SUMMARY
These analyses are part of the continued effort to reveal the extent and size of the elevation pointing errors induced by the reversal of direction in elevation that becomes so apparent during transit tracking (see the reports during April 2000 ).

• E-W differences in pointing data are revealed

• A simple correction for E-W differences allows a correction to parameter #2 of the telescope model.

E-W pointing differences
The difference between elevation pointing in the East and West is determined in the simplest way by analysis of all (SCUBA) pointing datasets taken during 2000 so far. Pointing data are assigned to 'East' and 'West' subsets on the basis of the azimuth. The mean residuals in azimuth and elevation in each subset are calculated and the differences (East-West) derived. The means and standard deviations of all such differences are tabulated below. Since datasets with only a few logged points might produce unrepresentative results, the calculations are repeated for datasets of various size. The columns in the table show :

1. the minimum number of data per dataset
2. the number of year 2000 pointing datasets involved
3. the mean of the (E-W) azimuth residuals
4. the rms of the (E-W) azimuth residuals
5. the mean of the (E-W) elevation residuals
6. the rms of the (E-W) elevation residuals

The azimuth residuals serve as a control of sorts on any result relating to the elevation residuals. The results involving small datasets show a bias in both residuals towards negative values. However, as the size of the dataset is increased, and as the (E-W) difference becomes less sensitive to rogue data, the azimuth difference trends towards (but does not reach !?) zero, which we expect as the default behaviour. At the same time, the elevation residual difference increases to -1" with some significance.

Application of simple correction to pointing
A similar analysis of the last dedicated pointing run, of 19991115, gave similar results and the e-mail message of 24 April 2000 relating the correction of a simple E-W difference is reproduced below :

I took the last dedicated pointing run, of 15 Nov 1999 :
   N=50 daz = 0.0 +- 1.4", del = -0.5 +- 2.0"
When run through FIT9 a new model with expected performance of
              1.2"           1.7" resulted
after changing the 7 parameters in the model by the following amounts :
   p1 (the N-S error in the direction of the main axis ) -0.9"
   p2 (the E-W ....................................... ) -0.7
   p3 (the non-orthogonality of the az & el axes       )  7.9
   p4 (the azimuth collimation                         ) 11.0
   p5 (the azimuth encoder zero point                  )  8.6
   p6 (the elevation encoder zero point                ) -0.9
   p7 (the flexure in the tetrapod                     )  0.5 

I used TPOINT to plot del vs both azimuth and hour angle but could not,
on the basis of this small (!?) dataset, choose which pattern
of residuals looked the most systematic.

Splitting the raw data in two by the meridian :

  29 points in the East had  del = -1.2 +- 1.9"
  21 ............. West ...         0.5 +- 1.6

I corrected all the 'Western' data by -1.7" giving
                             del = -1.2 +- 1.8" for all 50.

 - i.e. I treat the 'transit' phenomenon as a step function, although
it would be better to use an analytic form that has a smooth transition
for 5 or 10 degrees on either side together with any appropriate 
trigonometric functions - elucidating this functional form may  
require superimposing hundreds of data . . .

Running these data through the FIT9 program gave suggested changes to
the 7 parameters of 

   p1 (the N-S error in the direction of the main axis ) -0.9"  cf -0.9
   p2 (the E-W ....................................... )  0.0      -0.7
   p3 (the non-orthogonality of the az & el axes       )  8.5       7.9
   p4 (the azimuth collimation                         ) 11.8      11.0
   p5 (the azimuth encoder zero point                  )  9.1       8.6
   p6 (the elevation encoder zero point                ) -1.4      -0.9
   p7 (the flexure in the tetrapod                     )  0.2       0.5

The most significant change is the +0.7" in p2 (as expected).
This is roughly half of the step size (1.7"), (as expected),
and is in the correct sense - believe me, ask me to explain, or 
draw the diagram yourself :)

[ The changes to p3,p4 & p5 are not unexpected - these 3 parameters
  are a veritable tagteam and play off violently against each other at the
  slightest disturbance. The change to p6 is completely understandable, 
  that to p7 less so, but it is small enough. The zero change to p1
  is most encouraging.]

The new expected performance (rms scatters) in daz,del are  1.1" & 1.6".

Conclusions
-----------
1. The correction, simplified to a step function, generates the 
   adjustment to p2 expected theoretically (a la Per, 14 Apr 2000).

2. The overall improvement to the model is small (but desirable).

3. Attempts to refine the step amplitude and a search for an
   analytic form of the error continue. The error *could* 
   be implemented within the VAX telescope software as a lookup table.
   (although a continuous function is needed to provide accuracy during
    transit).