TiVo, Danni, Dave:
Two changes to h1tcscs: added RH_INVERSE filters for etmx and etmy, removed surplus 'ezca' parts for LSRPWR_MTR and replaced them with a connection between the ITM[X,Y] blocks and the SIM block.
DAQ was restarted.
Figures attached. Terminal/xterm windows did not list any CPS sensors which were over threshold. For BSC BS & ITMy had individual sensors which were a slightly higher, but Jim wasn't concerned when I showed him.
TITLE: 12/19 Day Shift: 16:00-00:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Commissioning
OUTGOING OPERATOR: None
CURRENT ENVIRONMENT:
Wind: 5mph Gusts, 3mph 5min avg
Primary useism: 0.08 μm/s
Secondary useism: 0.68 μm/s
QUICK SUMMARY:
Jenne is working on H1 (currently holding on CHARDBLEND). Useism is still fairly high, SEI_CONFIG is in WINDY state.
Craig, Dan
Tonight we have been tuning some TCS settings to try and improve the RF18. When we go from 2W -> 20W RF18 drops from ~80 to ~60 on a thermal time scale and we haven't had much luck on improving it.
Given the high absorption on ITMY the substrate lens increases significantly more than ITMX's during power up. To combat this the prevailing thought was to either turn down CO2Y or increase the ITMY ring heater to induce a negative substrate lens. However, this didn't work and only made the POP18 and REFL_RF9_Q worse it seems: Increasing the substrate lens using CO2X made a significant difference to the quality of the lock (see here around 20000s) and increased RF18 by ~2, increasing too much more didn't help. The idea behind this is that it makes the differential lens smaller as well as increasing the common lensing. What also helped was increasing CO2Y, surprisingly, this helped bring REFL_RF9_Q back towards zero when CO2X was increased too much and also improved RF18 slightly. It also increased the PRG and TRX/TRY so it seemed to be helping the overall mode matching of the IFO. Pushing this too high led to a CHARD instability though and lost lock. We also tried annular CO2Y (see here around 18000s), from the Hartmann output we were inducing a negative lens but this made POP18 significantly worse.
I've set CO2X=0.6W and CO2Y=0.3W when we reach 20W in the Guardian, POP18 should level off ~61 at increase power. However, going up to low noise it dropped further to ~57 and it starts becoming noisier again, which suggests we can get it back up with some alignment tweaks as well as TCS changes.
Lock losses:
Comparing the locks that Dan posted about the CO2 steps can give us an little more understanding about the pico-motoring we did relative to the IFO beam positioning. I attached four wavefront maps which compare the IFO beam to the CO2 beams, all in all, CO2Y looks a lot better than it previously did here.
Regarding the ezca error with bounce/roll mode monitoring checks in the LOWNOISE_ASC state of the ISC_LOCK guardian: After yesterday's move of these monitors from the h1oaf to h1susproc (LHO aLOG 46047) the $(IFO):OAF-BOUNCE_$(OPTIC)_RMS_LOG10_OUTMON channels have changed name to $(IFO):SUS-BOUNCE_$(OPTIC)_RMS_LOG10_OUTMON and likely the above team found that ETMX was the problem because it's the first element in the array in a for loop over checking that these channels values are low before moving on. I've updated the ISC_LOCK guardian code to look at this new channel name and reloaded it.
This is an interesting lockloss, which could use more investigation: 1229244374
This was a lock in which we had much more stable build ups than we have been seeing at 20W, thanks to TCS work today. We had been sitting in the CHARD BLEND state for more than an hour, when the refl centering loops (as well as SRCL, DHARD and to a smaller extent PRC2 Y) have an exponentially growing oscillation at 28.6 Hz. The verbal alarms indicate that an HTTS was saturating, but the lockloss website says that there were no saturations until ETMX saturated at the time of the lockloss.
Looking at the master outs for the RMs, they look like they should not have been saturating (second attachment, this is a 16 bit DAC so it should saturate when the request is +/-32768 counts). It would be good to compare what verbal checks for saturations against the master outs that are used by the lockloss tool to see if we are missing some saturations.
