I want to look at the effect of dropping a bad point on a different SLM. I picked ME+2/SLM2 because it was handy. Unfortunately, I don't have a list of which laser is closer to which DCOPS in this case as I do with ME+1 (thanks to Xiaofeng there!). The aperture is wider for ME+2/3 than for ME+1, and wider still for ME+4.
I modified the SLM definition to include all the DCOPS/Laser combinations, and for those combinations where there was no data I gave the data point an error of a kilometer. In this particular event (MEp2_SLM2_3T_Aug29.dat, 23:27:16, 0T) the DLC_45 values lacked a good reading for L. I interpret this as meaning that the laser was far enough away from CCD2 that the fan did not reach it (a known effect, BTW). So that point was dropped automatically. The other side with the other laser is DLC_64.
I do not know a priori which is the bad point, so I ran three times: once with the original datafile, once with the R point given the 1km sigma, and once with the L point given the huge error instead. I'm only interested in the effect on Z for now, since the effect on X and Y is smaller.
From the improvement in chi squared I'd guess that the proper CCD to omit is L. The resulting shifts in Z can be read out down the column in the corresponding table of differences below. Notice that the change in chamber center can be over 300 microns.
The next question, of course, is about ME+4, where the aperture is larger and the effect should also be larger. ME+4/SLM1 is handy, and once again one of the CCDs is already excluded: For DLC_156 it is L. So I run the three cases where DLC_145 is original or has L or R excluded. The resulting chi-squares (44 "dof") are shown below:
It appears as though the best improvement comes from omitting L
Once again the shift can be larger than 300 microns, and could easily be larger still if the other chambers were present. Note that at the ME+4/1/10 end the L CCD is already omitted (presumably correctly).
This shift (338 microns) would represent a 500 micron shift in the reconstructed position of the reference sensor. If the laser head's fan at the opposite transfer point were similarly oriented, the result would be a mismeasurement shift of the entire SLM in Z by 500 microns. If it were oppositely oriented (and since the lasers are superficially symmetric it might be) the chambers would appear to be tilted by .07 mrad.
Which CCD is bad? Use U/D/L/R notation. A "-" means it doesn't seem to matter which is removed. The ME+1 information comes from Xiaofeng. The others come from Cocoa fit tests. The question mark in the ME+1 line is because although the position of the laser illuminating L/R is clear, I still need a little clarification on the position of the one illuminating U/D. I think that the bad one is D, but I want to make sure of this.
(*) L and R were very similar, both with a large change from the normal.
Modified 18-May-2009 at 12:32
|Previous notes||Next notes||Main slide directory|