Improving the telescope - a risk we have to take

Last year we undertook several major upgrades of the telescope surface and the chopping secondary in order to improve the performance of the telescope. Any work on a telescope which is in constant use, always involves a risk. We try to minimise the downtime of the telescope. We cannot build in large safety margins if things go wrong, or if the weather is poor during the few recovery nights that we believe should suffice.

The second stage of the focal length change was performed in late January, and we had to shim a relatively large number of adjusters to get the panels into the right position. Although the mechanical operation went rather smoothly, we were plagued with bad weather, shaky adjuster electronics and problems with the holography receiver. Neither did we have well-placed planets for astronomical testing. It therefore took us into mid-March before we had the surface back to decent, although not perfect shape. Tests done later during the spring indicated that the focal length change was very successful in reducing the diffraction ring to an acceptable level. Currently there are no plans to do the final step (another 8mm), because the gain that could be achieved is too small compared to the efforts it would take.

Figure1. Beam maps of Mars at 800 microns.

(a) from July 29

(b) from September 8

(c) from October 11

The contour levels in each map start at 2% of the peak intensity with a step of 2% and after we reached the peak intensity in the error beam we continue with steps of 10%. In a) this occurs at 10%, in b) at 16% in c) at 22%, i.e. the surface appears progressively poorer.

During the spring we finally managed to diagnose the cause for the lack of homology of the telescope (the change in gain as a function of elevation, which had plagued us for several years). Five of the twelve conebars, which support the backup structure of the telescope, showed large movements in excess of 20 microns when the telescope was tipped in elevation. They are not supposed to move. After some testing in May of how to cure the problem, we finally decided to weld all conebars and scheduled the work for early July.

During a short heavy engineering period in July (July 7-8), the conebars were wedged and welded to eliminate the flexing in the backup structure. This clearly deformed the antenna. An in-focus beam map at 1.1 mm taken immediately afterwards showed a clear three-lobe error structure, which looked very similar to beam maps taken during the autumn of 1991, when the lack of homology first became apparent. Due to problems with the holography receiver, we had quite a bit of struggle before we managed to recover the dish. The holography maps looked rather good before the last adjustment, which was made on July 12. After this adjustment we could not get the holography receiver to work, but the adjustment went smoothly and should not have caused any problem. The weather during the whole period was too poor to permit observations at wavelengths shorter than 1.1mm, but out-of-focus (oof) beam maps made with RxA2 on Venus looked very similar to the holography maps. We therefore believed that we had cured the homology problem and successfully set the telescope surface.

Figure 2. Beam maps at 450 microns.

(a) Beam map of Uranus obtained in August

(b) Beam map of Mars obtained by Ned Ladd in October

In both maps the contour levels start at 10% of the peak intensity with steps of 10%. Note how poorly defined and elliptical the main beam is in October.

During an EAC shift on July 29th with marginal sub-mm weather we obtained several in focus beam maps both with UKT14 and RxA2. The beam maps showed surprisingly large error lobes (Fig 1a). Because the holography receiver was still broken, the decision was taken to wait until we got the holography receiver fixed (projected as early October). No reports from visiting astronomers reported any problems with telescope performance until after the heavy engineering period last week of September.

The heavy engineering in September, where the SMU was taken off and new stiffer flex-pivots, new LVDTs and the digital controller put back on- line, was finished late evening Sept. 30, 1994 (Friday). Testing done on Friday and Saturday night indicated that the chopper was working fine apart from some initial problems the first night which were fixed the following day. We checked pointing and focus (the 8 mm shim was also removed from the secondary), re-measured the chop scales and they appeared reasonable. The sky conditions were marginal and as in July we could not work shortward of 1.1 mm. Tilt tests of the secondary, however, indicated that there might be a tilt in the secondary and further tests were planned for the EAC shift on Saturday, October 8.

Since the sky conditions improved on October 2 and continued to stay good, PATT observers were asked to take beam maps at 800 and 450 microns so that we could compare the shape of the beam with earlier data. Tuesday evening (Welsh on 1st shift) was poor and no beam maps could be obtained, but improved radically on second shift and Ned Ladd took two large beam maps at 450 microns (Fig 2), showing an incredibly poor beam!! The Director (DDT, Sunday eve, October 2) also reported drop in UKT14 sensitivities at 1.1 mm and 800 microns, and so did Ned Ladd later in the week.

Figure 3.

