Following on from yesterday's attempt.  I made transfer function and spectra measurements with the input modecleaner locked and
unlocked.  In summary, there wasn't too much difference in the noise spectra above 100 kHz with the input modecleaner locked
or unlocked.

    Attached are transfer function measurements when the input mocdecleaner was unlocked and locked respectively (IMCULKD1.tif,
IMCLKD1.tif).  The noise spectra measurements are IMCULN1.tif and IMCLN1.tif respectively.  The mixer calibration was measured
to be 1.93/77.4 V/kHz.

    More to follow ...

Koji, Craig

- We ran the RF9 modulation depth change test Sheila asked us to do.  Koji will post an alog about those results.

- Koji had me lock the OMC with PZT2 at only 2 volts, rather than the usual ~40 V, for +9MHz 9th order HOM considerations.

- We measured the PRCL OLG after full thermalization at high power, and got a 30 Hz UGF.  This seems much lower than what was reported by Jenne a week ago.

- I carried the laptop in with me as I made my way down to the PSL racks.  
This time, I was able to make it without causing a lockloss, but when I got there the SR785 GPIB was gone, and when going to get it from the squeezer bay, while walking between HAM2 and HAM3 I killed the lock.
The area directly in front of the door in the LVEA, between HAM2 and HAM3, is the most sensitive to walking around, according to IMC_F.  Stepping on the lower tier floor directly inside of the LVEA door seemed to produce the largest response in IMC_F, including during my first excursion to the PSL racks.

- We reached nominal low noise again, but lost it within ten minutes, probably due to some CHARD, DHARD pitch ringup.  

- We reach NLN a third time, and this time held it for two hours, during which I was able to get to the PSL racks and measure all CARM spectra.  We eventually lost lock due to a super slowly growing ASC 0.9 Hz oscillation (lost lock about an hour after the ringup started).

- Had to readjust TMSY alignment in pitch significantly to get the guardian to continue locking in FIND_IR.

Figure 1 shows that yesterday the two jitter peaks in DARM were successfully damped and the peak amplitude reduced by a factor of 3 or more.

Recently I had identified the mounts that were responsible for the jitter peaks in DARM that remained after the reduction in table motion from the water cooling system (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=44460), suggesting a broad-band tuned mass damping system. Rick S. pointed out some leftover dampers that had been made for the PMC, based on OMC bench damping investigations (https://dcc.ligo.org/LIGO-T1600494). I recycled some of these parts and mounted them on the U200-A mounts that were creating peaks in DARM.

Figure 2 shows the system, the damper is bolted to the hole on the side of the U200-A that is used for mounting in the other orientation. The manufacturers hole is counterbored so the bolt head does not extend beyond the mount. The extra weight did not deflect the bottom periscope mirror enough to prevent OMC locking.

It took only a couple of minutes to install each, so I think this might be a useful system for all critical mounts in the future. We might make a smaller version for U100-As.

Cheryl also replaced one of the optics because she thought it might be damaged.

Robert, Cheryl

Tonight we were able to shake at EY after finishing EX last night. We found the EY vacuum enclosure to couple much more to DARM than the EX enclosure, and more than during O2. The Figure shows that factors of 10 increase in motion (which happens between 4 and 12 Hz when big trucks pass on 240), produce large scattering features in DARM. More later.

Robert, Kara

I turned off CW injections from the old injection channel H1:CAL-PINJX_CW_EXC at GPS 1231728548 (Jan 17 2019 02:48:50 UTC) and restarted them using the new injection channel H1:CAL-INJ_CW_EXC at GPS 1231728644 (Jan 17 2019 02:50:26 UTC).

I am not currently able to verify using ldvw / nds2 that the injections are appearing, but Keita confirmed via e-mail he can see them being routed to h1calex.

The screenshot shows the injections turning off in CAL-PINJX, but I can't yet show them turning on in CAL-INJ

We are still trying to nail down our non-linear gain measurement. Here's a comparison between two methods:

1) MIT method:

nlg = ((sqrt(pMax/pMin) + 1)/2)^2

2) ANU method:

nlg = pMax/pUnamp

pUnamp is seed transmission measured outside nlg region as I scan the OPO cavity (temperature set to ~51C, green blocked). Which is an unamplified seed power.

