Last Update: JeffMangum - 08 Nov 2007
- Relative Pointing Accuracy: 0.6 arcsec within a radius of 2 degrees on the sky
- Absolute Pointing Accuracy: 2 arcsec all-sky
Calibration Plan: How Long, Anomalous Refraction, etc.
I was just reading in the pointing section.
It seems that each time we move antennas (ie, twice a week),
they will most likely need to have a global pointing determination
(unless we can predictably separate antenna-based and pad-based
I see only one mention of anomalous refraction in this section, and I
believe that this will, in the end, dominate our residual pointing errors.
In particular, the short time scale refractive pointing jitter may totally
invalidate the "local pointing model determination" -- ie, instead of
fitting to some physically meaningful trends in AZ and EL, you will be
fitting to noise jitter generated by the atmosphere and not repeated while
observing the target source.
We need to crank through the numbers for the WVR correction of pointing
errors to see what sort of pointing error is implied by the thermal noise
on the WVR. Can the WVR do the pointing error correction AND the phase
correction at the same time? Surely we are leaning heavily on the
success of the WVR at correcting for the anomalous refraction, and should
at the very least state as much -- this is a totally unproven concept.
In the "How Long" section, a single reference pointing takes
90 s but a local pointing model takes 60 s. Why the difference?
I did not see the 90 s number in memo 189 (that doesn't mean it
isn't there) -- but I am guessing that 90 s includes observing
at each half power point N times so the atmospheric pointing fluctuations
average down by sqrt(N). However, the 60 s time for obtaining a local pointing model
surely does not include N independent observations of each half power
point. Also, it seems that 60 s for 5 sources might be pushing it
(but I am guessing -- 12 s per pointing source, including slewing).
- 28 Sep 2004
Frontend Requirements and Pointing
Richard Hills notes that, while participating in the Optics CDR he noted quite a few items that seemed to be missing or incorrect in the current version of the frontend specifications
. One of those issues is the frontend contribution to pointing errors. There is apparently no reference to this in the specifications but it is obviously an important item, given the emphasis on pointing errors in the antenna requirements. It may well be that this has all been covered by detailed design and analysis in documents already (awaiting Frontend IPT comment on this).
The key point regarding frontend positional stability and pointing is that, at the focal plane, a shift in the position of the receiver beam relative to the telescope axis gives rise to pointing error on the sky. By contrast a tilt of the beam at this point causes an error in the illumination of the telescope and phase gradient across the far-field beam but no pointing error. We are at present putting a great deal of stress on the beam tilts but apparently ignoring the possibility of shifts, which are arguably more important.
In the top-level specifications the pointing errors are in fact described as the "antenna" pointing, but it is clear that this has to be a seen as a system requirement. In the antenna requirements the pointing errors are given at three levels:
- "Repeatable": 90 arcsec,
- "Absolute": 2 arcsec,
- "Offset": 0.6 arc sec.
The first of these is a peak error. The others refer to the rms radial error - i.e. both axes combined. Unfortunately the antenna designs have essentially no margins on these figures, so we don't really have any "spare" budget to assign to the frontend. To get an idea of the sort of level we should be looking at, let us suppose that we can allow errors due to the front end (i.e. from the mounting ring down) to be one quarter of those for the system. IF the errors can be added to those of the antenna in quadrature this produces only a 6% penalty overall. In fact, errors arising in the FE are likely to be highly correlated with those in the antenna, so this is not a safe assumption, but taking a quarter of the antenna numbers seems a reasonable starting point.
The relationship between shift and pointing is set by the focal length, which is 96m at the Cass focus. This gives the secondary "plate scale" of 0.465mm per arc second. The allowable shifts are therefore: Repeatable ~10mm, absolute, 0.23mm, and offset 0.07mm. If both axes are to be given the same error budget, the absolute and offset numbers come down to 0.16mm and 0.05mm in each direction.
In the case of the antenna, the repeatable term is mainly gravity and the intention is that this is taken out by the sin and cos of elevation terms in the pointing model. Obviously there will be some flexure of the FE as the system is tilted, but so long as this is elastic it will be taken out in the same way. The figure must surely be a lot smaller than 10mm anyway. Of much greater concern would be the differential movement of one receiver with respect to another (see below).
In the case of the antenna the absolute pointing errors are dominated by thermal effects. These are much less likely to be a problem for the frontend given that it is in a temperature-controlled environment. In view of the "drift" problems seen in the beam direction, the immediate concern here is similar behaviour in the beam position - i.e. that it moves around in a way that depends on the history of the elevation moves. Here a tolerance of ~0.16mm does not seem to be extremely tight for a system like this, but we should certainly have a look at any analysis that has been done and see what the measurements that are being made at present say about this.
As with the antennas, by far the hardest requirement is however the "offset" pointing. This is intended to apply to the case where we make the pointing measurement at one frequency and the astronomical observation at another. This will surely be a common
circumstance since most of the best objects for doing pointing - quasars - are only strong at the lower frequencies. We are therefore concerned about the relative movements between the beam positions of the different cartridges. Although the offset requirements on the antennas are only for movements of up to 2 degrees on the sky and 15 minutes of time, we need to know the relatively positions of the beams of the different
cartridges "absolutely". That is to say, we would expect to measure the pointing offsets for each receiver with respect to the reference one (presumably Band 3) once, presumably using a "broadband" object such as an asteroid, and then use these offsets at all elevations and for a long period of time. It seems to me that a tolerance of 50 microns on the stability of these beam separations might be quite hard to guarantee.
If we are absolutely stuck, then having offsets that are a function of just one variable (elevation) might just be tolerable. Essentially it would mean that one or two parameters in the pointing model would have to be changed depending on which receiver was being used. Offsets that depend on the pointing history or offsets which drift on time scales of hours or even weeks, would certainly not be acceptable.
Richard Hills - 11 Oct 2007