I found a small mistake in the HAM_gain_matching_calculation script.
The Q of the GS13 model was wrong by a factor a few, causing a slight phase delay around the frequencies that matter for the sensor correction ([0.1Hz-0.4Hz]). The overall amplitude was thus altered.
This has been fixed and committed into the SVN:
/ligo/svncommon/SeiSVN/seismic/HAM-ISI/Common/Misc/HAM_gain_matching_calculation.m
Here's the correct values that we have so far (I though we had done HAM2 as well, but I can't find a good set of data):
HAM4 | HAM5 | HAM6 | |
X | 0.945 | 1.001 | 1.035 |
Y | 0.929 | 0.991 | 1.027 |
Z | 0.918 | 1.039 | 1.103 |
We still need to do HAM2&3, plus all the BSCs. Don't forget to put the ISI in high blend mode with sensor correction OFF when you want to calculate the gain.
Using the Guardian manager, brought down the entire ETMX SEI after capturing the current held position for HEPI. The Guardian code has been changed to restore all 8 dofs on HEPI to reduce or nearly eliminate alignment changes from lock to lock. Before HEPI only restored the RZ and Pitch at the ETMs.
The Guardian smoothly brought the system down with no trips. Restarted the Guardian code and then went directly to fully isolated first time. No problems with the idea then. I'll step this through all the platforms as commissioners allow.
Of note, none of the DOFs on the BSC-ISI are restored like this (As opposed to the HAM-ISI where all dofs are restored.) So, I attach a plot showing the before and after of this morning's restart. The op-levs are shown and the HEPI RZ. Additionally, all the rotational dofs of the two ETMX ISI stagesare show to reveal at least one lock to lock alignment change on the ISI. All the SEI channels are in nano units, Oplevs are urads I believe. The Stage1pitch is worst at 600nrad tilt, all others are much smaller.
250ml of water added to chiller.
As we found that the beam was too big on the IMC refl camera (alog 16166), I inserted a focusing lens (PLCX-25.4-20.6-UV) in front of it in order to obtain a better and wider view. The belows are pictures before and after the lens insertion.
Before
After
Both pictures were taken with the nominal exposure time of 19444 usec. In the second image, the main 00 mode is seen at the lower right. I am not sure whether if the upper part of the image is explainable by spatial higher order modes. It maybe a ghost beam. Note that this ghost-ish structure was visible with a laser card.
model restarts logged for Tue 20/Jan/2015
2015_01_20 00:05 h1fw0
2015_01_20 04:07 h1fw0
2015_01_20 10:29 h1iopsusex
2015_01_20 10:30 h1iopsusex
2015_01_20 10:31 h1susetmx
2015_01_20 10:31 h1sustmsx
2015_01_20 10:32 h1susetmx
2015_01_20 10:35 h1iopsusex
2015_01_20 10:35 h1susetmx
2015_01_20 10:37 h1sustmsx
2015_01_20 10:44 h1susetmx
2015_01_20 10:51 h1iopsusey
2015_01_20 10:51 h1susetmy
2015_01_20 10:51 h1sustmsy
2015_01_20 11:31 h1hpiham2
2015_01_20 11:31 h1hpiham3
2015_01_20 11:31 h1iopseih23
2015_01_20 11:31 h1isiham2
2015_01_20 11:32 h1hpiham2
2015_01_20 11:32 h1hpiham3
2015_01_20 11:32 h1iopseih23
2015_01_20 11:32 h1isiham2
2015_01_20 11:32 h1isiham3
2015_01_20 11:58 h1fw0
2015_01_20 11:58 h1nds0
2015_01_20 12:49 h1fw1
2015_01_20 12:49 h1nds1
2015_01_20 12:57 h1broadcast0
2015_01_20 12:57 h1dc0
2015_01_20 12:57 h1fw0
2015_01_20 12:57 h1fw1
2015_01_20 12:57 h1nds0
2015_01_20 12:57 h1nds1
maintance day preceded with two unexpected restarts. Restart tests on SUS ETMs. Power cycle of h1seih23 for noise hunting. Power cycle of DAQ solaris boxes, followed by DAQ restart for new slow controls ini files. Conlog frequently changing channels report attached.
