Wanted to get H1 to atleast DRMI before today's Press Conference, so jumped into locking first thing this morning, but no DRMI locks, and PRMI would not stay lock for more than a second, so opted for an Initial Alignment, but it wasn't trivial. A few notes from locking & alignment:
Another IFO Note: After the alignment, Krishna mentioned conditions were quiet so, we took the ITMs to the 90 blends and NO Sensor Correction.
BSC ISI Guardian Note:
Krishna noticed that Sensor Correction was turned off for ETMy & ETMx ISI's at around 1am last night by taking a gain from 1 to 0 by hand. We want to get in the habit of not doing this (because Guardian does not monitor these gain channels -at the moment-), and use our fancy new Guardian medm (sitemap/O-1/ISI Blend Filters.adl)
Daily activity log attached as pdf.
I checked the PSL chillers, per FAMIS work order 4155, assigned to me today, and found the chiller was filled yesterday, so did not need any water.
First post is in the ground and concrete is curing.
1505 - 1520 hrs. local -> To and from Y-mid Opened exhaust check-valve bypass valve. Opened LLCV bypass valve 1/2 turn -> LN2 at exhaust in < 1 minute -> Restored valves to as found Next over-fill to be Friday, June 17th.
Bad PEM corner station MAG and MIC channels
Richard, Fil, Robert, Dave:
We found that all the MAG and MIC corner stations channels have not been correct since the power recovery Mon 6/6. These channels span several ADC cards, so it did not look like a software issue. However to rule this out we power cycled h1oaf0, its IO Chassis and the PEM AA chassis with no change to the signals. Problem was identified as two DC power supplies which needed to be turned on after the outages (18V for MAG, 36V for MIC). Robert confirms these channels are now good. We are still unsure why all the MAG and MIC channels read about 3000 counts on the IOP and not zero, we may investigate further next Tuesday.
h1calcs new code
Joe, Jeff, Dave:
new h1calcs model went in. Some channels were promoted from standard single precision floating point to the new double precision floating point data type. The RCG appends the suffix "_DBL_DQ" to these channels, compared with "_DQ" for standard chans. The channels changed are:
[H1:CAL-DARM_CTRL_WHITEN_OUT_DBL_DQ]
[H1:CAL-DARM_ERR_WHITEN_OUT_DBL_DQ]
[H1:CAL-DELTAL_CTRL_DBL_DQ]
[H1:CAL-DELTAL_CTRL_PUM_DBL_DQ]
[H1:CAL-DELTAL_CTRL_TST_DBL_DQ]
[H1:CAL-DELTAL_CTRL_UIM_DBL_DQ]
[H1:CAL-DELTAL_RESIDUAL_DBL_DQ]
Initially H1:CAL-DELTAL_EXTERNAL was made double, but this meant the DTT display on the control room TV stopped working, this channel was then reverted to single precision.
ETMY Hardware Watchdog
Jim, Dave:
To verify the HWWD software is running, I pulled out the top monitor cable on the front panel of the EY HWWD unit. The LED fault light came on, showing that the software did start after the power outage.
DMT broadcaster channel configuration
Joe B, Maddie, Dave
Joe and Maddie requested some channel additions/removals from the DMT channel list, these were installed.
QUAD_MASTER change
Jeff
h1susetmx,y were restarted against a change to the QUAD_MASTER.mdl file. Next Tuesday QUAD_ITM_MASTER.mdl will be modified and the itm models restarted.
Resync of atomic clock to NTP GPS receiver
Jim:
Following the power outage, Jim re-synchronized the atomic clock to the 1PPS signal coming from the NTP GPS receiver. This provides an independent timing of the Timing Master.
