As a follow on from my alog last week (16811), I ran TFs on the PR2 while the largish IFO pointing offsets were enabled. With offsets of P = 1816, Y = 4315, the P and Y TFs look healthy. Phew.
With some guidance from T. Abbott, I have gotten the SUS Drift Monitor back up and functioning. I have updated all the suspension alignments to the most recent good lock stretch of 1108702216 GPS. After this update, a few suspensions were showing that they are misaligned from this previously good alignment.
PR3 - Pitch, not sure why this is aligned differently now. Trends show this changed Feb. 24 at 20:00 UTC. I could not find an aLog specifically stating that this SUS was touched.
SR2 - Changed Pit alignment to alleviate possible rubbing. See aLog 16831.
RM, IM, and OM alarm level are still in a state of flux. To be continued.
No significant changes from last week.
The version of root software used on x1work has been updated to root-5.34.21, the version of gds software for x1work has been updated to gds-2.16.12.4, to match what is used in the LHO control room.
The daqd and nds software for x1dc0, x1nds1, and x1fw1 has been updated to branch 2-9, r3965. It was fairly old, at r3626 and needed to be updated to test daqd protocol 12.1 with nds2 client software.
Since we started running feedforward on the BSC-ISI's, Jeff and Hugh noticed that the feedforward path is not controlled by Guardian. This means that if FF is engaged and the ready state is requested through Guardian, the St1 actuators will still be driving. From a safety standpoint, it looks like this probably isn't an issue. I took spectra this morning with the ITMX ISI off (FF off, too) and with FF only (no damping or isolation loops). The FF only state actually still provides some isolation above 1hz (see attached plot, dashed references are the off state, solid lines are FF only). I also turned HEPI off, then turned it back on while watching the ISI seismometers, and didn't see anything obviously terrible. The actuator signals stayed below a couple hundred counts the whole time, much like the ISI with damping only, and only the T240's got agitated while HEPI servoed to position, which is normal. It's still possible that whacking the HEPI piers could cause FF to do bad things, so...don't do that.
See JeffK's 16857 where he analyses the End Station's performance relative to the fluid pressure noise. The attached plots compare those final performance numbers for YEND from Sunday at 6pm pst with data taken 25 Feb at 0230utc -- after the EndY HEPI plumbing system maintenance detailed in 16899.
Bottom line--The DOFs still showing coherence on Sunday (Jeff's 16857) are all almost completely gone, there is still a bit of coherence in the HP dof in places between 10 & 100mhz. So this maintenance would appear to be very important. The DOFs remaining were the 'common' mode dofs. That is, the Z and the HP modes of the HEPI--those dofs where the Actuators move in common. All other dofs have counter moving actuators which one might expect a common noise like pump pressure to cancel out.
Now, understand, Jeff's work on the controller characterization and tuning provided most of the decoupling to the platform; this maintenance work just closed a few final loose ends.
The xml for these is /ligo/svncommon/SeiSVN/seismic/HEPI/H1/Common/2015-02-25_H1HPI_EndYStation_Coh_wISIT240s.xml
I spent some time tonight attempting to adapt the guardian codes to the new LSC and OMC models (alog 16893, alog 16904 ). I went through most of the guarduans by actually running them with the interferometer and tried fixing as many bugs as possible. Here is a summary of the guardian situation (red=bad, blue=OK):
In addition, a couple of associated parameter files (e.g. ISC_library, pdstep and etc.) were also edited for the same reason. Even though I could test most of the necessary parts, one should still pay attention to the codes as it is very likely that I have missed some bugs in there.
(For the ISC experts)
I have several items to mention.
Looking at some of the LSC demodulated signals, I found that the RF phases changed by some noticable amount. I then found that the CO2 laser on ITMX had been tripped at 13:51 local (or 21:51 UTC) today. I went to the rack at the floor and saw that a red indicator for 'TEMP' was on at the front panel. Power cycling the unit with the key untripped it. I pressed the red button for opening the gate and confirmed that the laser power was back to the nominal of 210-ish mW in the medm screen. The attached is a one day trend of the laser power, calibrated into the power at ITMX.
I did not do any deep investigations at this point.