Alexa, Evan, Koji, Arnaud, Sheila, Rana
The good news is that ALS is quite stable tonight, staying locked for up to 2 hours.
We have been having trouble transitioning to 3F tonight. We can transition MICH and PRCL fine, but we have seen a lot of noise at 300 Hz when we try to transition SRCL to 3F. When we try to transition we see noise at 23.5Hz.
ALS lock loss times: (all Jan 21st UTC) 0:03:02, 0:34:26
ALS lock loss +HEPI, ISI and sus trips: 7:30:15, 5:15:26
Lockloss and bad misalingment of IMC due to DOF4: 6:35
DRMI + arms off resonance lock loss due to SRCL 3f transition at 01/21/14 08:06:00 UTC
Keita, Sheila, Daniel, Alexa
Here is a screen shot of what happened that caused HEPI and ISI trips on DIFF lockloss last night. First, the arms loose the green lock, at that time the DARM output becomes very large, the gaurdian ramps this down over ten seconds, durring which time the integrator in L1 is still integrating. The guardian eventually clears this integrator. Before the integrator was cleared, the arm cavity flashes, although the suspension was still swinging so this produced a momentary lock. The tidal comes on, with a large output because the integrator in L1 has not been cleared yet.
We have done 4 things to help prevent this in the future:
The last thing that we need to do is add a delay to the logic for the tidal feedback, so that it does not engage prematurely on these momentary locks.
Comparison of the OMC length noise for H1OMC and L1OMC.
Note that floor noise of the L1 measurement was limited by the intensity noise at the time.
Math puzzle: What's wrong with this UGF Servo?
its installed for use in the OMC and LSC at the moment, and could be used in the ASC if we find we want to hold the UGFs constant during TCS tuning.
We want to measure . The algebra tells us that
,
.
When the excitation is sufficiently large that the noise is negligible, we get .
and
are complex quantities, and the real and imaginary parts are the I and Q outputs of the demods, or
,
.
In the version of the UGF servo that Rana posted, the phases would be chosen such that the imaginary (Q) outputs are zero, thus
.
This is OK as long as G only ever changes in magnitude, otherwise .
To get the correct measurement in the case of a changing phase, one must do the following:
This is implemented in the attached simulink model.
(First of all I don't know the answer yet)
I believe G only changes the gain most of the case.
The problem was
a/e = G/(1-G) and b/e = 1/(1-G) change their phase as you change the gain of G.
We usually don't care the phase of G, but only care the magnitude of G as the phase of G is fixed.
Therefore what we need is to take the ratio of the magnitude of a and b
|a| = |G|/|1-G|, |b| = 1/|1-G|
|a|/|b| = |G|, where |a| = sqrt(aI2+aQ2) and |b| = sqrt(bI2+bQ2)
(Ed: I'm suggesting to take SQRT(I^2+Q^2) to eliminate the frequently-omitted-effort of adjusting the demod phase correctly.)
Meh, I say it's overkill. As Nic mentions, this works just fine if the UGF phase is not changing, so long as you set the demod phase correctly. As Koji mentions, the UGF phase should not be expected to change that much. It is already a second-order effect, so what is there now should be fine, unless we really want to accommodate wild loop phase fluctuations at the UGF. Is there a reason that's ever something we want?
Dan, Koji
This is a followup to Koji's post that described our OMC work on Sunday. (Here I give details of the noise investigations; the measurement of the OMC actuator noise was described by Koji yesterday.)
