HARP has been used recently during regular observing (although it is not commissioned). Three nights' data are displayed below with no attempt to remove nightly collimation offsets:

The data separate into two (or 3 !?) blocks in several of the sub-displays:

• the centre-left plot of dS-vs-Z shows ~-25arcseconds separating 15 measures from the majority, while
• the Up-Down-Left-Right plot (lower left) shows 3 groups:
  • the aforementioned '15' loosely clustered on the right
  • 4 measures at high-elevation, transiting in the north, and
  • the bulk remainder
  They are separated by about 25" and roughly occupy the vertices of an approximately equilateral triangle.

The '15' measures come in two time-contiguous groups; measures 2 thru 9 on UT20061215, and measures 1-7 on UT20061217. Log entries preceding entry #9 on UT15 suggest 'kmirror engineering' (although 'exercising' may be a better description), but there is no commentary in the logs to explain the pointing jump on the 17th.

However, the k-mirror values themselves follow this pattern :

• - the bulk of these data were taken with -ve k-mirror values
• - the '15' are the only measures taken with +ve k-mirror values
• - three of the 'high northern four' have extreme -ve k-mirror values ( < -50 )

The '15' cover the same range of parallactic angle, Q, as the 'bulk', whereas 3 of the '4' are at uniquely +ve Q>150.

Dividing the data into these groups for analysis via TPOINT meets the same problems described earlier; with a full 7-parameter optimization leaving residuals of ~(4",4") in (dS,dZ) indicating, as before, and at the very least, the need for other terms.

20061220
These data don't look nearly so potent when plotted in what I consider a more raw form :

There are clear systematic errors (requiring adjustment to the mode parameters), and possible large-ish residual scatter (requiring new parameters, as previously speculated), but there's no sign here of the groupings seen above. In particular: compare the middle sub-plot here (del-vs-El) with what ought to be its mirror image above (dZ-vs-Z, also the middle sub-plot).