This morning around 8:30am, Nutsinee and I went out and restarted the TCS chillers and the TCS lasers. We also took the opportunity to swap the greenish filters. Recall, we had flushed both chillers last week, but when we installed a new set of filters we noticed that they were tinging green immediately even though we had just flushed the system. These filters sat for ~a week and we swapped them today in hopes that whatever immediately turned them green last week was just still pushing throgh the system. We also had to add another 750mL (to top off the half way full indicator to full) and again inspected for leaking water (none found).
We'll need to look up how to clean these recyclable filters.
Found that the dewpoint monitor IOC had frozen, the values on the MEDM screen were white. Killed the processes and restarted the IOC. Since these values don't change much normally, it's difficult to tell when it stopped running. A trend for the last 10 days shows no variation.
Richard restarted all the front end computers after the power glitch. The push button switch on h1susb123 stuck in, preventing the computer from starting. Once that was discovered, the models on the Dolphin network should have started but didn't. Suspecting another computer was down, I started the dxadmin tool on the h1boot server, and found that the dolphin node manager wasn't running. After starting that, the models started coming to life. Models on three computers didn't start, h1susey, h1sush56, h1sush2b, so those were killed and started manually. IRIG-B timing on the h1ioppsl0 model went negative, but recovered within 30 minutes. IRIG-B timing on h1iopasc0 and h1iopsush56 drifted positive, and are working down toward a normal value.
This morning, after the power glitch, I went out to check the settings for the analog communication voltage scaling for the TCS chillers. For both TCSX and TCSY chillers: High Volts: 10.0 Low Volts: 0.0 High Temp: 40.0 Low Temp: 5.0 These are the same as the defaults listed on page D-5 of the Thermo Scientfific ThermoFlex Recirculating Chillers (Deluxe Controller) manual (P/N U00939, Rev. 08/14/2014).
The analogue input voltages were never changed from the factory settings. The settings that were changed were the fault and warning settings for high and low temps. These should have remained after power cycling the chillers though.
The PSL is now up and running after this morning's power glitch. Once the front ends were alive the PSL came up without any issue. I reset the injection locking relock counter.
While it was down I took the opportunity to power cycle the diode chiller. The front panel of the chiller has been scrambled since the chiller started behaving funny (alogs here and here), the power cycling cleared the screen up; the water temperature is still oscillating. We are in talks with the chiller manufacturer about this issue.
TITLE: 09/30 Day Shift: 07:00-15:00 UTC (00:00-08:00 PST), all times posted in UTC
STATE of H1: Power glitched -- Front Ends are down (but Beckhoff seems to have survived). Waiting for people to show up to help with the recovery.
INCOMING OPERATOR: Corey
SHIFT SUMMARY: Kiwamu and Stefan left the interferometer locked and asked me to touch knobs to try to reduce broad bump around 100-1kHz. I gained about 15 Mpc by putting -0.05 offset to CSOFT pitch and CSOFT yaw (details and pretty plot in alog30085) but didn't get rid of the mysterious bump. Terra was here all night dealing with PI.
LOG:
13:08 Power glitched
13:29 Richard to CER
13:31 Richard going to power cycle stuff at end station
14:41 Jeff B. going to EX to check on dust monitor vacuum pump
14:56 Jeff done at EX, going to EY.
14:59 Turn over to Corey
Tasked by Kiwamu and Stefan to turn random knobs in a hope of getting rid of 100-1kHz bump, I was able to gain 15 Mpc by adding -0.05 offset to CSOFT pitch and yaw. I haven't had a chance to try every ASC knobs due to the power glitch. We were at 75 Mpc before the IFO went down.
Here's a note on everything I've tried (all time in PT):
3:40 added/subtracted 0.02 counts to H1:LSC-MOD_RF45_AM_RFSET -- Nothing improved
3:46 Added +6 gain to PRC1 LSC -- 27 Hz line became visible. But later found this line to come and go so this could totally be a coincidence
3:52 Changed CO2X power to 1W -- quickly dropped the power back to nominal when I realized that it's probably going to take a while to see the result
4:38-4:58 [In ASC land] Played with CSOFT pitch and yaw offset -- Settled at -0.05 for both. Slightly improved overall DARM spectrum but made 20-30 Hz wrose. Gained 15 Mpc in range here (if it's not calibration that's fooling with me). Any more negative number added would decrease AS90/POP90.
5:06 Added -10 to CHARD pitch -- nothing improved
5:13 Added -0.07 to CSOFT pitch -- nothing improved
5:14 Added -20 offset to CHARD yaw. -- Nothing improved
5:17 Added offset to DHARD pitch -- Nothing improved
5:27 Added offset to DSOFT pitch -- Nothing
5:48 Added -0.05 offset to DSOFT yaw -- Increased AS90/POP90 but no obvious approvement in DARM.
5:58 Added -25 offset to INP1_Y -- improved AS90/POP90 but no apparent improvement in DARM spectrum
6:08 Power glitch. Meh.
I also moved PR3 and changed something on SRC1 but didn't write down details (since it probably was not doing anything).
I put everything back in place before the power glitch except for the CSOFT gain that worked. I turned off all the gain buttons that I turned on except for CSOFT P and Y but not neccesary return the gain number back to what they were. I have them all written down though.