In any case, this lockloss is very similar to the ones we saw a couple of weeks ago trying to power up 45668
Verbal just looks for increases in the model overflows ({IFO}:FEC-{#}_ACCUM_OVERFLOW). On the h1sushtts model, there are only the RM's, I believe, so it wasn't any of the other HTTS's. Just for reference, here is the Verbal log when the saturations started to lock loss:
H_T_T_S (Dec 19 08:45:31 UTC)
H_T_T_S (Dec 19 08:45:33 UTC)
H_T_T_S (Dec 19 08:45:35 UTC)
H_T_T_S (Dec 19 08:45:36 UTC)
H_T_T_S (Dec 19 08:45:39 UTC)
H_T_T_S (Dec 19 08:45:41 UTC)
H_T_T_S (Dec 19 08:45:42 UTC)
H_T_T_S (Dec 19 08:45:45 UTC)
H_T_T_S (Dec 19 08:45:47 UTC)
H_T_T_S (Dec 19 08:45:48 UTC)
H_T_T_S (Dec 19 08:45:50 UTC)
DCPD (Dec 19 08:45:52 UTC)
H_T_T_S (Dec 19 08:45:53 UTC)
DCPD (Dec 19 08:45:55 UTC)
DRMI Unlocked (Dec 19 08:45:57 UTC)
ISC_LOCK state: DOWN (Dec 19 08:45:58 UTC)
Re-looking at Sheila's middle plot, RM2 is saturating, so verbal was telling us the correct thing.
TVo Danny Dan Georgia
The alignment of the CO2 lasers somehow at some point became bad. The image of CO2Y mode on the Hartmann looked like a sausage (probably clipping), and CO2X was clearly misaligned from the ring heater and IFO beams.
We adjusted the picomotors on the CO2X and CO2Y top periscope mirrors. To measure the alignment we turned the CO2 on after each step on the picomotor and looked at the spot position on the HWS. TVo, Danny and Dan took screenshots of the spot position measured on the Hartmann which they will post later.
Initial position (x, y) Position now (x,y)
CO2X -14935, 12200 -14935, 3200
CO2Y -48170, -27274 -50179, -28274
Since pico'ing we've acquired DRMI very quickly (once), and powered up to 20W without a problem.
I also tuned the calibration on the rotation stages for both ITMX and ITMY because Corey and TJ noticed that it wasn't close enough for the TCS power guardian to be happy and the rotation stage was jittering around a lot. This also gets us closer to the requested values for CO2 Power.
Nutsinee, Sheila
Summary: We went back through the data from last night's squeezing injection. We see that the noise added to DARM doesn't depend directly on the CLF power, but it is suppressed by the LO locking loop (which is feeding back to the OPO length). This suggests that OPO length noise is coupling to DARM but not through the CLF or seeding; back scatter is one possibility.
Details:
Our goal last night was to change the CLF power to see how it impacts the noise added to DARM by the squeezer. We did not change the demod phase at all to try to adjust the squeezing angle. Since the locks were short last night, we were only able to get one or two injections in each lock, and our alignment was changing from lock to lock. We looked back at the data to find the CLF fiber launch power, which we use to estimate how much CLF was injected into the interferometer, the RF3 demod power, and a few other things. Since squeezing injections were done over a couple of locks, and the squeezer alignment was adjusted after the first lock, the alignment of the squeezer to the OMC was different for different times. The number in parenthesis in the legends of the attached plots are an estimate how well aligned the squeezer was. We took the ratio ((10^(RF3/10)/sqrt(CLF launch power)))^2, and normalized by the best alignment from last night to make this estimate.
The first observation (second attachment) is that when the squeezer is badly misaligned, we add less noise to DARM (which is what you would expect). The second attachment shows that we have more noise introduced into DARM with better alignment, the same injected CLF power, and the same analog gain in the LO loop. This means that we had lower optical gain in the LO loop, but also that we have less light from the squeezer reaching the DCPDs.
The next more interesting observation from is that increasing the analog gain for the LO locking loop reduced the noise added to DARM for a similar CLF power and alignment efficiency (first attachment).
One explanation could be that increasing the LO loop gain (which is currently feeding back to the OPO PZT) decreased the OPO length noise. Nutsinee and Daniel today found that we were causing a lot of OPO length noise (46058), which is not suppressed by our OPO locking loop since it feeds back to the laser. Since the OPO length noise was the dominant noise in the LO error signal at the time of these injections, the OPO length noise could be decreased by increasing the LO gain.
The OPO length noise could directly couple to DARM through intensity noise on the CLF. Since the CLF is off resonance in the OPO, the OPO length fluctuations cause intensity noise on the CLF, which is attenuated by the OMC but could still add noise if large enough. We also have the comparison in the third attachment (below), where the analog gain stayed the same, the CLF was increased and the alignment was improved, but the noise in DARM improved. The DCPD 3MHz demod shows that there is 3dB more 3MHz signal at the time of the yellow trace (3:46:59 Dec 18th), which means the LO loop gain was increased by sqrt(2), so the OPO length noise and the CLF RIN could have been decreased by a factor of sqrt(2). The CLF power reaching the DCPDs was a factor of 4 higher, so if the CLF RIN was the coupling mechanism we would expect the noise to be larger by 4/sqrt(2) ~3 in the yellow trace, but the noise is actually lower, suggesting that the coupling isn't through the CLF RIN. We are planning to revisit 45858, since that projection of CLF RIN to DCPDs was based on a RIN measurement made on ISCT6, before the OPO length noise would have been imposed on the CLF intensity. '
This comparison also suggests that the noise isn't due to seeding that comes through the AOMs in the CLF path, since there would be more seeding when we inject more CLF into the OPO, and when it is better aligned.