I took nlg measurement with data from StripTool and oscilloscope. The result are consistent within 10% between the two methods and two devices. 

Pump input to coupler: 20mW

Pump reflected (OPO off resonance): 2.7 mW

Seed input to coupler: 1 mW

StripTool (IR trans PD): max = 0.0063 mW, min = 0.0019 mW, DN = -6.8e-5 mW

Scope (IR trans PD): max = 142 mV, min = 47.6mV, DN 8mV

Seed trans unamplified: StripTool = 0.003 mW, Scope = 76.4 mV

We lost lock while going through the 9MHz reduction state.  This might be the second time today, although the first time I thought it was due to the extra 6dB of gain in the CARM loop. 

Perhaps part of the problem is that the PD digital gains (which get increased to compensate for the actual modulation depth reduction) are ramped over 30 sec, while the analog EOM driver is just making 3 discrete steps over 6 seconds.  We need to look into this.

However, while looking at the state REDUCE_RF9MHZ_MODULATION_DEPTH, I saw that the ASC 45 MHz photodiodes were getting their digital gains increased by 3dB.  Ooops.  I have fixed this, however we'll need to pay attention to the ASC loops, that in NomLowNoise they'll now have 3dB less gain then they were having.  Should be okay, since all the loops are stable and were set before the modulation depth reduction.

On Monday evening, Koji and I went out and measured the shot and dark noise of the IMC/CARM PDs.  Then, Koji taught me how to think about photodiode noise.  Slides 35 and 36 of G1401145

Model:

where nV is the voltage noise floor, T is the transimpedance, e is the electron charge, R is the responsivity of e λ/(h c) = 0.858 A/W, QE is the quantum efficiency of 0.9, PDC is the DC power from the incident light, and Pdet is the power apparent from the intrinsic detector noise (i.e. the incident power where shot and dark noise are equivalent).

We fit the measured shot and dark noise spectra for IMC REFL I/Q, REFL_A 9I/Q, REFL_B 9I/Q, and at the CARM and IMC servo board OUT2 locations with the nominal gain slider settings (OUT2 connected to IN1, CMB IN1GAIN = 6 dB, IMC Board IN1Gain = 28 dB).  REFL_A spectra are plotted in attachment 3 as an example.
For dark noise I shuttered the PSL.  For shot noise I measured with 1W, 2W, and 4W input requested power, and recorded how much light was reported by the LF channels:

Light Level      REFL_A_LF [mW]     REFL_B_LF [mW]    IMC_REFL_LF [mW]
----------------------------------------------------------------------
Dark                          0                  0                   0
1W PSL                     3.51               3.24                1.87
2W PSL                     7.38               6.80                3.92
4W PSL                    15.08              13.90                8.01

Full Lock                  4.36               4.02                1.15


Then, I found the average noise floors for each spectrum between 3 and 10 kHz, and fit the model to the data.  (attachment two)
I have let the transimpedance number from the fits capture the servo board gain sliders.

Photodiode      Transimpedance [V/A]     Pdet [mW]    IN1 Gain Slider [dB]
-------------------------------------------------------------------------------------
CMBoard                        730.4          2.2                       6
IMCBoard                      2357.7          3.5                      28
IMC_REFL_I                     128.6          1.6
IMC_REFL_Q                     127.5          1.7
REFL_A_9I                      154.9          2.0
REFL_A_9Q                      151.8          2.3
REFL_B_9I                      152.4          2.1
REFL_B_9Q                      152.2          2.1

Whistles have been appearing in the top two rounds of hveto for the last 3 days. To get some better statistics about the troublesome frequencies the IMC VCO is crossing when whistles are produced, I grabbed the times of all the glitches hveto finds in the top 2 rounds. Note, not all the glitches in these rounds will be whistles, but the majority will be. I then looked at the IMC VCO frequency when these glitches were present and plotted a histogram (Figure 1). Whistles appear most often ~78.94MHz. 