ISC/SUS Code Changes
Daniel, Jeff, Kiwamu:
During the maintenace period the following models were restarted for new code: h1lsc, h1omc, h1alsex, h1alsey, h1asc, h1susetmx, h1susetmy, h1susitmx, h1susitmy, h1oaf, h1susbs. LSC to OAF Dolphin IPC channels were removed from the LSC, Daniel renamed the IPC channels being used by the OAF to their CAL equivalent channels.
DAQ Restart
To support the code changes and the new guardian system several DAQ restarts were performed.
Guardian Reboot
Jamie and Dave:
13:34PST the h1guardian0 machine was rebooted to test its autostartup. No problems were found.
RFM Diagnostics
Daniel, Jim, Dave:
To test the remaining IPC errors on the ETM SUS models, an ISC model (we chose PEM) was temporarily changed to read the same RFM IPC. The ISC receiver shows zero errors, so the problem appears to be within the SUS front end only. Investigation is continuing.
We perfomed some RFM tests on the LHO DTS, verifying the RFM loop transit direction.
[Stuart A, Betsy W] After earlier using the front-end to take a safe SDF snapshot of the Beamsplitter Suspension (see LHO aLOG entry 16886), thus stripping out the redundant read-only channels and alarms, I used Jamie's script to set all channels to be monitored (see LLO aLOG entry 15907) i.e. sdf_set_monitor 1 h1susbs_safe.snap. I've since manually edited the SDF snapshot setting channels we do not wish to monitor, which are essentially consistent with those reported at LLO (see LLO aLOG entry 15987). Finally, I've checked the safe SDF snapshot into the svn: /opt/rtcds/userapps/release/sus/h1/burtfiles/ M h1susbs_safe.snap Need to monitor for changes in the BSFM configuration as commissioning/locking continues overnight...
Added 352 channels. Removed 302 channels. There are still 7 channels unmonitored: H1:GRD-ISI_ETMY_ST2_REQUEST_S H1:GRD-ISI_ETMY_ST2_STATE_S H1:GRD-ISI_ETMY_ST2_STATUS H1:GRD-ISI_ETMY_ST2_TARGET_S H1:GRD-ISI_ITMY_ST1_LOGLEVEL H1:GRD-ISI_ITMY_ST1_MODE H1:GRD-ISI_ITMY_ST1_NOMINAL_S
They connected, just took a little while.
Kyle, Gerardo Added remaining bolts to HAM1 West door and finished torquing -> Soft-closed GV5 and GV7 (GV5 hard-closed on its own at only 20 psi) -> Pumped HAM1 with scroll pump via vent/purge valve (as such, atmospheric pressure remains on air-side of closed vent/purge valve) -> Opened GV5 and GV7 Richard M. is investigating getting signal cables to HAM1 gauges
Kyle, Gerardo ~0910 hrs. local, Wednesday morning -> Installed temporary pirani gauge controller on top of HAM1 so as to view pressure until CDS cabling gets installed -> HAM1 now showing 0.4 torr which is as expected. In retrospect, I should have not pumped HAM1 without this to monitor -> I had reasoned that the view ports are exposed to 2 atmospheres (gauge) differential (both directions) during bench testing so are considered "safe" -> I now consider the failure mode whereby a view port could have cracked during pumping resulting in a measurable change in the pump down curve -> In this scenario we would not have opened GV5 and GV7 -> Lesson learned
[Stuart A, Jeff K] After yesterday’s preparation and discussion (see LHO aLOG entries 16874 and 16868), this morning we completed the roll-out with the aim of providing AS WFS damping of roll modes, for all QUADs, as follows: (1) LSC, OMC and ASC models were restarted (see LHO aLOG entry 16893), which enabled the necessary IPC/RFM senders. However, there were some channel name changes to deal with... (2) ETMs H1:LSC-ETMX_DARM_ERR renamed to H1:OMC-ETMX_DARM_ERR (same goes for ETMY). (3) ITMs H1:LSC-OAF_DARM_ERR renamed to H1:LSC-ETMX_DARM_ERR (originated from LLO using the calibrated version of DARM ERR for ITMs and raw for ETMs). (4) Svn'd up the common MEDM screens (after remotely committing them from LLO): /opt/rtcds/userapps/trunk/sus/common/medm/quad/ U SUS_CUST_QUAD_M0_DAMP_MATRIX.adl U SUS_CUST_QUAD_M0_DAMP_MODE.adl U SUS_CUST_QUAD_OVERVIEW.adl (5) Removed LLO's Tidal implementation from the Overview Screen (see LLO aLOG entry 16753). (6) Rebuilt, installed and restarted the h1susetmx, h1susetmy, h1susitmx, h1susitmy models. (7) Noticed some white boxes on Overview Screen, which were site specific to L1, so I fixed these with a $(IFO) substitution and re-committed SUS_CUST_QUAD_OVERVIEW.adl to the svn. Screen-shots below show the updated MEDM screens. This should help inform ECR E1500090 and Integration Issue #1015, in determining which AS WFS channels are most suitable for damping QUAD roll modes at each site.
