ca-forum #22 minutes
Wednesday, August 13, 1997
- Reference Pointing problems (R. Sault)
- Polarisation leakage problems - pointing induced field rotation (R. Sault)
- Turret rotation problems (R. Sault)
- Amplitude closure problems (R. Sault)
- LO problems (M. Kesteven/G. Baines)
Dave Rayner had an observation corrupted by an aberrant reference pointing episode. The incorrect value was accepted, but unhappily the subsequent pointing updates were excluded having been judged unreasonably large. Several other observers report similar experiences.
Background : the reference pointing manual is available on the web. The machinery (CATAG) is also available. Briefly, a background task recognises that the observer has requested a reference pointing observation and processes the data to extract two pointing corrections (cross-elevation and elevation corrections) which will be applied to any subsequent "reference pointing" observation. The process attempts to filter out bad solutions; in particular, if the change since the previous check is large (> 20 arcsecs), and the previous check was close by in time (within a 2 hour window) and angle (20 degrees) then the solution will be rejected. Dave's bad luck was a bad initial solution at the start that was accepted, which meant that all subsequent (good) solutions were rejected.
Remedial actions
- Mark Wieringa will look to establishing a link between CAGET and assistance, so that assistance can alert the observer to rejected solutions, as well as to large jumps in the pointing solutions.
- The pointing parameters should be logged in the RPFITS file at the start of each scan. (This is relevant to the polarisation leakage problem as well).
- A task similar to LOPLOT should be available to monitor the pointing solutions on a regular basis.
- Mark Wieringa and John Reynolds will look to improving the robustness of the solution algorithm.
Dave Rayner's high precision circular polarisation experiment found time-variable polarisation leakage. This has been traced to the fact that we don't allow for the field rotation which accompanies the pointing model corrections. Ie, although the antennas all point in the same direction, their x-probes are not necesssarily parallel. An AT technical note, AT/39.3/071, is available as a postscript file.
- Bob Sault has provided a new Miriad task, TRANSFIX, which will correct the data using the pointing model that was in place during the observations.
- CAOBS will be modified to compute the field rotation for each antenna/visibility and have them written in the RPFITS file. The pointing model will also be logged. (M. Kesteven)
- ATLOD will be modified to apply the field rotations.
Dave Rayner's experiment also showed a step function in the polarisation leakage, consistent with a small turret rotation. The step occurred after two receiver changes (X->L; L->X). Subsequent experiments may have confirmed the diagnosis.
Graham Baines will organise some tests to check the mechanical stability of the turret. The basic question is whether the X-band receiver is always in the same location, relative to the vertex room floor irrespective of the elevation where the receiver was selected, (for all elevations).
Dave Rayner's experiment also revealed a problem with amplitude closure: Loops involving antenna 6 show an amplitude closure error of order 1%.
Bob Sault has extensive data on the problem. It appears to be independent of the source position in the beam; it is constant over time (12 hours to several months). It differs between XX and YY on frequency 1. Even though frequencies 1 and 2 were both set to 4.8 GHz, only the first frequency showed the effect. Interchanging the IFs (A C) at the correlator input showed that the effect was on the correlator side. Warwick Wilson and Bob have organised further tests.
During the 3 weeks of the CMB experiment LOPLOT was subject to some scrutiny. (The observers had little else to do). There is clear evidence that the round-trip correction is not complete - the astronomical phase contains a scaled-down version of LOPLOT. The scale factor differs from antenna to antenna.
On a number of occasions LOPLOT showed a jump in the round-trip phase for antennas 1,2 and 3. The astronomical phase (the calibrator) showed jumps on all baselines. If the round-trip phase machinery were operating correctly no jumps should be visible. At worst, a spike coincident with thejump might be expected. Further analysis showed that in fact the jump was of constant amplitude at all 3 antennas, but since the LOPLOT measurement differed from antenna to antenna, jumps were forced onto the astronomical data. This was confirmed by running CAOBS with the round-trip correction disabled. Careful analysis showed that the diurnal cycle evident in the LOPLOT data was also present in the astronomical data, although attenuated in amplitude.
The CMB configuration (all the antennas at the eastern end of the track) is notoriously bad. The behaviour is rather better in most other configurations.
- Running CAOBS with the correction ENABLED is still recommended - the correction is largely successful; it's just not quite exact.
- The jumps are a problem - Graham and Martin are continuing to address the issue. However, the observers should keep an eye on LOPLOT - and take action whenever a jump is observed.
CAOBS should log the round-trip phase corrections.
- The LO monitoring scheme that Derek implemented in VIS should be reactivated as a further diagnostic tool.