The Timing Solutions time code generator in the MSR has been resync'ed to the Symmetricom NTP server following the power outage of June 4,5. The following steps were performed: Sync the Symmetricom 4310B Cesium Frequency Standard to the Symmetricom S350 NTP Server. Attach the 1PPS output of the Symmetricom S350 NTP server to the Sync input of the Symmetricom 4310B Frequency Standard (aka atomic clock). Attach matched coax cables from the Sync input of the 4310B and the 1PPS Out of the 4310B using coax T's to an oscilloscope to measure the difference between the two signals. Difference between sync pulse and 1PPS out was 178mS. Send sync arm command to the 4310B via the RS-232 input to the 4310B. Observe that the difference between the sync pulse and the 1PPS out of the 4310B is now in the range of 30nS. Note that the jitter of the S350 is much greater than the output of the 4310B, so precise measurement is not really possible to the 1nS range. Also, it's only possible to sync the 4310B to a 1PPS input to +/- 100nS. Sync the Timing Solutions timing distribution system to the 4310B. Next, attach the 1PPS out of the 4310B to the Timing Solutions TCG 1PPS Sync Input. Using the oscilloscope, measure the difference between the TCG 1PPS Sync Input and the TCG 1PPS Output. This difference was in excess of 430mS. Adjust the slew using the Timing Solutions MCA user interface to bring the difference as small as possible. This involved several iterations as it is only possible to adjust the slew by a maximum of 100mS for each adjustment. Also note that the minimum adjustment is 10nS. The ending best difference obtained was 4nS between the 4310B 1PPS output and the TCG 1PPS output. The 4310B 1PPS output is connected to the TCG Advance 1PPS input to monitor coincidence to +/- 20nS between the 4310B and TCG 1PPS. No adjustment of the Timing Solutions TCT units at the end station is needed, as the only adjustment possible is to compensate for the length of the optical fiber between the corner and end stations which is a fixed value. The 1PPS output of the Timing Solutions TDS is connected to the corner station timing comparator input 1 while the Symmetricom S350 NTP server is connected to input 2. The MEDM screen for the timing comparator shows the difference between the timing master 1PPS derived from a separate GPS receiver and the various inputs.
Subsytem Updates
REMINDERS:
Progress on the E X Wind Fence. We will continue work on this project today also.
Carl, Terra, Rich A.
1. We used the never-before-tested LVLN ITMY ESD driver to ring up and damp two mechanical modes of ITMY, 15072 Hz and 14979 Hz. Below is the amplitude evolution of the 15072 Hz peak as we rang it up, allowed it to ring down naturally, rang it up again, and then damped it down fully with a gain sign flip.
We drove and damped similarly for the 14979 Hz peak. 15072 Hz is the vertical differential drumhead mechanical mode; 14979 Hz is the horizontal version. For both cases, we tightly bandpassed the H1:OMC-PI_DCPD_64KZ_A/B signal, added a +60deg damping filter, and added gain to the damping filter until saturation. Positive gain drove up, negative damped down.
I've fit the natural ring down with f(x) = a * exp(bx), where tau = - (1/b). Then Q = pi * 15072 * tau = 31.5 million.
2. Interestingly, we realized after the above tests that we had not turned the ESD bias on for either ITM. After turning on both to 100K cts (to DC offset), we just had time to ring back up the ITMY 15072 Hz mode before a lockloss. Below is a comparison of the ring ups (note we lost lock and did not damp for the ring down portion below). Green trace is the 15072 Hz ring up without bias, blue trace is with bias (time shifted for ease of comparison).
A slight slope difference is visible but we'll look into this more. Thoughts are that we could use the difference in responses to measure test mass charge coefficients as discussed here and here.
3. We turned ring heaters off and on for future mode-mass identification analysis. For the record (since RH messing with violin modes was a concern tonight):
We were also driving the ITMX mode at 15077Hz before and after the bias was turned on. The data is a little more confusing. As the phase was being varied before hand to try find the optimum damping phase. In the plot the amplitude is made to coincide when the damping phase with no bias had the largest response. There was no attempt to optimise the phase in the case where the bias was on other that to try the positive and negative of the previous 'best' phase. Interesting points are:
For the 15077Hz mode the phase that excited the mode was flipped whent he bias was engaged.
The response with the bias engaged is relatively larger when compared to the ITMY 15072Hz mode.
