The MAB DCOPS should be useful for fitting where the MABs are. Here I compare the original input (from Zoltan's PG survey) of where the centers are and compare it with the fit positions. Note that Line 2 is bad (at the top) and line 5 is poor (at the bottom).
If I rotate to the transfer line system angles and exclude the bad lines, the result still looks very odd. Some of the shifts are quite large.
I do not understand why the last iteration of printout of real vs simu hits shows such wild variation, when the fit values seem to agree well.
Or perrhaps I should say "better" if not well. Something is still strange. The MAB fits should only be constrained by the DCOPS, and therefore should land right on the nose. So why don't they? What makes Cocoa pull them some distance away from exact? The hand fit pulls things quite close, why not this?
Do observe though that the variations here are an order of magnitude smaller than those seen in the REAL/SIMU dump.
If I look at lines fit through the endpoints on Transfer Line 1, I can assume that they are driven by the PG displacements of the transfer plates, take the Cocoa fit results, or use the hand fit results. Cocoa agrees better, but not perfectly, with hand fit results. The disks are not allowed to move in the Cocoa fit, so the result should be entirely focussed on the transfer plate positions. The rear disks are set at survey, and the rear transfer plates are fixed at survey.
The only remaining degree of freedom I can think of at this point is the transfer plate rotation. If I look at the points that appear on both the transfer plate bench survey (round 2, with large targets) and the disk survey, for ME-4/TP1 I find that the DiskSurvey measurements are related to the Bench measurements using.
Offset= [484.018;1133.34;-9.74] Bench = Offset + Rot(0,180,0) * (1/R_MEp4_tp1)*((1/R_slm_p41)*(DiskSurvey-X_slm_p41) - X_MEp4_tp1)
The resulting differences (in Bench coordinate system) are shown below. Notice that survey z differs from bench calibration Z by up to a mm. Point 4101 is near the Transfer Line DCOPS,, while 4103 is halfway down the plate and 4107/4108/4109 are on the SLM DCOPS (and therefore provide the reference center). Their average deviation is +0.152. The deviation across the distance from one to the other is 1.3mm. The distance between 24101 and 24108 (about the center of the 3 DCOPS points) in X is 504.84-10.01=494.83, thus representing a rotation of 2.6mrad or 0.15 degrees. The error on the bench measurement I take to be of order 50 microns (looking at reproducibility), and that of the disk survey is supposed to be of order 300 microns, so the angle measurement has an error of .03 degrees.
This isn't a huge rotation, but it would be adequate (by construction) to give us an extra 1.3mm, and that in a situation where the difference between measurements has an RMS of about 1.6mm as shown in the histogramm above. We may have a smoking gun here.
|Point||X diff(mm)||Y diff (mm)||Z diff (mm)|
Unfortunately not all points are measured in both surveys, so there's a certain amount of hand checking to make the tables.
Raw data used: Bench first
First_24101=[10.01;-107.97;4.97] First_24102=[224.98;-18.52;5.00] First_24103=[424.92;-59.03;4.98] First_24104=[113.79;-133.80;99.97] First_24105=[84.95;-157.89;100.04] First_24106=[55.94;-133.98;99.90] First_24107=[475.94;-129.39;-9.55] First_24108=[504.84;-129.29;14.48] First_24109=[533.77;-129.32;-9.52] First_24110=[164.94;-99.94;37.94] Second_24101=[9.67;-107.97;22.45] Second_24102=[224.46;-18.66;21.79] Second_24103=[425.23;-59.26;21.31] Second_24104=[113.72;-133.77;99.98] Second_24105=[84.87;-157.86;100.05] Second_24106=[55.85;-133.94;99.91] Second_24107=[475.94;-129.50;-9.57] Second_24108=[504.84;-129.39;14.47] Second_24109=[533.78;-129.43;-9.53] Second_24110=[164.89;-99.91;37.92] Second_24111=[467.58;-45.19;90.13] Second_24112=[503.11;-45.18;54.93] Second_24113=[538.70;-45.17;87.17] Second_24114=[493.06;-128.23;-31.09]
And the disk survey info:
Disk_YEM3_4101=[7014.45;1675.77;-652.48] Disk_YEM3_4102=[6957.62;1882.78;-563.00] Disk_YEM3_4103=[6904.71;2076.51;-603.60] Disk_YEM3_4107=[6861.75;2117.31;-673.89] Disk_YEM3_4108=[6877.35;2151.60;-673.72] Disk_YEM3_4109=[6846.64;2173.14;-673.82] Disk_YEM3_4111=[6959.98;2135.42;-589.34] Disk_YEM3_4112=[6916.69;2160.53;-589.50] Disk_YEM3_4113=[6938.40;2203.35;-589.40] Disk_YEM3_4114=[6836.40;2128.29;-672.58]
It looks as though there are two types of vertical target in ME-4 calibration, one of which is nominally 20mm and the other of which is something else: either 16.5 or 21.5. It isn't clear from UR-0007 which it is, because the target isn't listed beside each measurement. At a wild guess I'd say 21.5, since that is the value (average) given for the Z of the 2nd pass for the 02 series targets, while 5mm (average) is the value given and the theoretical value and the target thickness for the first pass. Since not all measurements were actually taken, I will have to substitute theoretical values in places. Since the variation is generally quite small, this is acceptable. (A handful of them exceed 100 microns--one is almost 800 microns). Points 25407 and 24309 didn't have the target driven completely into the hole.
Modified 16-March-2010 at 04:28
|Previous notes||Next notes||Main slide directory|