Attaching collections of plots for are result from "individually" processing the data before they get used in the comparison script.

Let's compare pages of each SUS -- open each of the attachments in tabs, and line up the follow pages in your browse window (or download all the .pdfs and compare side-by-side).
As always, you can ignore the incoherent junk above 10 Hz. We'll get better measurements of this once we pump down the chambers, and we can tailor the excitation drive to better excite these cross-couplings at these frequencies.

Pages 4 -- P to L or L to P transfer functions.
    - Above 2 Hz, ZM1 and ZM2 have drastically less P to L coupling than ZM4 ZM5 and ZM6.
    - For All SUS, the L to P transfer function measures roughly consistent and appears to match the model, as opposed to the greater-than-2-Hz P to L TF.
    - The fact that ZM6 has similar levels of cross-coupling P to L magnitude to ZM4 and 5, but ZM2 has similar low levels as ZM1 implies, that its *not* an issue with the ZM4 and ZM5 HPDS PZT cabling.

Pages 5 -- L to Y and Y to L transfer functions.
    - Note, the model makes no attempt to model any L to Y coupling.
    - Kind of similar story here -- ZM1 and ZM2 are good -- ZM4 - 6 are worse (I don't want to say "bad" unless I know a requirement metric I can compare it against).
    - Also same story -- for ZM4-6, the Y to L coupling is worse than the L to Y coupling. Maybe this is coil balancing, and somehow ZM1 and ZM2 drew the lucky straws?
    - Looking at the phase relationship in the Y to L TF resonances, it seems like the relationship is random across all SUS (i.e. whether a cross-coupled resonance has a zero first or a resonance first)

Pages 6 -- the P to Y and Y to P transfer functions.
    - Again, the model makes no attempt at these TFs.
    - Same story -- ZM4-5-6 are worse than ZM2 -- though interestingly, ZM1's Y to P TF is only about half as large compared to ZM4-6, where ZM2 is the opposite -- the P to Y is half as large compared to ZM4-6.
    - Either way, ZM4-6 have a pretty nasty, complicated resonant structure, with TFs all of comparable magnitude, which no expects from a simple dynamical model.

Interesting... (in that "I don't really want it to be interesting at all, but sadly it is" kind of way...)

J. Kissel, R. Kumar

Rahul and I had a good conversation about his experiences through assembling, adjusting, and balancing the variety of HXDS flavors in hopes to narrow down the dynamical model parameter space that may have flexibility / adjustability and simultaneously might contribute to the differences we see in the above measured transfer functions. Here's a debrief / summary of that conversation.

In order to aide the conversation, we looked at 
    - The mechanical assembly of the "basic" HXDS, D1900352,
    - The diagrams of which model parameters correspond to which physical parameters, T1200364

I also add some whiteboard drawings that we used to help us work things out.

So, with all these visual aides in front of you, here's the list of things that we imagine might be affecting the differences we see in the P to P and Y to Y modes.
Thus far, the ideas are not backed up by actually noodling these parameters in the dynamical model to confirm that these parameter mobilities *actually* result in the changes we see; they're only guesses based on our intuition during the conversation. However, we think it's worth writing down to expose all of the ways that the HXPS are adjustable / able to be different for future reference. And over the next few days, we'll be following up with modeling exercises and mechanical / visual quantitative measurements to help corroborate our refute these ideas.

(1) IMG_5624.jpg aides in conveying that the top-down view, lateral alignment of the blade tips are adjustable, and the alignment reference is only visual. Rahul recalls that during the installation and alignment of ZM6, Betsy did adjust this lateral blade alignment in order to achieve the desired yaw alignment of the SUS. For all subsequent HXDS, ZMs 1-5, instead, the blades were left alone, and the whole cage was rotated. We imagine that if the blade tips are pushed differentially outward during adjustment, than this might effect the transverse separation of the blade tip suspension point, thus the parameter (2*) n0, which we imagine will have an impact on the yaw modes. We could also imagine how if they're laterally adjusted in common, that might misalign the principle axis of rotation with the center of mass of the top (M1) mass, which would increase cross coupling. Rahul expects that the slop in the lateral adjustment of each tip is ~+/-0.5 mm. Rahul's going to try to grab an inventory of blade-tip to blade-tip distance for ZMs 1-5 in HAM7 to see if there's a visible pattern there that matches the feature patterns in the TFs.