Continuing our look at the test mass drumhead modes, I've set up damping filters for
MODE5 ITMX Drumhead 8162 Hz , MODE12 ITMY Drumhead 8161 Hz
Was able to predict which drumhead frequency corresponded to which test mass based on frequency shift during previous night's 4 hour lock: known ~15kHz modes of ITMX shifted 0.09 Hz over the four hours, ITMY shifted 0.13 Hz over the four hours. Accordingly, 8162 Hz shifted 0.07 Hz and I was able to ring up on ITMX. 8161 Hz shifted 0.11 Hz and I was able to ring up on ITMY. Torsional modal frequency is proportional to sqrt(G/p) and longitudinal modal frequency to sqrt(E/p) where G is shear modulus, E is Youngs modulus, and p is density (see P080069).
We had minimal PI issues through the night; most PI related locklosses over the past week could've been prevented with simple early-action small phase changes and tonight reinforced that. Jim and Nutsinee were both successful at damping; I fixed Jim's earlier problematic MODE17 with a filter change and the rest of the 4.5 hour lock went smoothly, requiring only one phase tweak a few hours in. We lost lock from the power glitch.
We just had what appears to be a site-wide power glitch at 13:08 UTC. All front ends down. Richard is here and starting power cycling.
- Tried switching the guardian back to using REFL WFS instead of POPX WFS for PRC2. This is simply a boolean variable change in lscparams.py: use_popx_wfs = 0. We also updated the the beam diverter closing state to also use the same boolean variable.
- However, we need to retune the REFL WFS at the new modulation index - we saw the loop run away in Nominal Low Noise,
- We also tried to repeat the jitter coupling reduction with INP1 offsets. While the transfer function again was reduced, the ASC loops were having problems with the large offset - we saw CSOFT ring up at 1Hz. Also there was a 3Hzish signal visible in CHARD and CSOFT. So much for reproducability...
- Next we also tried a CHARD YAW offset. Indeed -3500ct in CHARD YAW started bringing down the IM4 YAW jitter coupling. As suspected this probably means we simply cancel an unknow coupling with a deliberate misalignment induced coupling.
I've updated the PI Damping Operator Wiki with step-by-step instructions for damping PI's. These instructions cover basic damping for modes we already know about and repeat exactly what I've worked through with almost all of the operators by now. They should be enough to walk you through damping most modes that might ring up during a longer lock. We've also added a verbal alarm to alert you if a PI has started ringing up.
Wiki can be opened from LHO Ops Wiki ('PI DAMPING') and from PI Overview MEDM screen:
Many thanks to Nutsinee for the awesome edits.
PSL trip took down the IFO, or vice versa. No commissioners are on site currently, so I'm calling it a night (with Corey's approval). Stefan and other commissioners made noise about coming back in so they try to recover, but I don't have the procedure to restart the PSL.
TITLE: 09/29 Day Shift: 23:00-7:00 UTC
STATE of H1: PSL down
SHIFT SUMMARY: Lock was generally going well until the PSL went down, PI's were a learning experience, vigilance is required.
LOG:
23:00 IFO almost locked at end of Corey's shift
1:15 DRMI flashes non-existent, did initial alignment through MICH_DARK
2:00 NLN, PI Mode 26 immediately started ringing up, but tweaking phase brought it back down.
3:30 4 PIs ring up, 17, 25, 27 & 2, breaking lock before I could make any improvemts on any of them, but 17 rang up first
3:50 NLN, PI 17 rang up around power up, but managed to bring it down again
4:47 Lockloss, PSL is down
I have restarted the PSL. Looking at the status screen of the PSL, I have determined that the situation was exactly the same as those from yesterday (30063). However, because it was already late in the night, I decided not to call the experts. Instead, I by myself did the same recovery procedure as I did yesterday. I added 200 ml water to the chiller.
Thanks for handling PI's Jim. You saved a few locks.
Got in right as Jim was leaving. Looking back at the spectrum, it was only MODE17 that rang up during the 3:30 time; all other modes just had increasing RMS due to bleed over from the extremely heightened mode (note MODE27 did not ring up, it was MODE10 - I've changed the color to make this more distinguishable). This one had not been problematic since I implemented stepping filters and a smarter guardian to handle (several days now).
At 3:50 power up, MODE17 was still significanly rang up (since Jim relocked so quickly), hence seeing it right away. If lock is lost due to PI and you immediately relock, pause at DC_READOUT before power up to ensure mode is damped first. I'll go through this with operators.
Carlos, Dave:
h1fs1 kernel panic'ed apparently during its 08:39PDT backup of h1fs0. We rebooted via power cycle and manually got the ZFS file system back into sync so the hourly cronjobs could continue to run.
TITLE: 09/29 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Aligning
INCOMING OPERATOR: Jim W.
SHIFT SUMMARY:
OK shift. Had issues locking in the middle of the day (completely dead DRMI), but Sheila tracked this down to a misaligned PRM (Stefan alogged this). After that, took H1 up to a state for the Commissioners, and that is where they are at right now.
LOG:
Updated Nuc6 to so the PI Strip Tool has Terra's new mode.
(Betsy, Gerardo)
This morning added 500 ml of water, and walked/crawled the length of the pipe looking for water drops, none found. The inside of the TCS-Y table was not inspected.
model restarts logged for Wed 28/Sep/2016
2016_09_28 11:38 h1nds0
2016_09_28 13:29 h1nds0
two unexpected nds0 restarts, hopefully memory upgrade will fix this.
Is this on top of the 500ml added yesterday? https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=30074
Added scales to the sides of the fill tubes. This will allow better monitoring of the chiller water levels. The TCS-X scale runs from 21 - 30, linearly, and the TCS-Y scale runs 0-10, linearly.
I have just opened and inspected the TCSY table - All water lines look good and no puddles found. The tubes also appear to be pretty solid with not many air bubbles.