J. Kissel, P. Fritschel

After all the investigations into the SUS loop performance to find the source of the excess motion Pitch and Longitudinal motion at 0.43 (, 0.56, and 1.0) [Hz], Peter suggested perhaps it's merely the expected coupling between L and P, with the large not-yet-awesome input motion from the BSC-ISIs. So, following the same prescription used to generate the curves in G1200712, I produced the predicted QUAD test mass motion due to the measured (BSC8-ISI, ITMY) motion on June 26th, i.e. E1200668, in both L and P.

Attached are the results.

We see that indeed, given the (BSC8-ISI, ITMY) input motion, the predicted motion is on the order of 2.4e-6 [m RMS] and 5e-6 [rad RMS], with the expected first L and P modes at 0.43 Hz and 0.56 Hz, which have (modeled) damped amplitudes of 1e-5 [m/rtHz] and 2e-5 [rad/rtHz], reasonably consistent with results reported in LHO aLOG 3363, LHO aLOG 3302, and LHO aLOG.

There are several ways in which this model isn't perfect:
- The input motion is representative of BSC8-ISI (ITMY), which has been commissioned further than BSC6-ISI (ETMY), so one might expect the input motion to ETMY to be somewhat worse (It only has damping, No HEPI, No Level 0 Isolation Loops). Fabrice is working on getting me BSC6 data.
- I took the input motions in X, Y, Z, RX, RY, and RZ -- which are defined about the center (in X and Y) of the ISI, and at the lower-zero-moment-actuation-plane of ST2 of the ISI -- as direct inputs to the suspension point of the QUAD: (Y->) L, (X->) T, (Z->) V, (RY->) R, (RX->) P, (RZ->) Y, which means that the estimate for P does not account for the ~0.5 [m] lever arm between the ISI ST2 origin and the SUS point origin. See T1100617 for details (B. Lantz, C. Kucharcyzk, and I are working on getting the correct transform.)

The bonus attachments, relevant to the discussion, are as follows
- [2012-07-09_testmassmotion_bscinput.pdf] A plot of the motion used as input (identical to what is shown in E1200668)
- [2012-07-09_modeltf_*toP.pdf] The rarely-looked-at transfer functions between all degrees of freedom input to pitch in [rad/m] or [rad/rad].
- [2012-07-09_testmassmotion_P_resgndbudget.pdf] A break down of the predicted residual P motion from all degrees of freedom.

One VERY interesting revelation from these bonus plots: 
- T @SUS point transmission to P @ test mass becomes comparable to L to P in [rad/m] above ~1 Hz
- R @SUS point transmission to P @ test mass becomes roughly a factor of 10 greater to P to P in [rad/rad] above ~0.5 Hz
These mean we'll have to focus on reducing the T and R motion *just as much* as reducing the L and P motion.

HAM-ISI Unit #6 was balanced, and CPS readouts were within acceptable range, on 07/03 when we left it for long TF measurments. When checked this morning, CPS readouts featured an offset of approximately +/-3000 counts:

There is no drive/offset on MEDM channels.

Coil drivers were turned off. The CPS readouts remained unchanged.

CPS interface chassis were turned off, and then turned back on. The CPS readouts remained unchanged.

Dataviewer plots are attached. They show the drifting of the CPS readouts.

We experienced high temperatures over the last few days.

J. Kissel, M. Evans

As mentioned in LHO aLOG 3312, the previous measurements of the open loop gain TFs of the L P and Y loops for H2 SUS ETMY with the loops open. This method of getting the open loop gain would be fine if these degrees of freedom were entirely independent. However, L and P are fundamentally cross-coupled (and sounds like there may be some non-fundamental cross-coupling between L and Y). Further, if we want to obtain not only the open loop gain ( -G ), but the suppression ( 1/(1+G) ) and closed loop gain ( -G/(1+G) ), the loops need to be closed. If you're interested in the math behind these truths, as well as derivations of each of these figures of merit from the available excitation and test points, check out T1200336.

And so it goes. 