(2) IMG_5625.jpg From a "face on" view, if there were differential vertical position and alignment of the (cage to top-mass) suspension point blade tips with respect to each other, then longitudinal / pitch drive from the back of the top (M1) mass would create some roll and/or yaw. In terms of model parameters, these blade tip heights would correspond to an assymetric "dtop" (which I don't think the model can accommodate at the moment). We don't have observability of the roll transfer function with this top mass OSEM arrangement, but they might perhaps be a source of excess L to P, L to Y, P to Y cross coupling. Rahul expects that the blade tip heights have adjustment of ~+/- 1 mm. Note, we don't expect that the dynamics would be impacted if both tips are collectively in sync vertically, and collectively high or low. That would just result in a DC shift in the vertical position of the whole suspend chain w.r.t. the cage. 
    Counter point -- but it *might* have an impact in misaligning the coils with the magnets, given that the BOSEM coils are mounted to the cage in essentially a fixed position.  
    Counter-counter point -- the flags position within the BOSEM coil / sensor should be more limiting that +/- 1 mm. 
    Counter-counter-counter point -- do the BOSEMs on the HXDS have cam-nut, individual positioning adjustment, and if so, how much is the range of motion? 
Rahul will also take visual (and as quantitative as he can) inventory of the symmetry of blade tip heights for ZM1-5.

(3) IMG_5626.jpg Rahul's experience with the HPDS (ZM2, ZM4, ZM5) has revealed that the M2 payload mass that contains the PZT actuation system on the optic, is pretty "front" (mirror, HR side) heavy, where as the others, ZM1, ZM3, and ZM6 were relatively easy to balance at a neutral pitch. As such, one could imagine that, with the amount of adjustment mass that alignment-to-neutral required (which is affixed symmetrically to the back in groups of 4), the pitch moment of inertia might be altered. The relationship *between* the M1 and M2 moment of inertia of a non-extra-balance-mass HXDS vs. an extra-balance-mass HPDS might mean than pitch zeros are in different places. Rahul assures me that he's kept good track of what adjustment mass he's used, so we're going gather an inventory of that mass to bound how much we might expect to be able to noodle that M2 pitch moment of inertia dynamical model parameter.

(No visual aides, but also perhaps useful information)
(4) HPDS M1 to M2 Cabling has a clear impact on the amount of 1.6 P and and 2.2. Hz L modes cross coupling into the Y to Y TF. As indicated in the main aLOG above, though, there doesn't seem to be a pattern that ZMs 2, 4, and 5 have worse L to P, or P to Y transfer functions because of this -- because ZM6 (a cable free HSDS) breaks that pattern with equally worse L to P and P to Y TF magnitudes as the HPDS. Rahul says there's no science to finding the right cable routing that reduces the unwanted 1.6 and 2.2 Hz mode features in the Y to Y transfer function, he just iterates re-setting the routing and re-measuring the Y to Y TF until the mode impact is minimized.

(5) The order of construction for these SUS was ZM6, then ZM45, then ZM2, then ZM1 and ZM3, where the groups are several weeks apart is we (well, Rahul and Betsy) spent those installing them as needed. Rahul does feel as though the assembly them got better at constructing the HXDS as they went along, developing tricks that helped clamp wires more consistently, paying closer attention to the alignment of certain elements etc. He shares that because ZM1 shows the closest alignment to the model in terms of P2P and Y2Y inter-mode zeros, where as ZM456 -- even though 6 is an HSDS and 45 are HPDSs -- have a spread of where these features land.