J. Kissel During the furious model re-arrangement this morning, I took the opportunity to remove IMC X channels from the top level of the MC2 front end model. After the changes to the front end model, this required a recompile, reinstall, restart and restore. The IMC is now locked happily. I want to remove these channels because they are now to be derived in the CAL model, like every other of the IFO's cavity length signals. I'm still in the process of creating a set calibration filter that makes sense for the actuation chain. I would copy and paste what's in the IMC X bank, or what's in the OAF bank, but it turns out the filters that have been in use since Apr 2013 (LHO aLOG 5945) have been wrong. Namely, the M3 stage filter is a copy of the M1 stage filter. The M3 stage filter should (at least) fall off as 1/f^2, which, of course, the M1 stage filter does not. Further, this has been copied, in one way or another to every other place where we've tried to calibrate the IMC length signal since, i.e. the OAF model. So, I'm trying to break the cycle, and stick in sensible filters, and even update them with the what ever current non-sense damping scheme we're using by applying the real damping filters and gains to the SUS model. BUT I'm running onto the classic problem of turning something in matlab into something that foton will chew on. Mostly it's because basic functions like minreal and zpkdata are not cancelling right-half-plane poles in a sane way. I'll bang my head on this more tomorrow. If you want to fight with me, check out /ligo/svncommon/SusSVN/sus/trunk/HSTS/H1/MC2/Common/H1SUSMC2_GlobalTFs_to_Foton_20150224.m Two plots attached: - the bad IMC X / OAF calibration filters. - What the transfer function should look like for MC2 with the current filters and gains. Screenshot shows that the three outputs near the bottom of the main, yellow, MC2 block -- M1, M2, and M3 _LOCK_L_Outs -- are now terminated. I've committed the changes to the userapps repo, jeffrey.kissel@operator1:/opt/rtcds/userapps/release/sus/h1/models$ svn commit -m "Removed IMC X calibration channels from top level. These will now be calcuated in CAL-CS." h1susmc2.mdl Sending h1susmc2.mdl Transmitting file data . Committed revision 9902.
It has been noticed in the past that ISI's sometimes trip when de-isolating. This causes Guardian to pause and wait for the operator to untrip the watchdog, which gets inconvenient when trying to bring multiple chambers to a de-isolated state. The GS-13s seem to be the main culprit, as the BSC St1's don't seem to trip when de-isolating. We may want to try running the GS-13's in low gain, to reduce the odds of tripping on them. I looked at the perfromance of ITMX ISI with the GS-13's in 3 different states, our nominal high gain mode, low gain and "ultra-low" (see Hugh's post 16416 and subsequent comments by JeffK) . It looks like in an isolated state, none of these configurations hits ADC noise below 100hz (see first attached plot, red is high gain, blue low, green ultra-low, ADC noise is brown(?)), although "ultra-low" looks like it might be getting close above 100hz where the ISI is not actively isolating. I also looed at the performance of the ISI in the low and high gain states (second figure, dashed are high gain, solid are ultra-low) and it looks like even in ultra-low gain we still have good performance. I looked at HAM6 (third plot, red is high, blue low, green ultra-low, brown is ADC) as well, not in as much detail, but I suspect we could run in low gain on the HAM's, too.