The attached measurements show the open loop gain (for the diagonal degree of freedom; L to L, P to P, Y to Y) and the closed loop gain for both the diagonal and off-diagonal terms. The plots compare three different data sets:

- 2012-06-29, Loops Open, No Offset is the same data set as was taken two Fridays ago. The damping loops were all off (input disabled -- hence "Loops Open"), so the open loop gain, G = P * K = IN1/EXC was obtained by driving through the DAMP filters. "No Offset" indicates that the P and Y offsets used to align the cavity were removed, for fair of saturation during the excitation.
- 2012-07-07, Loops Closed, No Offset is new data from yesterday that was taken with no offsets again, but this time with damping loops closed. Because the loops are closed, the ratio of IN1/IN2 yields the open loop gain (again, see T1200336 if you don't believe me). The loops were configured as the were for the 2012-06-29 data, in configuration (1) listed in LHO aLOG 3306.
- 2012-07-07, Loops Closed, With Offset Because two things are different between nominal operation and black, the offsets and the loops being closed, we wanted to be sure that the offsets were not causing the difference between black and blue. So, I repeated the measurements with the (what were as of yesterday) nominal offsets ON (P: -2841, Y: 14290). Only L needed 1/2 the drive, the P and Y levels I had used for loops open worked just fine; there were no saturations during any data set.

Note that since IN1/EXC gives you the closed loop gain when loops are closed, the 4 page of each attachment only shows the closed loop gain for the latter two data sets.

The things that are evident from these results:
- There is an obvious difference between the Loops Open and Loops Closed, L and P, open loop gains.
- Because there is little to no difference between the No Offset and With Offset data, we can conclude that the OSEMs (sensors and actuators) are linear over a wide range of offsets (thank god).
- Since there were only two (obvious) differences (Offsets and Loops Open/Closed), and blue to red rules out the offsets, we conclude there is indeed significant cross coupling between L and P that's effecting the performance of the loops.
- We still see a much larger L to P than L to L closed loop gain, where the P to L is much less than P to P in the closed loop gain.
- The shape of the L to P closed loop gain is weird, but a plot of the P to L closed loop gain with loops closed against the open loop gain with loops open show that all the resonances are in the right place (see pg 4 of L plots, in gray)
- As was mentioned many times before, excess motion, is focused around 0.43 Hz. But one can see comparing gray to brown, that open vs. closed loop shows that the reduction is dramatic, and it shows that what is left is not particularly "peaky" in that frequency region.
- The P to L and Y to L closed loop gains are both significantly well below their respective diagonal terms, and in fact show little coherence so even what is shown may be an upper limit.
- There is little to no change in the Y to Y open loop gain between loops closed and loops open. This is what we would expect from an independent degree of freedom.

So. No new answers here, just relevant data that (perhaps) has ruled more things out, and is steering us to a different direction (which Matt seems to have already found some unexpected but good leads with the T loop, as well as a ~0.3% L to Y coupling, most likely do to [mismatched/imperfectly matched] actuators, see his aLOG below).

The next thing to do is calculate the pitch motion at the test mass, given measured ISI input, and using P sensor noise (and do the same thing for yaw), similar to what was done for L and V in G1200712.

------------------
Data can be found here:

/ligo/svncommon/SusSVN/sus/trunk/QUAD/H2/ETMY/SAGM0/Data/
2012-06-29_1425_H2SUSETMY_M0_Mono_L_WhiteNoise_OLGTF.xml
2012-06-29_1425_H2SUSETMY_M0_Mono_P_WhiteNoise_OLGTF.xml
2012-06-29_1425_H2SUSETMY_M0_Mono_Y_WhiteNoise_OLGTF.xml
2012-07-07_1811_H2SUSETMY_M0_Mono_L_WhiteNoise_OLGTF.xml
2012-07-07_1811_H2SUSETMY_M0_Mono_P_WhiteNoise_OLGTF.xml
2012-07-07_1811_H2SUSETMY_M0_Mono_Y_WhiteNoise_OLGTF.xml
2012-07-07_2012_H2SUSETMY_M0_Mono_L_WhiteNoise_OLGTF.xml
2012-07-07_2012_H2SUSETMY_M0_Mono_P_WhiteNoise_OLGTF.xml
2012-07-07_2012_H2SUSETMY_M0_Mono_Y_WhiteNoise_OLGTF.xml

ALS Input Pointing QPD->PZT Loops

I started this fine Saturday checking the PZT alignment loops while JeffK measured SUS TFs.  The idea was to make sure that the loops were fast enough to ensure that sub-Hertz alignment loops would not be bothered by sluggish PZT response.  I found all of the loops in good working order with UGFs between 3 and 10Hz.  While I was there I added cut-off and boost filters (loops are still unconditionally stable) and set all the UGFs a little closer to 10Hz.

SUS Damping

In search of something to blame for our large angular motion, much of which appears at 0.43Hz (the first longitudinal resonance), I measured the Transverse loop (side OSEM), which also has a resonance around 0.43Hz.  It wasn't doing much, so I broght it up to speed with the Longitudinal loop boost filter.  I also changed the cut-off somewhat, as there was very little phase margin for the highest resonance (near 4Hz).  The result was a much more damped T DOF, but not much else that was obvious.

