SUMMARY
Three further tests were carried out of the 's-correction' - the lookup table that corrects the elevation pointing across transit. The results are still frustratingly inconclusive, and a redetermination of the underlying error function is suggested.

The 'transit' problem was noticed first in April . A correction was installed on 16 May, but recent checks in May and June have been inconclusive for various reasons.

The new tracking tests were done during poor weather over the past weekend, using the SCUBA map16 method :

UT Date	20000702	20000702	20000703
source	3c279	Uranus	Uranus
HST start	18:13	02:00	01:51
HST end	19:21	04:16	04:10
elevation at transit	64o	55o	55o
typical seeing	1.0"	0.2"	0.2"
Pointing residuals

The elevation pointing residuals from the run on 3c279 suggested that the s-correction term was perhaps deficient in accounting for the elevation errors properly, with the observed residuals showing a possible 2" residual s-function. However, the odd behaviour of the azimuth residuals at azimuth 195 suggest other possible problems, perhaps with the stability of the atmosphere in the early evening.

Thus, a second experiment was performed that night, on Uranus. The Uranus residuals are notable for their good azimuth performance (rms errors of 0.6" in more than 2 hours of tracking) and the strong elevation systematic, of amplitude 4", but lying asymmetrically around transit. The poor SCUBA noise performance at this time is noted : it might have invalidated data of a fainter source, but the brightness of Uranus must surely be enough to overwhelm the extra bolometer noise.

Nonetheless, the experiment was repeated on Uranus the following evening - when the noise on the SCUBA LW array had abated - with essentially the same results. Both the Uranus results, though, are quite different in character from the 3c279 results, and an explanation has not yet presented itself.

Had we only these two self-consistent Uranus runs, my inclination would be to repeat the experiment without the s-correction activated in order to re-establish the underlying error function. But how do we explain the difference between the 3c279 data and those on Uranus ?