Design Study 1

A design study was performed by NGD while at the Anglo-Australian Observatory in 1994-1995 and worked out in detail by Damien Jones of Prime Optics. Instead of having an instrument which rotates as shown above we decided to split the light by using a rooftop construction consisting of two gratings. These gratings diffract the light to the left and to the right, splitting the pupil, and thus an image of the galaxy is formed by each of the two cameras. Wheras the orientation of the field is the same for both, the dispersion directions differ by 180o.

A summary of the specifications:


PNS budget estimation, December 1996
N.G. Douglas

Design Study stick with Prime Optics design (cost approx $4000)
Fabrication ICOS quote $58,000 (adjusted for inflation)
Assembly and Test 3 months in house 17,500
Grating Milton Roy $7,000
[O III] Filter(s) ? $7,500
Design, Drawing 4 months in house $23,000
Fabrication 6 months, mixed in house and under contract $40,000
Materials   $10,000
Controller (x1) SDSU/Leach 25,000
Cryostat and Chip (x2) Tek $45,000
Project Manager, Administration 30,000
Total:   $A263,000


Here we explore some implementations of the PNS-I design as they might appear at specific telescopes - the 4.2-m William Herschel Telescope (top) and the 8-m ESO Very large telescope.

In PNS-1 the diffracted beam is separated from the incoming beam by about 20 degrees. The design would be modified to the latest specifications and for use with a 2048² CCD with 15um pixels. Note that the collimated beam size is about 120mm, as shown here. The sketches are to scale.

The `classic' PNS-I design calls for pupil-splitting to give two simultaneous images -- this is shown in the lower sketch -- but for economy we might build the instrument with only one camera instead. It is better to assume this while designing the optics since the dual camera version can always be built later by making a second camera, while the minimum camera for the split-pupil design has a slightly smaller aperture than that required for the one-arm design.

The field of view of the instrument is slightly rectangular owing to the anamorphic effect of the grating. For the f/15 VLT the long side is about 260mm in the focal plane. For comparison the corresponding value for the f/8 AAT, for which the design study was done, is 100mm. Accordingly the VLT collimator will need to be much larger and possibly very expensive. To capture all the field, the first lens (`Field Lens') of a VLT collimator would have to be 350mm in diameter. For the f/11 WHT the Field Lens has to be 260mm in diameter. These two circular apertures are shown circumscribed on the corresponding apertures in the smaller drawings on the left.

In the case of the VLT, the inscribed circle is also about 260mm in diameter. If this defined the aperture, the corners of the field would not be imaged but the cost of the optics would be comparable to that of the WHT collimator. Perhaps even the same Field Lens could be used.

The f/15 collimator would no doubt be close to the PNS-II design just arrived at by Damien Jones of Prime Optics although the F=238mm camera would probably be more similar in design philosophy to that used in PNS-I.


The success of the PNS depends critically on the efficiency achieved. Because the instrument will be optimised for a small wavelength range, the efficiency will far exceed that of general-purpose instrumentation. We expect the instrument to have about 62% efficiency, giving just about 36% total efficiency into both arms as shown here. For comparison, ISIS at the WHT has 10-20% efficiency.

Filter AR coated narrow-band 0.85
Detector Loral LR2048 or equivalent 0.95
Telescope Cassegrain focus (two reflections, loss at central obstruction) 0.722
PNS instrument efficiency 0.622
Total 0.363



This shows the performance of the PNS-1 design at a number of different observing sites (with appropriate change in collimator).

Assumptions: 2000² 15um pixels, 160 mm collimated beam, efficiency 36%, sky 21.4 mag/arcsec², galaxy 21.8 mag/arcsec², 40Å bandpass, 28,000s integration time

Dtel f seeing (") image scale (pxl/") PSF (pxl) F.O.V.(') resoln (km/s) S (e-) SNR
3.9m 8 1.0 2.76 x 2.31 2.76 12.1 x 14.4 5.9 5363 36
4.2m 11 0.8 2.97 x 2.49 2.38 11.2 x 13.4 5.2 6220 41
8.0m 15 0.5 5.66 x 4.74 2.83 5.89 x 7.04 5.9 22568 114
The dispersion is 0.227Å/pxl in all cases
PSF is the size of the seeing disk, in pixels, in the undispersed direction. The PSF in the dispersed direction is slightly smaller due to the anamorphic effect of the grating, though it also contains a contribution due to the natural linewidth of the PNe (about 0.17Å).

Scale and FOV given as (spatial) x (dispersed) values.

Resolution is here defined as that corresponding to a centroiding accuracy of 0.25 the PSF FWHM in the dispersion direction, or 0.4 pixels, whichever is larger. After determining the displacement between two images this centroiding is improved by sqrt(2) in velocity units.

The background level (B) is 16,473 counts per pixel in all cases.
The signal (S) is for the brightest PNe in a galaxy at D = 20 Mpc (these PNe have a flux of 1.3e-15 at CenA (assume 3.3 Mpc) and SNR = S/(sqrt(S+B))

For convenience we give for one case the corresponding values for the PNS-2 design in which the collimated beam diameter is reduced to 120mm (this is the only important change):

Dtel f seeing (") image scale (pxl/") PSF (pxl) F.O.V.(') resoln (km/s) S (e-) SNR
8.0m 15 0.5 5.13 x 5.07 2.57 6.50 x 6.57 13.6 22568 114