jeffrey.kissel@LIGO.ORG - posted 11:32, Thursday 06 September 2012 - last comment - 11:51, Tuesday 11 September 2012(4109)
Updates to QUAD Models (wire rehang, erm, thincp)
M. Barton, J. Kissel, J. Shapiro, S. Stepleskwi Mark has spent a good bit of time, (a) creating a new parameter set of the QUAD model to represent the current, odd-ball main-chain of H2 SUS ITMY configuration (a wire hang of the a glass mass) called 'wirerehang' (originally reported in LHO aLOG 3017), and (b) while creating the awesome new QUAD model mode shape wikipages, discovered a few bugs in the reaction chain model parameter sets ('erm' and 'thincp'), and has fixed them. Here, I document these changes by (1) Comparing the updated model against the previous version (in the case of the reaction chain parameters), and comparing against what we would have used otherwise (in the case of the wire rehang models), (2) Comparing both models to representative measurements we have of each, and (3) Explaining / justifying the details of the parameters that have changed. For (1) and (2), see attached plots for each of the three updated models, and for (3) see the tables below. Note, only parameters that have changed are shown, and of those, only the changes in hard-coded (as opposed to derived/calculated from hard-coded) parameters are shown. As of this entry, the updates to ${SusSVN}/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/quadopt_wirerehang.m ${SusSVN}/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/quadopt_thincp.m ${SusSVN}/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/quadopt_erm.m which are options for the buildType argument of the function ${SusSVN}/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/generate_Quad_Model_Production.m which produces models all production analysis software suites have been committed to the svn, under rev 3304. Please svn up! Executive summary of changes: ----------------------------- Wire Re-hang Major Changes: - Changed sign of TOP ('n' stage) off-diagonal moments of inertia due to misinterpretation of coordinate system used in Final Design Document (though this has little affect on the predicted dynamics because the magnitude is small) Small Changes: - TOP ('n' stage) and UIM ('1' stage) mass weights ('m' s) and moments of inertia ('I' s) adjusted to reflect QUAD retrofits - Radius, mass and moments of inertia of TST ('3' stage) changed to reflect the (glass mass and prisms) vs. (metal mass and prisms) - Change of d's to better reflect measured pitch frequencies of (glass mass and prisms) vs. (metal mass and prisms) - PUM to TST horizontal separations ('n3', 'n4','n5') changed to match glass mass and prisms vs. metal mass and prisms - Spring stiffness ('k') updated to match Brett's 2010 fit to vertical TFs (vs. a prior, 2008 fit). From the plots: - Most features remain identical, and match up well with either the wire measurements or the one wire-rehang H2 SUS ITMY measurement. - As usual, Pitch, the most sensitive degree of freedom, is now better matched to the data, but have changed no more than 10%. - The changed parameters only affect Pitch dynamics, so the model hasn't changed, and still matches the data exquisitely. Thin CP Major Changes: - Changed sign of TOP ('n' stage) off-diagonal moments of inertia due to misinterpretation of coordinate system used in Final Design Document (though this has little affect on the predicted dynamics because the magnitude is small) Minor Changes: - UIM ('1' stage) mass weights ('m' s) and moments of inertia ('I' s) adjusted to reflect QUAD retrofits (specifically the pitch adjuster). From the plots: - Without lacing cables, the model continues to match the data extremely well. - With lacing cables, in L, T, V, R, and Y, there's still a bit of difference between the model and measurements, though only in stiffness as expected. The resonant frequencies still match just about as good without lacing cables. - Pitch, the DOF most affected by the lacing cables (who's surprised?) is the worst, but still, only the lowest two modes in frequency are increased (i.e. modeP1 and modeP2 which involve the TOP and UIM masses, which have the most lacing cables running through them). - Any and all remaining discrepancies between model and measurement are not of much concern, since this is a reaction chain which has no where near the control-noise performance requirements as the main chain, so we'll most likely rely on a simple set of damping filters that are robust against the differences. ERM Major Changes: - Corrected TST stage thickness, it had the Thin CP value (copy'n'paste oversight). - Changed sign of TOP ('n' stage) and PUM ('2' stage) off-diagonal moments of inertia due to misinterpretation of coordinate system used in Final Design Document (though this has little affect on the predicted dynamics because the magnitude is small) Minor Changes: - Updated TST stage moments of inertia for an ERM, as opposed to a Thin CP. From the plots: - For L,T,V,R, and Y, either model doesn't perfectly capture the changes brought on by lacing cables, or from switching from metal to glass -- specifically the increase in stiffness (from cables), and the bifurcation of the second trans mode -- but, as with the Thin CP, it's expected and not a big deal. - For Pitch, neither model gets the stiffness right, cables or no cables, though we should triple check this against other no-lacing cable, ERM chain measurements. Unclear why this is. Naturally, when you add lacing cables, the lowest resonant modes increase in frequency, but this is exactly the same as on a Thin CP, so it's understandable since they have the same cables, the same configuration of routing, and the mode shapes are the same (see modeP1 and modeP2 of the last version of the production models). - Any and all remaining discrepancies between model and measurement are not of much concern, since this is a reaction chain which has no where near the control-noise performance requirements as the main chain, so we'll most likely rely on a simple set of damping filters that are robust against the differences. Full details of parameter changes --------------------------------- Table 1: Wire Rehang 'Param' 'Former Production Value' 'Updated Value' 'Differnce' 'Difference' '' 'Model: wire' 'Model: wirerehang' '(Absolute)' '(Percent)' 'mn' '21.9' '22' '0.0696' '0.317%' 'Inyz' '4.65e-05' '-4.65e-05' '-9.3093e-05' '-200%' 'Inzx' '0.00172' '-0.00172' '-0.0034368' '-200%' 'm1' '22.3' '21.5' '-0.81226' '-3.64%' 'I1x' '0.509' '0.505' '-0.0040466' '-0.795%' 'I1y' '0.0711' '0.0724' '0.0013481' '1.9%' 'I1z' '0.518' '0.518' '0.00013532' '0.0261%' 'I1xy' '-0.0132' '-0.0132' '5.271e-06' '-0.0399%' 'I1yz' '0' '1.37e-05' '1.3742e-05' 'Inf%' 'I1zx' '0' '-8.08e-06' '-8.084e-06' '-Inf%' 'm3' '39.6' '39.6' '0.031' '0.0783%' 'I3x' '0.598' '0.568' '-0.029963' '-5.01%' 'I3y' '0.418' '0.42' '0.0011272' '0.269%' 'I3z' '0.4' '0.411' '0.010141' '2.53%' 'dm' '-0.00351' '-0.00353' '-2.2308e-05' '0.636%' 'dn' '0.00328' '0.00423' '0.00095167' '29%' 'd0' '-0.00174' '-0.00175' '-1.0004e-05' '0.575%' 'd1' '0.00299' '0.00399' '0.00099789' '33.3%' 'd2' '0.00709' '0.00708' '-4.0147e-06' '-0.0566%' 'd3' '0.001' '-0.00116' '-0.0021609' '-216%' 'd4' '0.001' '-0.00116' '-0.0021609' '-216%' 'n3' '0.176' '0.172' '-0.00425' '-2.41%' 'n4' '0.171' '0.172' '0.00075' '0.438%' 'n5' '0.171' '0.177' '0.00555' '3.24%' 'ln' '0.445' '0.449' '0.004192' '0.942%' 'l1' '0.311' '0.309' '-0.002415' '-0.777%' 'l2' '0.339' '0.331' '-0.008213' '-2.42%' 'l3' '0.604' '0.604' '0.00029403' '0.0487%' 'r1' '0.000355' '0.000356' '5e-07' '0.141%' 'kcn' '1.41e+03' '1.43e+03' '17.9919' '1.27%' 'kc1' '1.65e+03' '1.65e+03' '-1.8307' '-0.111%' 'kc2' '2.42e+03' '2.38e+03' '-40.5505' '-1.67%' 'kw3' '1.15e+05' '1.15e+05' '-56.022' '-0.0487%' Table 2: Thin CP 'Param' 'Former Production Value' 'Updated Value' 'Differnce' 'Difference' '' 'Model: thincp' 'Model: thincp_20120831' '(Absolute)' '(Percent)' 'I1y' '0.073' '0.0734' '0.00040601' '0.556%' 'I1z' '0.519' '0.518' '-0.00077135' '-0.149%' 'den3' '3.98e+03' '2.2e+03' '-1780' '-44.7%' 'Inyz' '4.36e-05' '-4.36e-05' '-8.7197e-05' '-200%' 'Inzx' '-0.00171' '0.00171' '0.0034279' '-200%' Table 3: ERM 'Param' 'Former Production Value' 'Updated Value' 'Differnce' 'Difference' '' 'Model: erm' 'Model: erm_20120831' '(Absolute)' '(Percent)' 'I1y' '0.073' '0.0734' '0.00040601' '0.556%' 'I1z' '0.519' '0.518' '-0.00077135' '-0.149%' 'tx' '0.1' '0.13' '0.02996' '29.9%' 'tr' '0.17' '0.17' '-1.5e-05' '-0.00882%' 'den3' '3.98e+03' '2.2e+03' '-1780' '-44.7%' 'I3x' '0.376' '0.376' '-6.6301e-05' '-0.0176%' 'I3y' '0.21' '0.225' '0.014932' '7.12%' 'I3z' '0.21' '0.225' '0.014932' '7.12%' 'Inyz' '4.36e-05' '-4.36e-05' '-8.7197e-05' '-200%' 'Inzx' '-0.00171' '0.00171' '0.0034279' '-200%' 'I2yz' '-2.38e-05' '2.38e-05' '4.751e-05' '-200%' 'I2zx' '-4.41e-05' '4.41e-05' '8.8146e-05' '-200%'
Non-image files attached to this report
Comments related to this report
A couple of notes: * The CP and ERM models correspond to cases mark.barton/20120831TMproductionCP and mark.barton/20120831TMproductionERM of the Mathematica model. * Wiki pages for the CP and ERM models were generated on 9/11/12 and linked to from https://awiki.ligo-wa.caltech.edu/aLIGO/Suspensions/Background/QUAD/Models . * Also on 9/11/12, the resonance wiki page at https://lhocds.ligo-wa.caltech.edu/wiki/Resonances was updated with the new (but almost identical) mode frequencies. * The wire rehang model r3304 is based on Mathematica case mark.barton/20120601TMproductionTMrehang but has had one additional edit by Jeff K to supply a parameter pend.bd not used in the Mathematica. A new case to be called mark.barton/20120831TMproductionTMrehang that has the same MOI fixes as the CP and ERM updates is in the works.