Prelude
The pointing model software is now in a new location :
          subdirectories of /jac_sw/itsroot/src/tcs
The track model was not that 'installed' on 20 May 2001, but the errors incurred by having the previous model still active are likely to be small. The pointing model used tonight was that installed on 18 May 2001 by NPR.

Summary

• A transit tracking experiment on CenA was performed to monitor the size of the transit elevation pointing step. Tonight's value was ~5" (full step size).

• Allsky pointing was done. Large azimuth & elevation systematics are seen throughout. A new model was derived and installed. It resembles version;-2 more than version;-1. Either the intermediate model was unnecessary, or the need for a completely new model for the second time in a week is disturbing.

• Detailed analysis reveals a possible algorithmic solution to both allsky pointing and transit tracking.

Transit tracking of CenA was performed :

which suggest a (full) step size of about 5"*cos(el*). This is to be compared with 6.7" on 20010517 and 7.5" on 20010501. The variation may reflect noise or a real change - we must continue to monitor this.

Allsky pointing was also done :

It includes many points on IRC+10216 between azimuths 145 and 205. These were taken to provide good coverage of this critical region, but it is also possible that they bias the resulting fitting procedure. The IRC+10216 data are shown below :

The elevation residuals show no sign of a step per se (as expected for such a high transit el*evation), while the azimuth residuals show much unexplained structure.

The entirety of data were run through TPOINT. The display below is the FIT9 result, which for all purposes is the same as the TPOINT result :

There are still worrying non-random features in the plots of daz & del -vs- azimuth (top-left & center-left). The fitted scatters (2.5", 2.1") (which equates with the 3.4" total from TPOINT) are also much larger than our expected (1.5",1.5"), so this is by no means the end of the story.

The new model was installed at 01:00 HST 23 May 2001.

So far, no account has been taken of the particular elevation problem (the 'step') at transit. Dividing the data now into East and West, we find :
          East : daz = 0.1 +- 2.9 , del = -1.5 +- 1.6
          West : daz =-0.1 +- 1.9 , del = 1.6 +- 1.3,
so our canonical elevation pointing accuracy appears potentially recoverable given appropriate E-W differentiation.

Further, below is the middle of the 9 subplots above (ie after the model change) showing the elevation residuals plotted against elevation :

The increased scatter at low elevations may be ascribed to increased measurement noise . . . but, see below how the data are divided between those points in the east (white) and those in the west (red) :

Negating the western data and folding E & W together yields the following plot of elevation residuals -vs- cos(el), below left :

The slope of 4.5" +- 0.8" is the half-step size of the effect, but is not too disimilar to that seen in the original transit step data, above right (full step = 7.5").

Depending upon which dataset you like best, it would appear that correcting all pointing/tracking by approximately +4"*cos(el) in the east and -4"*cos(el) in the west would resolve our current elevation pointing problems.

In the future we must continue to monitor the transit step size by doing tracking of low, bright objects like CenA (13^h22, -43^o) and G343.0 (16^h58, -43^o). These tracking data will provide the coefficent of the relationship above : they are a faster and more direct measure than the allsky pointing provide. However, since we must, at least for now, be aware of possible evolution of this effect, I suggest we adopt 4.0" as being valid currently.

The algorithm
The algorithm reflects the gross +4"*cos(el) described above, plus fine details at transit. These latter are derived from the many transit tracking experiments performed over the past month - particularly those done on 01 May 2001. They generally show narrow (1.0^o full-width) sine curves, located essentially at precisely 180^o (360^o). The parameters of the curves as plotted in these pages recently show centres displaced from these precise values, but they do not take into account the time lag between the mid-point of each integration and the logging of the data. Correcting for these leaves the curves centred essentially at transit.

The total correction to the elevation pointing is then

   for     0 < az < 179.5    del = +COEFF*cos(el)
   for 179.5 < az < 180.5    del = +COEFF*cos(el) * sin 90*(180-az)/0.5)
   for 180.5 < az < 359.5    del = -COEFF*cos(el)
   for 359.5 < az < 360.5    del = +COEFF*cos(el) * sin 90*(az-360)/0.5)
   for 360.5 < az < 450.0    del = +COEFF*cos(el)