Slow Length Loop

After figuring out that the PHASE and PDH fast monitor signals are switched in digital land, I was able to close a slow loop (< 100mHz UGF) which off-loads the VCO signal for locking the arm cavity to the ETM.  At the moment, the locking filter is in the SUS-ETMY_M0_LOCK_L filter bank rather than the ALS_EY_ARM_LONG bank.  This is because ETMY_M0_LOCK_L_IN1 is recorded at 2kHz, which is useful for making spectra (see below).

Calibrated Spectrum

By stepping the gain of the slow loop after a large output had accumulated (with the input switch off), I was able to modulate the length of the cavity without unlocking.  This modulation is visible in the signal send to the VCO for cavity locking, and in the ETMY M0 OSEMS, so I cross-calibrated teh VCO signal using 40nm/cnt as round number for the OSEMS (see fig 1).  This gave 0.8nm/cnt for FMON, which might be 40kHz/V for the VCO if I got all of my conversions right... not so far from the 100kHz/V I would have guessed.  I made 3 spectra from locks separated by ~1 hour, all of which overlap nicely (see fig 2).

Lock Stability

The main threat to lock stability was a large Long -> Yaw coupling that would misalign the cavity as the slow length loop pushed on ETMY_MO_LOCK_L.  I fixed this with a small element in the drive align matrix (L2Y gain = 0.0036), indicating that we will probably have this problem with all of the suspensions just due to small magnet strentgh mismatches and BOSEM misalignments.  With this in place, the cavity will stay locked for hours with ~10% power flucutions (see fig 3, 4).  For reference, I include the alignment slider values in fig 5.

Matt called to report a chilled water temperature alarm at ENDX.

I was able to start the second chilled water pump which in turn enabled the second chiller. Water temperatures are now falling. We'll investigate the tripped chiller on Monday.

We were lucky yesterday when we connected the REFL DC to the Tabletop Interface Box on the EY ISC table.  The signal arrived in digital land in H2:ISC-ALS_EY_PDH_DC as expected.  None of the other 3 DC signals work (IR_PWR, GR_PWR and BB_PWR).  I connected BB DC to the TTIF box Ch4, and found that it arrives at the concentrator in the field rack (D1102045 #36, on cable 73), but I didn't continue tracking the problem beyond that.

The EY ISI was also tripped.  I reset it, so we have damping, but I don't know the secret handshakes for turning on the isolation, so we will live without.

It isn't obvious to me how to reset this watchdog.  Has this system ever been made to work?  For the moment we are "floating on oil".

Some time this week, somebody changed PEM corner station model, and as a result it seems like one of the baffle diode channels, H2:PEM-CS_CHAN_30, is gone. The channel doesn't exist any more.

Please give us the baffle diode back. You don't have to revert the model back, all we need is another PEM channel (e.g. H2:PEM-CS_CHAN_26) and we'll fix our MEDM screen.

We're done with Faraday replacement. We might have to tune the optical paths on the ALS table a bit, but basically we're done.

We don't see any serious clipping in the PDH error signal, the arm locks, and the alignment is already reasonable (50% reflection visibility).

Now that we can do things without worrying too much about the clipping on the Faraday, we quickly tested to change the cavity length by pushing on the ETM while maintaining the PDH lock, and it worked. We were able to (manually) relieve VCO tuning signal by merely giving an offset to the ETM TEST L filter. We'll implement (or Matt has already implemented) a fully featured low frequency feed back to the ETM.

We haven't tuned WFS demod phase yet.

Just a heads up, I'll be starting some measurements using ETMY nowish (2pm ET, July 7th 2012) which should last for 2 to 3 hours. I will post when I finish, and then results and analysis will be to come. I'll be launching DTT measurements from cdsws1.

For those who are interested (though I think I emailed everyone who might be):
Matt Evans and I, before I left last friday (June 29) had thought we'd stumbled on something interesting regarding cross-coupling vs. excess cavity motion while measuring the open loop gain transfer functions of the damping loops. See LHO aLOG 3312. Specifically the 4th page of the L and P sets of plots showing that, with the damping loops open, the cross-coupling from L to P is *huge*, where it's not huge in P to L. But, as I mention in the log, the right thing to do is to measure the open loop gain with the loops closed (i.e. IN1/IN2), and take a look at the results. These are the measurements I'll be taking now.