The summary is that the excess ~1kHz noise in the OMC length error signal is not consistent with intensity or alignment fluctuations, and isn't being amplified by a resonance on the OMC breadboard. This agrees with Zach's investigations from L1. The peaks are consistent with cavity length noise that is driven by acoustic noise coupling into the chamber via the ISI, but our measurements aren't conclusive. We will have to see if the noise appears in DARM -- at L1 this doesn't appear to happen.
1) Intensity noise suppression by ISS first loop.
First we investigated the stability and noise suppression of the ISS first loop. We found we could reliably close the first loop once the input offset had been tuned, but oscillations would develop when the integrator was turned on.
The first plot is intended as a qualitative reference of the performance of the ISS first loop. The intensity noise measured by the LSC REFL photodiode is improved above 300Hz with the first loop on, but the noise below 300Hz is unchanged, and there are peaks (at 700Hz, etc) that are not suppressed. Is this noise coming from the IMC?
(We had to use the REFL PD to measure the intensity noise after the IMC because the beam going to the ISS second loop PDs is misaligned.)
2) 1kHz peaks in OMC noise - intensity noise, alignment jitter, or OMC length noise?
Next we locked the OMC on a half-fringe and studied the noise. If the peaks at 1kHz are due to intensity noise (maybe due to a bad ISS), the coupling should increase as we move the OMC lock from the half-fringe to the full fringe. If the noise is due to beam jitter, we can change the coupling by moving the alignment into the OMC. If the peaks are due to real cavity length noise it should couple more strongly at the half-fringe lock.
The second plot tells the story -- there was no change in the peaks when we misaligned the cavity (compare purple and black), which seems to rule out beam jitter. There was some change in broadband noise as we moved the locking point up the fringe (compare blue, to cyan, to purple); this indicates that we were limited by intensity noise. In particular the peak at 817Hz in the lower plot is consistent with an intensity noise explanation. But, the other peaks were mostly unchanged as we detuned the cavity (although the 877Hz peaks seemed to be slightly worse at the 50% fringe). This appears to rule out intensity noise, but it's not really consistent with cavity length noise, either. So, we're baffled.
(Regarding beam jitter: we had tuned the alignment into the OMC by using offsets in the QPD loops to minimize the linear coupling between excitations of the tip-tilts to the OMC TRANS spectrum. Turning the offsets on reduced the height of alignment dither lines by 10x. So, we should have seem *something*, if the peaks were due to beam jitter.)
For reference, the frequencies of the largest peaks are the following: 817, 877, 937, 991, 1055, 1132, 1165 Hz. The location of the DTT file is at the top of the figure.
3) Comparison between PDH error signal and OMC TRANS.
Next we steered the OMC REFL beam onto ASAIR to use the PDH error signal as an out-of-loop sensor for the cavity length. Here, the cavity was locked at 80% of a fringe, to move the PDH error signal into the linear region. The third plot compares the PDH error signal (out-of-loop) to the OMC TRANS noise. (We rotated the ASAIR_A phase to maximize the signal in the I_phase - see coherence in top panel.) In the lower panel I've used a calibration factor to overlay the two curves. The peaks around 1kHz appear in both signals, which also rules out intensity noise as a source. The peak at 817Hz only shows up in OMC TRANS, not the PDH signal --> still consistent with an intensity noise source.
4) Transfer functions from OMC SUS and tip-tilt SUS to PDH error signal
The only hypothesis that remains is real length noise in the cavity. This idea is supported by Robert's measurements of the ISI, which showed that noise from outside the chamber can be transmitted through the ISI due to resonances at ~1kHz.
We thought that the ISI resonances could be lining up with resonances in the OMC breadboard. Koji had measured a breadboard resonance around 1kHz in bench tests, although the boundary conditions of the suspension points were different from the in-vacuum suspension.