Summary:
A new ETMX M0 P damping filter was made, together with a new damping gain (first and second attachment), to make the L1 length to L3 PIT coupling at 0.5Hz smaller.
This makes the coupling at 0.5Hz smaller by 8dB or so, but it increases the coupling at 5.5Hz and 3.5Hz, and overall it seems to give us about a factor of 1.5 or so of reduction on average.
This was only done to ETMX, and moving to the new damping of course affects the WFS, so I left it with the old configuration.
Details:
(All of the measurements were made without oplev damping.)
People say that a higher BW WFS is necessary during CARM reduction, and that oplev damping is necessary for ALS. The problem could be a large (and noisy) length drive coupled to the angle.
Instead of optimizing the top mass damping for making test mass as quiet as possible without drive, I tried to minimize the coupling from L1 length drive to L3 PIT. I injected a [0.2Hz, 0.8Hz] uniform noise to L1 Lock L filter so the L1 coil drive does not exceed +-5000 counts or so peak to peak (otherwise the resulting PIT motion becomes too large so the OPLEV starts to miss an entire quadrant).
I started by simply making the gain of the existing filter up or down. This didn't do much, as the steep phase slope of 0.5Hz boost filter makes it such that when you gain something at 0.5Hz you quickly loose at 0.4Hz and 0.55Hz by gain peaking.
After some trial and error I ended up with a broader version of the original filter (second attachment).
The third attachment shows my old/new/old/new test (Old = Red traces, FM2 on, FM3 off, gain = -1. New = green traces, FM2 off, FM3 on, gain=-4).
The top panel shows the transfer function from the L1 length drive to the oplev PIT, and the bottom panel is the oplev PIT spectrum with the length drive into L1. As you see the TF shows the reduction at 0.5Hz but some increase due to gain peaking at higher and lower frequency, and the overall RMS reduction in the TM angle due to the length drive is about a factor of 1.5.
A reduction in the RMS of the test mass angle when we're not driving L1 length is not large, but it is reduced by a small amount (4th attachment, red old, green new).
It would be very difficult to reduce the l2p coupling at 0.55Hz just by the top mass local damping, as there's a sharp dip in the OLTF of the top mass damping that should come from the lower suspension structure (5th attachment, blue=OLTF, red = top OSEM P to oplev P, both measured while driving M0 DAMP_P_EXC). A gain of -3 or so with the new filter to make the damping less aggressive, combined with some other means of decoupling, seems necessary to obtain an advanced awesomeness.
Looks like a little upkeep can go a long way! Still waiting for more time on system to collect high resolution spectra for comparison but the early views look like the ASD of the Pump Pressures are lower by several factors and the remaining coherences still around after the controller changes JeffK engineered have now all gone away.
This was achieved by
1) strapping the common legs of the Pos & Neg power supply to ground. Verbose: The + side of the Neg voltage output is tied to the - side of the Pos voltage output but this was not explicitly strapped to the green ground plug. When I did this before taking the Pump System down, the swing of all pressures dropped immediately by 4x at least--See 1st attachment: 10 minutes of second trends. I have tried this at the corner station as McCarthy suggested long ago but I did not get the same clear affect there--needs more study.
2) Charging up all the Accumulator Bladders--Most of these were extremely low and needed full charging of N2.
See the second attachment, 3 hours of minute trends of the pump station pressures, remote pressures from the BSC10 and their difference and the controller Out signal, for the maintenance period play-by-play. 1) Grounded Power Supply 2) Maintenance work while pump is off: grease motor, check Accumulator pressures, Bleed laminar flow resistors. 3) System back up while retreaving full N2 Bottle 4) Pump back off while Accumulators are charged 5) Pumps back up when complete.
Note: Obvious change in all pressure signals and the reduction of the CONTROL_VOUT which drives the motor from the quieter DIFF_PRESS. I don't know if it is obvious or just wishful thinking but maybe all the signals are quieter with the appropriately changed Accumulators.
Most of the Accumulators were not within the spec of being charged to 60-90% of operating Fluid-Side pressure. One Accumulator on the Pump Station was leaking and held no charge. I was able to tighten the Schraeder Valve and it holds N2 now.
I will post comparative coherence plots in a few more hours.
