Bad weather kept us closed until about 02:00 MDT.

During preflight checks, we had a recurrence of the secondary E actuator failing to finish its 
move on a full step, so that the collimate process reported errors.  We homed the E actuator 
(the position error was 1), which solved the problem.  We moved the mirror arround at various
altitudes and could not reproduce the error.  The PR (2560) was updated.  

We opened under partly cloudy skies at ~02:00 and tested the pointing model installed during
shakedown.  We focused to 1.7" FWHM images with ~10 per cent asymmetry (11.7 pixels vs. 10.2 
pixels), which is fair under gusty east winds.  The out of focus images showed more signs of
a small collimation error but the cone fans were only turned on when we opened, so it is 
probable the primary was not yet in thermal equilibrium, and with variable seeing anyway, 
it was not worth trying to improve the collimation at this time.  

We observed a 20-point grid (TGRID:GRID_5_4_25_80_T_30.DAT).  It was clear during the 
observations that the errors in target position in the engcam images were large.  When we
used tpoint to calculate the mean error between the observed and fitted positions using the model
obtained on 2001/08/06 (tdat:telmod.dat;20) we obtained an rms error of over 10 arcsec, with all 
the error vectors pointing east-west.  We confirmed that the observations from the previous model
had be fitted correctly by re-fitting them in tpoint, and indeed found that it gave a 1.7 arcsec
rms fit to those observations.  

When we allowed all parameters to vary with out new 20-point observations, we got a 2.6 arcsec 
rms error in the new fit.  The largest change in parameters from the old fit was in AW (azimuth 
axis west tilt), which moved closer to zero but with opposite sign:

       coeff       change     value    sigma

  1     IE         -0.119   -185.37    2.759
  2     IA         -8.339   -268.02    2.566
  3     AN         -0.432     -8.97    0.894
  4     AW        +12.095     +4.98    0.825
  5     HWCA2      +2.323     +2.54    1.164
  6     HWSA2      +1.133     +1.20    1.270
  7     HNCA1      +1.651     +0.92    1.471
  8     HNSA1      -1.821     -2.67    1.429
  9     CA         +4.141    +21.97    1.878
 10     PZZ1       +5.212   -121.24    7.593
 11     ACEC       +1.956     +4.86    1.514
 12     ACES       +0.301     -2.00    1.487
 13     TX         -2.531     -8.31    3.258

Sky RMS =   2.58
Popn SD =   4.59

We found this very surprising: what changed in the six days since the last model?  We repeated
the observations again, as we were worried we had made a mistake in rotator mode or in clearing
initial offsets, but obtained almost identical results.  So, we implemented our new (sparse
20 point) model, then observed a 25-point grid (TGRID:GRID_6_4_27_87_T_0.DAT).  It was
immediately obvious that the new model was giving much smaller errors.  We calculated the
errors between these observations and the 20-point model and got an rms error of 2.7 arcsec, 
thus confirming the new model produces repeatably small pointing errors.  

If anyone can explain why the pointing error should change so much in a few days, please do so!

We closed at 06:00.  We do not think we lost science time tonight, as at no time were the
skies clear enough for useful spectroscopy.