Beam map of Mars obtained with RxC2 in December 1994

The contour levels are 1%, 2%, 3% and 10% and continue in steps of 10%. Note how clean the beam appears at 495 GHz.

At this stage we dropped everything we were working on and started to investigate what could possibly have gone wrong. A fair amount of telescope time was used for testing, but most of it was either done during daytime or using service and DDT time, which could be paid back later. We did use regularly scheduled PATT time as well, and Matt Griffin, who had a long run in mid-October gave us a fair amount (~1.5 shifts) for carrying out various tests. In the beginning a lot of the work was concentrated on the secondary, because this was the logical point to start at. We spent quite some time checking the tilt of the secondary, but this was just a red herring. Even though tilting the secondary can remove some of the telescope errors, re-focusing the telescope will effectively cancel out the tilt. However, we could not really find anything obviously wrong with the secondary. The next step was to eliminate UKT14, because it appeared to show a poorer performance than our heterodyne receivers. We therefore checked the alignment of UKT14, and in the end we decided to take it off the telescope and open it up, but in the event nothing amiss was found. Neither could we find anything wrong with the conebar joints or the telescope backup structure. However, we got more and more evidence suggesting that either the surface or the secondary had deteriorated.

Our tests showed conclusively that the performance of the telescope was poorer than we have ever had it before. Beam maps showed large error-beams and at 450/350 microns, the beam was largely elliptical, often split into two peaks. Although it is clear that the telescope performance was never fully recovered after the welding of the conebars in July (because the holography receiver broke down), analysis of the available data prior to the heavy engineering period in September appear to show that the telescope performance was satisfactory, although far from good.

Figure 4. Beam maps of Mars deconvolved with DBMEM.

(a) The 800 micron map is an average of three beam maps obtained, two obtained in the December 1994 and one in January 1995

The size of the map is 95" x 95" and contour levels are 0.5%, 1%, 2%, and 10%. All contours thereafter are in steps of 10%

(b) The 450 micron beam maps is an average of two maps of Mars obtained during the same night in December, but with two different chop throws: 20" and 25"

The size of the map is 40" x 40" and contour levels are the same as for the 800 micron map.

However, after the September heavy engineering period the performance appeared much poorer, but this could partially be due to the fact that we had no decent sub-mm weather before October. The prevalent view is that the cause for this dramatic decrease of telescope performance was due to deformation of the chopping secondary, which could have somehow been damaged during the heavy engineering. As we later found out when we started to investigate this matter in detail, the secondary could have been damaged during the crash tests, that were made on the chopper. The main support for this hypothesis is in the apparent large scale errors of the telescope deduced from oof-maps and holography, which show a north-south squint of the telescope. What somewhat contradicts this view is that the in-focus beam maps show the same error beam pattern (strong elliptical error lobe mostly oriented east-west) from late July onwards, only much more amplified in October (Fig 1c & 2b). If the poor performance is due to the secondary, it must already have had some damage before the heavy engineering period. An alternative hypothesis (considered mechanically less likely) is that the telescope surface was stressed by wedging and welding the telescope in July. The stress put into the dish by the conebars was gradually released and redistributed over the dish. In this scenario it is the relaxation of the dish, which eventually made the surface almost unusable for sub-mm work, but this does not adequately explain the squinting of the dish seen in the holography maps.

After solidly verifying that the problem was in the telescope surface or deformation of the secondary, we decided to take out the errors by adjusting the primary. This was done in November (Nov. 3 - Nov. 6, 1994), and fully recovered the performance of the telescope. The telescope is now better than we ever had it before. Extensive testing done in December 1994 and January 1995 shows that the homology problem has been cured by welding the conebars.

We now have the best surface we ever had on JCMT. With relatively small and rather safe efforts we can improve it even further. Holography, oof-maps and aperture efficiency measurements predict a total surface error about 30 microns. Extensive sets of beam maps show very clean and symmetric beams. At 490 GHz we can hardly see any error lobes at all (Fig 3) and beam maps of Mars at 800 microns and 450 microns (deconvolved with DBMEM) show that the error beam adds about 10 – 15% of power at 800 microns and 20 – 35 % at 450 microns (Fig 4a,b). We now have the holography receiver back in operation and most of the problems with the adjuster electronics have been solved as well. We have set up a regular monitoring of the dish. No further changes have been seen since our last adjustment in early November.

We will continue our work on improving the telescope, with the aim of minimising any problems that it may cause our users.