So we tried to excite the modes using the SUS. The fourth plot is a swept-sine transfer function of the OMCS vertical mode to the PDH error signal. There's no coherence at the frequencies of the noise peaks; we believe this rules out the OMC breadboard resonances as the source of the peaks. But at other frequencies our excitation was coherent with the length error signal, and the transfer function is pretty flat; the implication is that the OMC breadboard has a flat response, approximately, to excitations in the SUS, which means that noise from the ISI might couple into the cavity length. Not good.
We also tried to measure the beam jitter coupling in the same band, using a pitch excitation on OM3. This is the fifth plot, and the story is the same - coherence in some bands, but not at the peaks in the noise spectrum. Also not good. (It wasn't clear to us why beam jitter noise was coupling into the PDH error signal. The cavity was locked using the length dither with a 30Hz UGF, so it's hard to imagine jitter at 900Hz being impressed upon the cavity length via the dither servo.)
5) Encore noise investigation: IMC DOF4 error point and sideband detuning
Right at the end of the day I was setting up a measurement of the IMC cavity pole. With the OMC locked, we noticed that changing the frequency of the 9 and 45 MHz sidebands slightly off the IMC resonance would cause the OMC TRANS signal to drift down. Eventually we realized that moving the sidebands by 20kHz changed the reflected power from the IMC, which moved the error point for the DOF4 loop and slowly misaligned the IMC. See the sixth plot. This isn't really surprising, but it was pretty confusing for a few minutes. (The many dropouts in the plot are due to the ISS becoming unstable - these show up in the DC power signals but not in the ASAIR_A_RF45_I_ERR.)
h1calcs: Dave, JeffK, JoeB
WP5016. This install was postponed from today, the model is going to be reworked. WP is open through next Tuesday.
ETM SUS FE reboot tests; JeffK, Jim, Dave
Following some problems starting the ETM suspensions front end computers during last week's RCG update, we tested this further today by rebooting h1susex and h1susey from a powered down state. The procedure was: stop user models, stop iop model, take out of dolphin fabric, power down computer, via IPMI power computer back up.
We did not get a clean restart of all models and awg_tp_man processes. The restart problems were not reproducible on both end stations. Some manual intervention was needed. Investigation is continuing.
Power cycle of DAQ Solaris QFS writers: Jim, Dave
the frame writer spontaneous restarts have become more frequent since last week's RCG2.9 upgrade. Today we power cycled both QFS writers in the MSR to see if this will make the DAQ more stable. The procedure was (for both h1ldasfw0 and h1ldasgw1 systems, staggered restarts so one framewriter is operational at all times):
on nds: stop monit, stop daqd process, stop nds process, un-mount ldasgw NFS mounted file system
on fw: stop monit, stop daqd process, un-mount ldasgw NFS mounted file system
on ldasgw: power down with "poweroff" command. Power up by pressing front panel button. Manually mount the QFS file system, manually share the file system
on fw: start monit. Verify daqd runs and writes frame files
on nds: start monit, verify daqd and nds process startup.
The only surprise was when h1ldasgw0 was powered back up, it started a continuous hardware diagnostics which I manually killed after 10 minutes of running. h1ldasgw1 did not do this.
We will know if this has helped in a few days of running.
DAQ reconfiguration, Dave:
I created the latest Beckhoff slow controls INI files and restarted the DAQ after the above solaris reboots had completed.
OS updates, Dave:
all servers accessible from outside of CDS were upgraded to the latest OS patches. cdslogin and cdsxegw0 required a reboot to complete. cdsssh was rebooted yesterday to complete.
Filter module FE load status, Dave, TJ:
The following models had partially loaded filter files and were fully loaded from the GDS-TP screen: h1sussrm, h1oaf, h1lsc, h1omc, h1iscey.
The following models have modified filter module files which have not been loaded, further investigation is needed (modifcation time is given): h1isiitmx (Tue 13jan2015 16:16), h1susmc2 (Sat 17jan2015 17:04)
FE code with local modifications pending SVN commit, Dave:
Attached file shows local mods to model mdl files.
inicheck fix: Dave, Jim
We found inicheck was reporting a warning if a channel's UNITS setting was different from that in the default block, problem arises from RCG2.9 upgrade. Jim has fixed this.
