This note describes the properties of the telescope and its "Bottom end" systems (as they affect offsetting to faint sources), indicates likely performance, and makes recommendations.

Telescope Offsetting

Offsetting over a couple of degrees or so using the telescope encoders is extremely common. Standard observing practice at UKIRT is as set out below:

Procedure for Telescope Offsetting

1. Using the HST Guide Star Catalog (current default) or another source (e.g. USNO a catalogues, digitised sky survey, etc.) select a guide star near the target object. (See Crosshead constraints under "Acquisition and guiding systems".)
2. Point to a nearby star (i.e. one withn a degree or two of the target) with accurately known position. (Stars with positions accurate to <0."1 are currently taken from the Carlsberg Meridian Catalogue (CMC), which has around one star per square degree.)
3. Begin fast guiding on the star.
4. Set system offsets to zero. The instrument and autoguider now share the same pointing properties.
5. Slew to the target object position. It will be approximately centred (± a few x 0."1) in the instrument aperture.
6. Drive the crosshead to the position of the preselected guide star.
7. Switch on the guider.

Limitations of telescope offsetting

The precision with which the telescope will point to the target object (which, if the target is bright enough, can be measured by the change in pointing when the autoguider locks on after the slew) is determined by three factors:

• The accuracy of the pointing model over the offset range
• The encoder performance in RA
• The encoder performance in DEC

The first of these should be a negligible error source over distances as small as this: the pointing model is good to a couple of arcsec (typically 1."5 rms over the sky soon after a pointing test) over the accessible sky, so should be well below 0."1 over a degree or so.

The encoders have digitisation resolution of 0."07, but there is a cyclic error of 0.3 arcsec RMS amplitude (it is NOT a sinusoid) and period ~20" on the sky. In RA, without guiding, this results in a ±~0."3 EW wobble with a period of 1.3s. In a period of twenty seconds or so, as the fast guider sends averaged error corrections to the telescope the encoder position error will be averaged down to ~0."1 or less.

In DEC, however, we do NOT get averaging, so a position in DEC is uncertain by 0."3 RMS (equivalent to ~1" peak-to-peak).

The result is that telescope offsetting will locate the centroid of an image to perhaps ±0."1 in RA, but only to about ±0."3 in DEC.

Telescope offsetting thus has definite limitations. These add to the uncertainties of using the bottom-end systems "backwards", as is commonly the case, by offsetting the telescope to a source and then "tweaking" the crosshead to get locked on to it or a guide star.

After all these strictures, however, it should be noted that many users of the current spectrometer CGS4 apparently find Telescope Offsetting quite adequate for the acquisition of most faint point sources.

Back to pointing.

Crosshead offsetting

The focal plane scale of UKIRT at the instruments is FPS=1.525±0.001 arcsec/mm. The scale at the crosshead location (below the pupil imaging lens) is estimated at FPSC=1.5724±0.001 arcsec/mm.

The crosshead, manufactured by the SKF bearing company, can position the fast guider assembly with RMS accuracy ~10 um. It displays negligible attitude-dependent flexure, across its whole range of travel. The precision in TRANSLATION is therefore as good as we could possibly need (±0."015). Because it is a precison machine we do not expect the slides to show any measureable (by us!) departure from orthogonality, but the accuracy of their N-S/E-W rotational orientation is not very well known.

The above figures imply a fractional uncertainty in the FPSC of ±0.0007, which corresponds to an angular positioning error on the sky of ±0."14 if a source is placed on axis by guiding on a star 200 arcsec off axis, at about the limit of the crosshead travel. This is close to what we would expect to be satisfactory for use with the present spectrometer CGS4, with its smallest slits (0."61 wide).

At present, therefore, only the uncertainty in the angular position of the crosshead is likely to be a source of problems. While we believe that these will not be very significant, it is probably best, if accuracies of <0."1 are desired, to use stars closer to the target than half the maximum range of the crosshead (±3.5 arcmin, ~200").

NB: Although measurements suggest that the fractional error in the FSPC is ±7/10,000 and we believe that the NSEW orientation error is small, thus far we have not measured the latter at all. It is at present wise to assume that the overall error is at least twice that given, i.e. at least 0.15%.

Back to Pointing.

CAVEAT: The perils of falling off the dichroic

It may quite easily be forgotten that the dichroic coating is carried on a glass substrate 10mm thick, which being inclined at 45 degrees has the effect of translating the transmitted beam sideways.

Thus the on-axis beam is moved 5."27±0."05 away from the instrument in use, relative to the off-substrate beam, and right at the edge of the substrate a double image can be formed. (The uncertainty results from uncertainties in the refraction coefficient of the glass and the small position-dependence of the displacement, which varies across the FOV.)

As seen from the N and S ports the dichroic spans the crosshead field-of-view in an E-W direction, so there are no restrictions on guide star positions in that direction. However about 130"±10" N or S (nominally 123" N and 140" S, from the IRCAM port, but not constant) the beam will fall off the dichroic and the shift will vanish. The absent shift is a fruitful source of "missed" guide stars and must always be borne in mind.

Recommendations

1. Offsets from guide stars to invisible spectroscopy targets need to be determined in advance to a precision of ~0."1 so high-quality local relative astrometry is required. Note that, e.g., VLA positions may not always be on the same system as optical positions; it is always preferable to measure positions from a plate or image showing both the intended guide star and the target source.
2. If your guide star is more than about 140 arcsec N or S (for N and S ports) of the target, remember to apply the "dichroic correction": the image will be 5."27 closer to your instrument relative to an on-axis image.
3. Do not rely on guide stars between ~115" and ~145" N or S of the source (with N and S ports) without checking them at the telescope, in case they fall on the edge of the dichroic.
4. The measured offsets should be applied to the CROSSHEAD, i.e. the crosshead should be used to translate the guider to the predetermined offset guide star position, so that when it is locked on to the guide star the source will be centered in the spectrometer slit.

Procedure for cross-head offsetting

1. Peak up carefully on a nearCMC star while guiding on it;
2. Slew to your target object using telescope offsetting (q.v.);
3. Apply to the crosshead the accurate offsets you measured from your (recent!) image, thereby moving the autoguider away from the the target and onto the guide star, with the full precison of the crosshead and your astrometry. The autoguider will then be almost aligned on the guide star (certainly well enough to capture it).
4. Start guiding on the selected guide star. The process of capturing the guide star will move the telscope a small amount which will place the target accurately in the instrument aperture (slit, row)

Unguided exposures

In RA the pseudo-sinusoidal encoder error causes an image on an unguided exposure to be elongated E-W, though its centre will be accurately positioned in RA.

In DEC we thus have the opposite situation to that in RA: a short unguided exposure will NOT be trailed in DEC, but its position will be uncertain by ~0."3.

All this is true modulo the effects of wind: the telescope drive system bandpass is not broad enough to correct for windshake vibrations. The DVS should be closed if there is any suspicion of windshake.