SUMMARY
SCUBA POINTING data are analysed with and without REMSKY, and a quantification of the benefits of REMSKY is attempted.

The Data
Some 7860 POINTING data taken with SCUBA since the beginning of 1998 are analysed. No attempt has been made to vet this sample for sources that might be unsuitable astronomically. It is assumed that the vast majority must have satisfied the astronomical needs of the programs for which they were taken. In any case it is the difference between the centroid determinations with and without REMSKY that is important to this analysis, not whether or not this will provide a fair description of telescope pointing at the time.

The Method
The POINTING data were requested from the JAC SCUBA archive , and they were reduced using the ORACDR pipeline. The method is described further here .

The results
The results comprise plots of the differences between the centroids derived with and without REMSKY. The plots below show the errors in azimuth and elevation plotted against the brightness of the source weighted by the number of integrations. The plots on the left show errors of as much as +10", while those on the right zoom in to the errors between +1" :

Error type	+ 10 arcsec	+ 1 arcsec
DAZ
DEL

As expected the errors increase exponentially for fainter sources and/or shorter integrations. There is some quantization at the 0.1" level of the elevation residuals generated by the centroid analysis - otherwise, the distributions of errors appear to be unbiased.

The vector errors ( = (daz² + del²)^0.5) are shown below :

	+ 15 arcsec	+ 2 arcsec
VECTOR

The 'fluxes' are roughly in Jy. The faintest of the blazars we use are quoted as roughly 0.3Jy (at 1.1mm, actually - so they'll be a little fainter at 850um), while 3c273 is 10Jy at 1.1mm and Uranus is approximately 60Jy at 850um. The brighter sources are usually well observed with N_integrations=2, and so the x-axis value of zero corresponds to a 1Jy source observed for N=2, or a 0.5Jy source observed for N=8 etc. Faint sources will require increased numbers of integrations to provide pointing accuracies similar to brighter sources (obviously !), and exactly how many more integrations are needed may now be judged from these plots.

The previous diagram is plotted again with a green line dividing, roughly equally, the number of points above and below it :

Some data lying particularly far above the line are more likely to be rejected by a critical observer, (and so the line represents only a guide to the average difference at any flux level), although others represent cases when REMSKY has provided exactly the improvement in analysis that is expected of it !

Conclusion
The following table shows, for N_integrations=2, the variation of desired accuracy with source strength independent of the activation status of REMSKY; i.e. this is a parameterization of the green line :

Recall :

• x = log((Source Strength/Jy)*(N_integrations/2)^0.5), so x=0 may correspond to (1Jy, N=2), or (0.5Jy, N=8) etc.

Acknowledgement
Thanks go to Tim Jenness for showing me both the method of extraction of data from the archive, and the flexibility of ORACDR to analyse them in a simple (single-UT) way, and to Remo Tilanus for help in overcoming some quirks in the database.