jeffrey.kissel@LIGO.ORG - posted 18:42, Wednesday 07 December 2011 - last comment - 14:32, Thursday 08 December 2011(1852)
H2 SUS ITMY M0, Long to Yaw Coupling Investigation
In order to investigate whether the Length modes coupling into H2 SUS ITMY M0 Top2Top Yaw transfer functions are real or noise, I've taken a high resolution (2 mHz) measurement of the same transfer function (same amplitude, BSC-ISI unlocked but undamped, ITMY M0 damping OFF, ITMY R0 damping ON).
I attach three plots for discussion.
(1) A full-frequency-range plot measurement itself,
(2) A zoom in on the resonance that we've been concerned with, and
(3) A plot of what cross-coupling we expect from the model (i.e. zip, nadda, zilch).
Note that neither of the first two plots are calibrated properly, but the relative amplitude should be accurate
One can see from the full-range plot that not only is the lowest Longitudinal mode present (at 0.43 Hz), but the second L mode (at 0.98 Hz) also creeps in. Regrettably, I'm now convinced that this (these) resonances are actually a measurement of physical motion, not just unlucky in coherent noise.
Now, is it a show stopper? No (yet).
A useful tip from the good Dr. Lantz: physically cross-coupled modes typically show up as pole-zero pairs, as opposed to what we see here -- just a sharp pole.
Other pertinent information: the excitation for this drive is a continuous, broad-band, white noise excitation across the measurement band, for the duration of the measurement. You'll note in the second plot attached (upper right panel), that the coherence (i.e. the measure of the *linear* coupling) between the Y drive and this particular L resonance is ~0.25, which is roughly consistent with what we know to be noise in the rest of the band.
However, the lower right panel shows the OSEM basis response of F2 and F3 to Y (in PHASE); the sensors that compose this DOFs Euler basis signal. Here, (though it's tough to see with the black cursor overlayed -- sorry) the sensors are identically in phase, implying real longitudinal motion.
Why don't I think this is a show stopper (yet)? We have found from experience that moving around these suspensions, after locking and unlocking, that these sharp cross couplings come and go. Case and point -- we don't have a smoking gun of what might have happened between the 2011-11-19, 2011-11-29, and 2011-12-02 measurements that might have caused such a gradual increase in coupling, except for *better* aligning the chains. Further, I expect that the coupling will be significantly reduced once we take a similar measurement with damping loops ON (we'll be sure to confirm this of course) -- which is the default "plant" upon which we'll apply ISC control loops (if there are any at this stage).
But most importantly, let's just see what we get after we install the cartridge. We'll have to lock and unlock the suspension, and will have to do another round of BOSEM centering (in and out, not necessarily laterally). We may get lucky and the coupling may be reduced, or we could get unlucky and have much worse coupling. Further, we have yet to use what's in our digital back-pocket: diagonalizing the drive using sensors. This may help as well. From what I've seen of the remaining degrees of freedom, I'm confident that the suspension's mechanical system is behaving well.
--------------------
Data for this measurement can be found here:
{SusSVN}/sus/trunk/QUAD/H2/ITMY/SAGM0/Data/20111207_1700_H2SUSITMY_M0_Mono_Y_0p002to50Hz_TF.xml
Non-image files attached to this report
Comments related to this report
B. Lantz, J. Kissel, B. Shapiro Brian guesses that this excess cross-coupling maybe be from air currents in, on, and around the suspension. The notes leading up to the hypothesis: - The amplitude at this frequency (0.43 Hz) in both the Yaw2Yaw and Yaw2Long transfer function is incoherent (~0.2 coherence, consistent with what we know is noise, or non-linear coupling at other frequencies). - One difference between the 2011-11-29 measurement and 2011-12-02 measurement is that the BSC ISI is unlocked (and undamped), and we know the BSC-ISI "is a big sail" when it comes to air currents**. - The clean room forces air current to move in, on, and around the QUAD, as well as the BSC-ISI. - Remember F2 and F3 are the Long (in common) and Yaw (in differential) sensors; they're in line with the vertical center of mass at the top stage. - OSEM response to linear drive goes incoherent on *expected* resonances, because the SUS is swinging with large amplitude outside the (linear) range of motion of the Flag/LED/PD system (think -- at the edge of the range, the signal flat-lines at open or closed light and is no longer proportional to the drive). Non-linear response to drive = still get amplitude, but no coherence. - "The OSEMs are linear to "+/- 0.7 mm" peak to peak." I put in quotes, because though this is the number we always quote as a spec, this number is eye-balled from the curves measured of a few OSEMs, on an independent jig, ideally aligned. Mark has shown the linearity to vary with alignment (see T1100455) and we know OSEMs can have ~50% variability in sensitivity). - Suspension Q's are "a billion," so we often cannot resolve their actual absolute motion. - Another plot is attached -- the calibrated amplitude spectra of the motion during the transfer function excitation, and after late at night during a quiet time (thank you data stored in frames!). The hypothesis: - Air currents are exciting the longitudinal mode by lots, but in an incoherent manner (such that it might be misconstrued as yaw). Because there is so much incoherent motion, it bleeds into the yaw sensors, and therefore into the amplitude of the transfer function. Devil's Advocate questions: - Why would there be so much more motion at longitudinal, vs. other degrees of freedom? (Not sure. BSC-ISI Y [not yaw but cartesian Y, aligned with IFO arms, and therefore ITMY's L direction] resonances are at ) - Wouldn't the air current excitation be broad-band? (Well -- so is the intentional excitation. We insert uniform white noise across the measurement band as our excitation) - Is there really a mechanism where longitudinal motion can be sensed as yaw? (If, for example F2 goes non-linear before F3 as the pendulum swings through the edge of OSEM range in L, then you'll get more amplitude in F2 than in F3; a differential signal, which appears as yaw.) - Why don't we see the same coupling on the reaction chain? (We did -- in the 2011-11-29 measurement (see page 6, magenta curve of allquads_111202_H2SUSITMY_ALLR0_TFs.pdf), arguably just as strongly, but it went away in the 2011-12-02 measurement) - Why don't we see anything on FMY? (Maybe because FMY is not aligned with any of the fundamental modes of the BSC-ISI?) **Auxiliary/Curiosity Questions: - What're the BSC-ISI Modes in the L degree of freedom? (See second attachement -- for this QUAD and ISI, the L and Y/RX modes are roughly aligned. for the BSC-ISI, those are at [1.0, 1.75, 5.15, 6.95 ] Hz)
Non-image files attached to this comment