The obvious way to try to deal with the Z situation is to see if it is possible to fit for the Z positions of the MABs as well as the rotations and translations of the disks, assuming that the transfer plates are at the PG positions and that all shifts are due to rotations, translations, and MAB target positions. Since there is an overall Z-shift, I have to fix one of the measurements, which I take to be the center of YE+1, which I set to be 0. We can use the Link measurement of this to then find the overall positions of everything.
Unfortunately the disks don't bend uniformly, but we can try a first approximation here.
gp < SolveZ.gp | sed -f clean.sed
Unfortunately the results are not encouraging. The fit MAB targets have a spread of 171.4mm, which doesn't seem possible. The X and Y rotations seem locked together more closely than I'd expect too. The Z centers are also quite a bit farther from nominal than can be accounted for by bending.
Target1= -934.8508607266579404529143171 Target2= -862.1334737003287223899279111 Target3= -790.3514738980112554637288944 Target4= -813.2588116970387671050306728 Target5= -894.2549530109558160542986926 Target6= -961.7664601719863401996152679 YE2Z= 1340.149305968572976442185798 YE1RotX= 0.005461262500092890670900182724 YE2RotX= 0.005612981178782552844386748852 YE3RotX= 0.006281847941107928822785590352 YE3Z= 2415.021927006819581112756912 YE1RotY= 0.01043750089255111452513481067 YE2RotY= 0.01029359043073513485960134951 YE3RotY= 0.01051157838544715898965646133
Conclusion: The Z model is wrong. The predicted average MAB target position is about 6690mm (using nominal for YE+1), which is consistent with estimates using Excel, so the fit isn't unstable. The model must be wrong. But where?
Modified 28-May-2010 at 16:10
|Previous notes||Next notes||Main slide directory|