For the 15077Hz mode the amplitude response with bias is about twice the amplitude response without bias.
Mode identification.
In the attached plot the relative change in frequency of the four likely drum head modes around 15200Hz is plotted as a function of time. The ring heaters were adjusted as follows:
ETMX 0.5W per segment to 0W 02:29
ETMY 0.5W per segment to 0W 03:00
ITMX 0W to 0.5W per segment 03:33
ITMX 0.5W per segment to 0W 04:02
ITMY 0W to 0.5W per segment 04:06
ITMY 0.5W per segment to 0W 04:36
The response in frequency shows that the 15218Hz mode is a cooling ETMX, 15219Hz is a cooling ETMY, 15197Hz a heating ITMX and 15192 a heating ITMY mode.
As we were only operating at 2W input the signal to noise ratio of modes is a lot lower and many fewer modes are visible compared to Livingston measurements T1600141.
The following is the list of modes identified.
Measured Frequencies | Simulated Frequency | ||||||||
ITMX | ITMY | ETMX | ETMY | ITMX | ITMY | ETMX | ETMY | Shape description | |
6044 | 6042 | 6055 | 6054 | 6057 | 6054 | 6053 | 6053 | Butterfly | |
8162 | 8160 | 8161 | 8158 | 8194 | 8190 | 8189 | 8188 | Drumhead | |
9812 | 9809 | 9829 | 9831 | 9827 | 9827 | 9832 | 9832 | Drumhead | Vertical |
9860 | 9863 | 9881 | 9882 | 9879 | 9881 | 9885 | 9885 | Drumhead | Horizontal |
10415 | 10437 | 10434 | 10433 | 10432 | 3 point ripple | ||||
10423 | 10419 | 10428 | 10426 | 10462 | 10460 | 10463 | 10462 | Drumhead | |
12992 | 12988 | 12999 | 13000 | 13022 | 13026 | 13035 | 13036 | Drumhead | Differential |
15077 | 15072 | 15093 | 15096 | 15102 | 15103 | Drumhead | Horizontal | ||
15197 | 15192 | 15218 | 15219 | 15227 | 15227 | 15232 | 15231 | Drumhead | |
15540 | 15545 | 15541 | 15544 | 15544 | Drumhead | Vertical | |||
15627 | 15635 | 15632 | 15634 | 15634 | Drumhead | Horizontal | |||
19555 | 19553 | 19589 | 19588 | 19595 | 19594 | Drumhead |
Stefan, Nutsinee
A very quick summary: ITMY modes rung up tonight. All the damp settings has been changed. I suspect ETMX damp phase might have been changed as well. ITMX didn't seen to have a problem. We suspect the ring heater changed the violin mode frequency.
Below I attached a screenshot of what works on ITMY modes tonight.
We looked at the frequency of mode 8 last night, two nights before around GW151226 and after a O1 power ouitage (20151025 9UTC). (attached)
Two days ago the mode was heavily damped, so the frequency estimate might be a bit off. At any rate, all frequencies are within 3mHz (or 22deg of feed-back phase in the current band-pass filter). So we ough to be able to damp this mode consistently.
Chandra, Gerardo -Adjusted CP5 Dewar pressure regulator 1/4 turn CW to raise Dewar pressure (nominally at 15 psi). Will wait hours/days to see results. -Measured Dewar vacuum jacket pressure = 20 microns, satisfactory. -Magnehelic readout is noisy - jumps from 35-40 in. of H20, even when CP5 is in manual mode. -Manually tested full range of stroke of electronic actuator - looks OK. -Tightened actuator support/brace nuts along unistrut. -Noticed needle of valve has wiggle (more than CP6). Verified packing nut is adjusted properly and coupling nut and needle are fully threaded. So now we wait to see if adjusting pressure of Dewar has an impact.
Not able to trend - was sitting around 100% but maybe just big swing?
Sheila, Haocun
We had lots of locklosses today, and some were probably caused by the side OSEM driving of the beamsplitter M1, as shown in the figures attached.
