 3-layer composite which declines rapidly beyond 800nm. TAURUS has a much better red performance than LDSS.

    Appendix E.   Calibration of TAURUS++ field aberration and plate scale.

    Experiment I.   On 16 Nov 2000, JBH did some tests with the red 786nm VPH + matrix mask using CuAr+W as the source, MITLL2A detector. The first tests were using the R1 filter as a blocker. The emission line wavelengths are given here; wavelength increases to right. You can download the original fits image from here and simply run imexamine and polyfit under iraf. This takes all of 5 mins if you are downloading locally. See the image subsets below: the LHS image is the centre of the field, the RHS image is the top corner of the field with the worst aberration. In fact, the images are very good across the entire field with typically 2.45 pixels FWHM.  The PSFs are the same as the best focus image ever achieved with TAURUS. Here is a summary of the psf properties in the centre of the field. A microscope was used to measure the matrix holes in the range 150-180um from centre to edge. 150um is 1" at the detector.
    filename wavelength (nm) line FWHM (A) psf eccentricity dispersion (A/mm; A/pix) aberration
    R1_VPH.fits 710 7.0 1.01 190.64; 2.859 none
    R8_VPH.fits 910 7.7 1.16 197.17; 2.957 spherochromatic

    The next set of tests used the R8 blocker. The emission line wavelengths are given here. As we found from an earlier run, the spherochromatic aberration is worse but still usable. Nod & shuffle should do a good job to remove the faint cometary tail in the sky subtraction. You won't see this tail in the faint, noisy emission lines. PSF fits over the field give 2.7-3.3 pixels so still quite usable. You can download the original fits image from here.

    Conclusion: the aberration at red wavelengths was not anticipated, at least judged from the Zemax file (see Appendix D). This appears to arise from a misaligned element in the TAURUS camera optics. (A much weaker effect is expected from any bowing within the aperture mask.) In practice, this aberration is not expected to adversely affect microslit observations below 800nm, although this could result in a significant loss in SNR (~15%) at the reddest wavelengths.

    Experiment II.  On 18-19 Dec 2000, GFL, WS and LK carried out more tests using the same set up in Experiment I, although this time in excellent observing conditions. Their plan was to determine experimentally (a) the difference in plate scale between TAURUS and LDSS, (b) the throughout performance of TAURUS++ at all wavelengths. Two fields were observed. The guide stars for the Orion field were deemed to be too faint for the bright background. The AXAF field demonstrated that the plate scale of the LDSS mask was in error by 0.2-0.23% as expected from theory. Conclusion: we now fully understand the difference in plate scale to the extent that exact masks can now be manufactured for TAURUS++ with the existing CAD/CAM interface.