Rana, JeffK, Kiwamu,
We did some maintenance items on IMC ASC. The major activities are:
Also, tomorrow or sometime this week, we might insert a lens in the REFL gigE path as the REFL beam now is a bit too big to be able to cover the spatial higher order modes.
Some issues with the current implementation:
Tonight we have twice had MC lock losses where MC1+MC3 got completely misalinged, although they didn't trip. The second one of these was at the same time as a lockloss of the ALS, but we don't know if the IMC caused it or ALS. This was at 6:35 UTC, Jan 21st.
We made the decision not to increase the spot size. This was done at LLO to decrease the jitter coupling into the ISS. The in-vacuum ISS doesn't require this.
The IMC seem to have worked much more reliably, before we added DOF4. Please turn it off for good.
Alexa, Evan, Dan, Sheila
We have been having intermittent problems for the last two or three days. This evening we traced the problems we've been having with ALS COMM (alog 16129 ) to an oscillation of the PSL noise eater. The tell tale symptom was amplitude noise at around 900 kHz on the PSL light. We don't know of a good indicator of this problem from the control room.
We do not know if this was the cause of our mode cleaner lock losses over the last few days (alog 16128 ), or to tripping of MC1+MC3 suspensions and HAM2 ISI.
After I toggled the noise eater switch the ISS first loop was unable to lock. For now we have turned it off.
We had the outputs held on the IMC WFS DOF4 from late last night until 8 pm today. We didn't see any trips of MC1+MC3 today. Now we have turned DOF4 back on.
I turned on the ISS first loop. For the OMC characteization, we needed some kind of ISS.
1. Changed REFSIGNAL (H1:PSL-ISS_REFSIGNAL) from -2.248 to -2.135 to match it with H1:PSL-ISS_PDA_AVG
2. Push "On" of AUTOLOCK
This allowed me to engage the ISS loop. The out of loop "lsd" monitor (H1:PSL-ISS_PDB_LSD) shows 1.2e-8/rtHz.
There was recent check of the Noise Eater mon at LLO (log 13353). Wasn't that useful, but the binary NE mon is supposed to tell us when the NE loop is oscillating.
There were also numerous instances of this during eLIGO; the 'solution' then was to turn the servo OFF and then ON. Maybe if the monitor is now mistuned, we should adjust the resonant circuit to operate at 900 kHz.
Apparently we do have at least two ways to tell this is happenening from the control room, which means we could have some automated error checking for it.
First, this was already done in the PSL ODC, which seems to have degraded at least at LHO. (alog 9674) The PSL ODC screen now looks like the attached screen shot, I don't know what hapened to it but it might be helpfull to restore it.
The second screenshot shows that the RF mon on the COMM PFD was around -1dBm even when the X arm was unlocked while the noise eater was oscillating. We can add an error check for this in the COMM PLL beckhoff code. This is similar to what we did for the end station lasers (alog 10273 )
The attached plots show 2 weeks of the PSL Noise Eater channels as well as the ALS COMM demod mon.
NPRO_NEMON doesn't show any change and I don't know what it is connected to.
NPRO_RRO is the binary indicator of whether the NPRO Relaxation Relaxation Oscillation monitor is indicating a high noise state: around -5800 means OK, around -300 means Oscillating.
COMM DEMOD RFMON shows the non-bandpassed RF noise (in units of dBm):
-35 dBm corresponds to the bare noise on the laser without the arms locked
-1 dBm seems to be what we see with the arms unlocked an the NPRO NE oscillating
+5 dBm corresponds to the X arm locked and there's a good beat note between the green PSL and the green X trans beam
* the RRO indicator on the PSL screen had the threshold set too high; I've changed it to now change from green to red at -2000 counts rather than -200 (which would make it always show GREEN)