(H1:SUS-BS_M1_MASTER_OUT_SD_DQ & H1:SUS-BS_M1_DAMP_T_IN1_DQ)
It seems to have happened again at 22:20:12 UTC June 15th
Tagging SUS and CDS.
Tonight I had a look at several other of these locklosses, and other than the one at 23:57:59 that Hoacun plotted, it seems mostly like the glitch in the side osem comes after the lockloss. Many of these locklosses seem to happen durring the CM board switching.
It does seem like it would be worth investigating this side osem issue, since we know that it causes at least some locklosses.
J. Kissel WP #5932 Today's model modifications come in three parts: (1) Independent Synchronized Oscillators for each isolation stage of the QUAD for calibration lines (2) Updated CAL-CS model with better demodulation scheme that spits out coherence and uncertainty (3) Changed all CAL-DELTAL and CAL-DARM channels to double precision and removed whitening Below are all the details. ------------- Part (1) In order push the PCAL vs. SUS actuation calibration line cancelling scheme for ER9/O2 forward (see T1600218 and T1600215), I've installed independent, longitudinal, synchronized oscillators in each of the UIM/L1, PUM/L2, TST/L3 stages of all of the QUAD models. See first attachment for screenshot of the QUAD OVERVIEW screen that now has the CAL_LINE block beneath every LOCK bank. Sadly, due to the recent split in QUAD library part models, I'd forgotten to add the needed EPICs channel for GPS synchronization to the ITMs. We'll install them tomorrow morning. I've committed the following new models: /opt/rtcds/userapps/release/sus/common/models/ Sending FOUROSEM_DAMPED_STAGE_MASTER_WITH_DAMP_MODE.mdl <-- Used in PUM/L2 stages Sending FOUROSEM_STAGE_MASTER_OPLEV_TIDAL.mdl <-- Used in UIM/L1 stages Sending QUAD_MASTER.mdl <-- Added EPICs record for GPS sync time on TST/L3 stage for ETMs and accordingly modified MEDM screens: M SUS_CUST_QUAD_ITM_OVERVIEW.adl A + SUS_CUST_QUAD_ISTAGE_CAL_LINE.adl D SUS_CUST_QUAD_L3_CAL_LINE.adl M SUS_CUST_QUAD_OVERVIEW.adl ------------- Part (2) In addition, as a continuation of the improvements to using those calibration lines to track changes in the DARM loop parameters, I've followed Joe's instructions (see LHO aLOG 26491) and updated the CAL-CS model to demodulate PCAL and DARM error signals at given frequencies and produce parameter estimates, coherence estimates, and even uncertainty estimates. Darkhan will be working on commissioning the infrastructure with Joe. ------------- Part (3) Finally, I've converted all of the calibrated DARM channels to double precision. *EXCEPT* for DELTAL_EXTERNAL, which is needed to display on the wall in the control room via DTT (which sadly, cannot yet handle double precision; see CDS Bug 1003). As a result, and as per Joe's advice I've turned OFF all dewhitening filters that condition the following (new) channels stored in the frames: H1:CAL-DARM_CTRL_WHITEN_OUT_DBL_DQ H1:CAL-DARM_ERR_WHITEN_OUT_DBL_DQ H1:CAL-DELTAL_CTRL_DBL_DQ H1:CAL-DELTAL_CTRL_PUM_DBL_DQ H1:CAL-DELTAL_CTRL_TST_DBL_DQ H1:CAL-DELTAL_CTRL_UIM_DBL_DQ H1:CAL-DELTAL_RESIDUAL_DBL_DQ We need to investigate if we're now happy with the signal fidelity without whitening or if we just need to re-assess the whitening. Again, I've retained single precision on H1:CAL-DELTAL_EXTERNAL_DQ, so its whitening filter remains ON.