The fourth attachment is perhaps a more convincing comparison to see that CLF intensity noise and seeding aren't the cause of noise in DARM. At the time of the mint colored reference (4:33:40 UTC) there was 6dBm more 3MHz signal than at the time of the black reference (4:30:26 UTC), but the LO loop gain would have been kept the same since Nutsinne increased the analog gain by 6dB for the lower CLF power. (The shelf at around 100 Hz is a transient, we were having these kinds of transients which could be fringe wrapping during all of our locks last night).
Here are the times that we used. Some of these were very short measurements.
| Dec 18 2018 UTC | CLF fiber launch power (mW) | DCPD 3MHz power dBm |
| 2:19:13 | reference no squeezing | |
| 2:35:21 | 0.7 mW | -24 |
| 3:46:59 | 0.7 | -16 |
| 3:50:35 | 0.38 | -19 |
| 4:33:40 | 1.018 | -14 |
| 4:30:26 | 0.295 | -20 |
Increasing the CLF power will increase the optical gain in the LO, which will decrease the noise as seen in the first plot. Potentially, this could mask an increase in the noise due to the CLF power.
Daniel agreed with the conclusion of this alog (that the noise is independent of CLF power) after looking again at the 3rd and 4th attachments.
Here is some information that we can use to make some estimates about noise added due to backscatter:
We had 20mA on the DCPD sum at the time of this measurement, the DCPD responsivity is 0.858A/W (45736), so there was about 23.3 mW of carrier heading toward the AS port (I'm not taking into account OMC losses and mode matching)
While in chamber, we measured 620 ppm of the light heading towards the OMC is sent towards the squeezer by the OFI (41229). (about 14.4 uW arriving at the beam diverter or the VIP)
In the optics lab we measured the isolation of the squeezer faraday to be -27.8dB (39897), so we expect about 23 nW of carrier from the interferometer to reach the OPO.
We are planning to work on getting a calibration of the LO error signal into meters of OPO length, so that we can estimate the amount of backscatter from the OPO we would need to explain the noise in DARM. If we get another chance to inject squeezing into the interferometer we plan to try a fringe wrapping measurement and to check how the noise measured on the DCPDs depends on the local oscillator power (by changing the DARM offset).
Daniel, Nutsinee
Attached an LO IMON (in-loop) before and after we removed the feed forward that we plugged back in the night we injected squeezing the first time (Dec 10). Since our configuration makes the laser follows the OPO, any left over noise that LO have to deal with would be the OPO length noise and turned out the feed forward set up (going from ADC through a pomona box to in-vac PZT driver) introduced a lot of OPO length noise. The rms at low frequency came down from 100mrad to 10 mrad after the cable was unplugged. Daniel reconfigure the loop shape that puts us below 10 mrad (3rd attachment, 30kHz UGF). The 1.4kHz peak now dropped by a factor of 6. The 1.4kHz wasn't improved after we unplugged the feed forward cable. It disappeared from TTFSS PZT mon when LO loop was engaged and came back when LO loop was disengaged (our LO loop feeds back to the OPO PZT). This suggests that 1.4kHz may have come from the OPO itself.
Part of the noise >25kHz can be made better or worse by adjusting sqz angle phase. So the gain has to be reconfigured for the squeeze angle phase when we know where to leave them.
Josh, Jeff B., Guillermo
Scattering arches in DARM between 40 Hz – 100 Hz are coincident with scattering arches in the ASC TR A and B photodiodes at frequencies between 3 Hz – 10 Hz.
Summary:
Results:
Note: scattering predictor in Hz is defined as F_fringe = 2*abs(v_surface)/lambda and the scattering harmonics are the fringes or arches generated after the scattered light hit more than once the perturbed surfaced; see Noise from scattered light in Virgo's second science run data.
there's no laser beam up near M0 or R0, so that's not likely. But perhaps the tool should be reconfigured to consider the difference between the motion of different surfaces. That may point more directly to the scattering path (e.g. between the mirror and the vacuum viewport, or between the mirror and the primary mirror of the transmon, or even a three way interaction)
I traced the source of the small peaks in DARM around 280 Hz that were coherent with the EX ESD power supply voltage monitor channels (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=45241 ) to the fans in the IO chassis at EX. This coupling is present even after the switch to the new dedicated 18 and 48 V power supplies for the ESDs.
