Report to the JCMT Board; November 2007

Introduction
This report is the first since April 2006. Following 6 months of engineering in 2006 - preparing the telescope to accept SCUBA-2 - operations resumed with many new components, including HARP, ACSIS and a new operating system. One of the major challenges in this new regime was the optimization of HARP pointing - a task that is still on-going.

Pointing with Receiver A
In the absence of SCUBA, Receiver-A has become our receiver of choice for pointing, since, unlike the antenna structure and the telescope control software, it remained unchanged through the engineering. Changes to the telescope might reasonably be expected to change the telescope pointing model, and we needed to re-establish satisfactory pointing with RxA before trying to do likewise for the more complicated, new (or newly-refurbished) receivers, like HARP and RxW.

Prior to the shutdown rms residuals in (azimuth,elevation) were of order (1.5",1.5"). The first pointing runs, in August 2006 were poor but not awful, with RMS residuals in (daz,del) of about (3",3"). An immediate improvement, to (2.3",2.3") was achieved at the start of September, but bad weather was blamed for a disturbing increase in the elevation residuals (to 4") through the middle of September. The performance settled back to being moderately poor (2"-3") through October, but became unsettlingly bad again as measured in a dedicated pointing run on 28 October - (daz,del)=(4",3"). Several subsequent updates to the model occurred - and some improvement was obtained - but always there was the suspicion that the weather was somehow never good enough. (Hmmm - shurely not !) By 07 November residuals were again in the 2.0"-2.5" range, but residuals during dedicated pointing runs in December and January 2007 increased to more than 3".

It was at about this time, that problems with the secondary mirror unit (SMU) and ACSIS continuum levelling became apparent. Pointings done as fully-sampled continuum or spectral-line maps were recognized as being vulnerable to a multiplicity of problems and so were thenceforth restricted to spectral-line FIVEPOINTs, with the multiple analysis options also reduced to one (spectral-line, gaussian). Subsequent pointing runs in February yielded RMS residuals of less than 2" for the first time, and the RxA model created on the basis of data obtained on 12 April has remained in place since then, with consistently good performance, similar to our advertized performance prior to the shutdown. (Phew !)

Pointing with HARP
HARP is a B-band array, sitting at one of the Nasmyth foci. It would be no surprise to discover that the pointing model suitable for HARP needs, amongst other things, a different flexure term to that for RxA - which is in the cassegrain cabin. Similar concerns in 2005 about RxA and SCUBA (at the other Nasmyth focal station) eventually proved unfounded, but HARP presents other complications, due mostly to the field rotator (the 'K-mirror') that sits between the tertiary mirror unit (TMU) and the right Nasmyth focus. Its construction and placement, like other aspects of the telescope, will not be perfect, and so its misalignments - with respect to the elevation axis or optical axis and with respect to the centre of the HARP array - must be parameterized and corrections applied.

Per Friberg has described how the pointing errors caused by these imperfections are related to the orientation of the antenna (elevation, E) and field rotator (K-mirror angle, K):

    dAz =  - xo * sin(2K)            dEl = + xo * cos(2K)
           - yo * cos(2K)                  - yo * sin(2K)
           + x' * cos(E-2K)                - x' * sin(E-2K)
           + y' * sin(E-2K)                + y' * cos(E-2K)

where

• (xo,yo) are the alignment errors of the rotation axis of the K-mirror with respect to the elevation axis

  and

• (x',y') are the alignment errors of the center of the HARP array with respect to the elevation axis

HARP's 16-element footprint on the sky needs to be 'jiggled' in order to produce a fully sampled image of the sky. It was understandably tempting to do pointing with HARP by creating such jiggled images and analyzing them for their bright (continuum) centroids. This mapping technique had been used for early RxA POINTINGs, too, with the resultant problems described above. Since early 2007, we have restricted pointing with HARP also to the traditional (spectral-fivepoint) method, using one of HARP's four central receptors (H10, a.k.a.R11) as a 'pointing' receptor.

Determining the K-mirror coefficients
The relevant data are available mainly in two forms:

1. regular AZ/EL FIVEPOINTs, where (E-2K) is either 0 or 90 degrees. The x',y' terms then contribute only to the constant offsets, but the great numbers of allsky AZ/EL pointings should still prove useful in determining (xo,yo) - eg the 108 screened data of 20070830. FIVEPOINTs done in the RA/DEC frame exercise the K-mirror more but have not been done in great numbers.
2. Arc-data, where the K-mirror is forced to a series of values in quick succession while the elevation is essentially constant. The errors in the K-mirror alignment cause the subsequent images to move around on the HARP array in an arc. The figure below shows such data taken at an elevation of approximately 32 degrees:
   

   The black symbols and lines show the azimuth and elevation residuals of 7 'arc' data taken with K values vaying from about -50^o to +50^o. The best fitting arc is shown in red, and the K-values at multiples of 30degrees are shown by red squares, with two values (-30 and +30) labelled. The green vectors show the differences between each datum and the arc at the value of K pertaining.

Several (7) such arc datasets are now available and together yield the optimal parameter set

            (xo,yo,x',y')=(+0.3",+0.7",+3.4",-0.1").

This global solution successfully fits some of the arc data - like the set above, with residuals of approx (1",1") - though not all, and not the latest two (October 05). It is possible that a change to the orientation of one of the fore-optics mirrors (M5) in September may require further data and a new solution. Also, theoretical application of the various K-mirror models, including the latest above, to all-sky pointing does not appear to improve all-sky residuals - which are currently of order (2.5",3.5"). There is scope throughout for misunderstanding of sign conventions in the application of such corrections, but efforts are on-going to establish such conventions beyond doubt.

However, there is reason for optimism that a K-mirror model is not far away and that it will restore pointing accuracy with HARP to something similar to that currently achieved with RxA.

In the future, when a valid K-mirror model has been determined and installed, local pointing (i.e. prior to a science observation) will be done in an RA/DEC frame, so that the K-mirror is in essentially the correct orientation for the subsequent science observation. Although current local AZ/EL pointings essentially nullify any remaining systematic errors in the basic (RxA) pointing model, they do not do likewise for the systematics due to the K-mirror. Thanks to the initial, precise installation, these are fortunately small, and our current operation - with (xo,yo,x',y')=(0,0,0,0) - is unlikely to generate large pointing or registration errors.

Inclinometry
Measurements of the antenna track profile are made every couple of weeks or so, and show only minor changes from one run to the next. Even the 6.7mag earthquake of 15 October caused only a 2" tilt in the overall plane of the antenna track, an amount small enough to be accounted for purely in software.

Transit step and central bearing race defect
Characterization of both these effects demands the highest possible resolution in azimuth and time, previously provided best by SCUBA in jiggle-map (continuum) mode. The race defect has been coarsely measured since the engineering using a HARP mapping routine; it is no worse than before - i.e. being of amplitude no larger than 3" at the horizon -- the effect varies as cosine(elevation) -- but it needs repeating at higher resolution and S/N. The transit step is yet to measured since the shutdown.

Temperature corrections
Algorithms that correct the pointing and focus for changes in the ambient or structural temperatures have been checked a couple of times this year since the RxA model settled down and no significant and consistent changes are warranted.