J. Kissel, J. Betzwieser Joe called to let me know that I'd misinterpreted the fixed oscillator part that we chose to use (of the two; see T1600143). Turns out this version of the fixed phase oscillator automatically synchronizes to GPS time 0 (i.e. Jan 4 1980), so there's no need for the extra EPICs record input to the oscillator that specifies the GPS time. So, in opposition to what's said above, the ITM models are fine; we need to take the GPS EPICs record *out* of the ETMs. Will do so tomorrow.
J. Kissel All QUAD Models have been restarted sans EPICs record for synchronization (the ITMs needed a restart to absorb the removal from the L1 and L2 stage common library parts). MEDM screen has been updated as well. I've committed the following to the repo. /opt/rtcds/userapps/release/sus/common/medm/quad/ Sending SUS_CUST_QUAD_ISTAGE_CAL_LINE.adl /opt/rtcds/userapps/trunk/sus/common/models/ Sending FOUROSEM_DAMPED_STAGE_MASTER_WITH_DAMP_MODE.mdl Sending FOUROSEM_STAGE_MASTER_OPLEV_TIDAL.mdl Sending QUAD_MASTER.mdl
New today is the the SEI_CONF Guardian node that will manage all of the other configuration nodes that we have made the past few weeks. This will hopefully make it much easier to choose a configuration for all of SEI. Sensor Correction for the entire SEI can be turned OFF from here, and then be brought back, of course.
The states and their names are very likely to change in the near future. So keep an eye out for updates, and we will also try to keep documentation up to date.
I also move the GUARD_OVERVIEW.adl around a bit since it was getting crowded in spots.
I changed the Blend_PAGE_MAIN.adl to include all of the config nodes and removed the SC ON/OFF switches. I have also included a guide to help choose states, but take it lightly because I'm sure it will change again soon. The right corner shows if the MATCH bank gains are 1 (green) to allow for sensor correction (Enabled) in that chamber.
Carl, Terra, Ross, Tega
Tonight we used the freshly installed LVLN ITMX ESD driver to ring up and damp two mechanical modes of ITMX, 15063 Hz and 15077 Hz.
After sorting out some phase settings, we drove the ITMX ESD close to saturation in a differential drumhead pattern. Negative gain rang up 15063 Hz. Flipping the gain sign to postive then damped this mode and rang up 15077 Hz. The amplitude plateaued as a the saturated drive signal approched a square wave. Figure below tracks amplitude of 15063 Hz and 15077 Hz (seen in OMC trans), with gain sign flip occuring around the 0.15 time mark.
Also attached is a spectrum of H1:OMC-PI_DCPD_64KHZ_A_DQ during no gain, negative gain, and positive gain times, i.e. on either side of the 0.15 time mark from the plot above.
At about 0.23 hours the gain was turned off and the mode rangdown. The fit to this ringdown indicates the mode has a Q factor of (omega_o)/(2alpha) = 1.2 million.
Settings: Power 1.9 W, DC bias 100k, butterworth BP filter, iWave bypassed, -60deg damp filter, damp gain 300,000.
Such a wonderful conclusion to the installation and commissioning of this system. Much thanks for the great support I received from all involved.
I suspect that 15077 Hz mode is an aliased 48923 Hz mechanical mode (64 - 15.077) kHz. The FEA gives an interesting mode at 48944 Hz (mode shape attached). Observation of the analog channel PSD on transmission is required to confirm if this is a case. The 15077 Hz mode is only 14 Hz above known 15063 Hz mode. I am not very familiar with linetracking filter but I assume that the two resonant lines 15063 HZ and 15077 Hz cannot be sufficiently separated and the signal with higher resonant peak will be eventuality phase-locked. It may be interesting to observe the transition from one mode to another. If this all is true you measured the Q-factor of one of the modes in the range were PI may also show up at higher circulating power than during O1. It looks to me that the 48923 Hz mode is very sensitive to off-center position of optical TEM00 (that's probably why can be seen on OMC) and can be used for centering of TEM00 on ITMs.
In the Q estimate 'f' was used in place of 'omega_o' introducing a 2 pi error. The corrected estimate of the Q factor of the 15077Hz mode is 7.4million.