The DARM coupling is consistent with fan contamination of test mass actuation: Sheila and I found that the fan peaks were in DARM when the control was at ITMX, but only if the ETMX bias was on. The coherence went away when the ETMX bias was turned off (I should note that this test was before the ESD power supply switch).
Figure 1 shows the fans in DARM, after the power supply switch, during E13. The peaks were identified by individually manipulating the fan frequency by partially blocking the fan ports. All 8 fans from the 4 I/O chassis (h1iscex, h1susex, h1susauxex, h1seiex) appear in DARM, the ESD power supplies, and a monitor of the voltage difference between the field rack and the vacuum enclosure. The presence of, for example, the SEI chassis fans in DARM, and in the voltage monitor channels that pass through the ISC IO chassis, suggests that the fan coupling mechanism is not the previously identified mechanism where the fans for a particular IO chassis contaminate the channels in that chassis by producing ripple on the chassis power supply ( this mechanism has virtually been eliminated in the new IO chassis that are starting to be installed (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=43425 ). Instead, the fans are contaminating external power, even the ESD supplies; I think that the most likely mechanism is that the fans are contaminating grounds.
There appears to be significant a significant current and voltage difference between the grounding bar for the ESD HV power supply and the bar for the 48 and 18V ESD supplies. It may be that moving ESD power supplies (and ring heater supplies also) to the same bar would reduce the coupling. But Im not yet convinced that this is the main problem and Id like to do some more grounding investigations when I return.
Robert, Sheila
Removed and replaced the annulus ion pump for GV15. No issues were encountered during the replacement, after new pump was installed the annulus system was pumped for about 3 hours with an aux pump cart. After 3 hours the ion pump was able to maintain vacuum, thus the aux cart was decoupled and turned off. GV15 Annulus ion system back to nominal.
Done per WP#8009.
FRS ticket 12030 closed.
The IMC REFL beam on the bottom periscope mirror was found to be low and off in yaw, which meant there was clipping and scatter on the high power beam dump path, and required an additional beam dump for a set of second reflection beams, and an iris in the path to the REFL camera to eliminate extra beams.
The REFL beam path has been aligned at the periscope, and restored through the rest of the path (with some help from Jenne, alog 46042, who centered the IMC WFS in pitch and yaw with the picos, after Jason and I centered for total power).
To get the best elimination of extra beams, the pick off mirror, IO_MCR_M14, was moved further upstream from IO_MCR_BS1, that sends a beam to the REFL DC diode, and a razor blade beam dump was installed behind the optic.
The beam dump, IO_MCR_BD5, was adjusted to block as much of the second reflection from IO_MCR_AR1 as possible without clipping the main beam, but the main beam and the AR1 second reflection separation is minimal, and smaller than my experience of the past. The behavior of the junk light present at BS1 was unexpected, so it's not clear if the two expected beams, or the two plus an additional beam, are present, but regardless, it was not possible to fully dump the junk light without clipping the main beam.
Snapshots of IMC REFL and IMC Trans attached: REFL unlocked, REFL locked, TRANS locked
- Jason, Cheryl
Took pictures of face of ITM-X and ITM-y under green light as the IFO unlocked while setting up. Will try to get a set of shots under IR at at least 20W before the holiday break.
There are two shots of each optic at different exposures and focus points.
| Optic | Frame # | Shutter | ISO | Focus |
|---|---|---|---|---|
| ITM-Y | 111 | 1.0 sec | 6400 | 1725 |
| 112 | 1.6 sec | 6400 | 1725 | |
| IYM-X | 110 | 3 sec | 800 | 1745 |
| 117 | 3 sec | 800 | 1625 |
to do precise image subtraction in the future (to detect small movement in the point scatterers) we would like the camera position to remain the same. It would be nice, therefore, to design a camera holding jig that would get us back to the same place w.r.t. the viewport. Would be easier to make that analysis quantitiative this way rather than do a lot of image scaling/aligning with camera at different positions.
Rana, Very good points. The camera/lens assemblies are bolted to the Viewport Camera Housing (D1500073), which allows for positioning the cameras. This assembly is attached to the viewport on the Spool Flange via the Viewport Guard Assembly (D080367). This provides a rigid and locked positioning of the cameras. The whole mounting assembly is then covered by a light and dust proof housing. The cameras are controlled, focused, and fired remotely, so there is no touching of the actual camera assembly. Unless the system is disturbed is some way the image should not shift in the frame. Also the edges of the baffles and the EQ stop mountings are present in the image, which should provide a fixed point of reference. These are just being deployed and any input to make the system more useful would be greatly appreciated.
T.J., Dave:
The CDS_CA_COPY guardian node did not fail in the way we hoped when the LSRPWR_MTR channels were removed from h1tcscs, resulting the camera-copy part no longer running. We may need to split the node out into two.