After some spectacular detective work by Anamaria back in September, she had identified a bug in 
${SusSVN}/sus/trunk/Common/MatlabTools/TripleModel_Production/generate_Triple_Model_production.m
(the function used to generate all models of triple suspension [HSTS, HLTS, BSFM] dynamics, both damped and undamped) in which the connection matrix -- used to close the damping loops around the undamped state space model -- had double-counted the index of the input and output ports, effectively closing the damping loops around the M2 (Middle) stage instead of the M1 (Top) stage. This, as she said: "[...] made some of the [transfer functions] look believable [namely, the diagonal TFs], while some of the cross-couplings were obviously wrong (when compared to the undamped case). This [...] also explain[ed] why we needed a strange gain fudge factor to make the Qs [match measured data]."
As of the entry in September, Anamaria had fixed this bug [though added some other hard-coded stuff, which makes the script only valid for HSTSs for the time being].

Since I wrote these functions, I have always had similar problems with predicting reality on the QUADs. Because the Triple function was a copy, paste, and reduction of the similar function for the QUADs,
${SusSVN}/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/generate_QUAD_Model_production.m
written by the same jerk (i.e. me), I suspected this same subtle bug was present. 

Today, I looked, and *BUHZINGA*, the same bug.

As of this entry, and SusSVN rev 3738, I have fixed this bug in the QUAD function. Please update the following corner of your local SusSVN repo:
${SusSVN}/sus/trunk/QUAD/Common/MatlabTools/QuadModel_Production/
(Note that you'll receive some other updates as well, but no other changes are substantial.)

I'll now work on comparing it with measured data, but I assume it will work out nicely.

Other notes:
- This does not affect any previous *undamped* modeling results, this bug only manifested itself on *damped* model predictions.
- Attached is comparison of select transfer functions between various stages produced by the buggy model ("OLD," solid lines) and fixed model ("NEW," dashed lines). One can see, as Anamaria had described, that most diagonal degrees of freedom (highlighted with thicker lines) are not terribly different [just some subtle changes in Q of the given resonances], but the cross-coupled transfer functions (thinner lines) are significantly different. OK, maybe the plots are too busy to actually see, but take my word for it.
- The default damping filters used by the model are identical to those used for the H2OAT, but the gains are not well-matched. So, although this gives a *better* prediction of the damped QUAD dynamics it does not yet exactly reflect reality. More updates to come after comparing with measured data.