I then started to look at other VCOs (regardless of where they are) just to see what their frequencies were when whistles appeared. I grabbed ALS COMM, ALS DIFF, ALS X, ALS Y and SQZ. I then made scatter plots of the different VCO frequencies against the IMC VCO frequency when whistles were present (ALS COMM - Figure 2, ALS DIFF - Figure 3, ALS X - Figure 4, ALS Y - Figure 5, SQZ - Figure 6, apologies for the odd rendering of the y axes). It looks like ALS COMM and ALS DIFF are happily parked away from the troublesome IMC VCO frequencies. The SQZ VCO also wanders away from the IMC VCO frequency range around 79.168MHz. Even though ALS X and ALS Y are many km away from the IMC VCO I still looked at them just in case it was interesting. These VCOs do wander the closest to the IMC VCO. 

Added a few filter modules to the ASC-OMC_A and ASC-OMC_B QPD segments to account for the calibration:

• FM5: "–12dB" is the inverse of the whitening gain
• FM6: "cts2V" calibrates counts into Volts 6.1035E-4
• FM7: "trans" goes from Volts to Amps and is the inverse of the transimpedance gain. The later is 2000Ω including the differential amplifier.
• FM8: "A2W" goes from Amps to Watts.
  • For an InGaAs diode at 1064nm this is 1/0.8 = 1.25.
  • For a silicon diode at 532nm this is 1/0.3 =3.333.
  • For a silicon diode at 1064nm this is 1/0.16 = 6.25.
• FM9: "milli" goes to mW (multiplies by 1000)

TITLE: 01/16 Day Shift: 16:00-00:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Commissioning
INCOMING OPERATOR: None
SHIFT SUMMARY:  new Guardian hardware installed the morning, useism increasing
LOG:

• 16-01-19 16:11 Vanessa to EX to clean
• 16-01-19 16:14 Karen to EX to drop stuff off then wants to go to the LVEA
• 16-01-19 16:39 Peter King to optics lab
• 16-01-19 16:41 Karen leaving EY
• 16-01-19 16:52 Vanessa done at EX heading to MX
• 16-01-19 17:00 Marc to LVEA to HAM6
• 16-01-19 17:10 Karen out of the LVEA
• 16-01-19 17:35 Nutsinee in EE Marc out of LVEA Terry here
• 16-01-19 18:26 Peter King done in optics lab
• 16-01-19 18:35 Nutsinee to HAM6
• 16-01-19 18:40 Terry to HAM6
• 16-01-19 22:00 Kyle back from MX
• 16-01-19 22:00 students to EY to look at a cable in the VEA
• 16-01-19 22:47 Nutsinee to ISCT6

Yasmeen, Ana, Stefan

Went into MSR at 1:30pm local to visually check which 1PPS cables feed into which Timing Comparator 1PPS input Port. No changes were made. Photos taken and attached.

Sheila, TVo, Georgia, Craig, Danny

• Last night as well as today we saw that there is a slow (~60 second period) oscillation of PRC1Y (AC coupled QPD to PRM), which seems to be fighting the dither loop.  We tried to turn off the PRC1Y twice, leaving the dither loop on, and both times an instability rang up in CSOFT P at 0.5 Hz.  Tonight we reduced the gain of CSOFT P by 20dB, and the oscillation went away. We were then able to turn off PRC1Y, and the low frequency oscillation is also gone. 
• Georgia added 10dB of gain to the YAW3 dither loop (which is for PRM), which she is putting in the guardian. 
• TVo and I saw that the X QPD combination used for the CSOFT P loop wasn't orthogonal to the CHARD combo, which we think it should be.  We put the matrix that is orthogonal to CHARD in CSOFT P A, and switched to that (0.111 for XA, 0.159 for XB, and 0.039 for Y B).  We increased the CSOFT P gain, and saw something ring up at 0.5 Hz again. 
• We sat with the CSOFT P gain low for about a half hour again, before something rang up at 0.5 Hz again, most visible in CHARD Y.  We increased the gain of CHARD Y by a factor of 2, and this made the oscillation go away.  With the CHARD Y gain increased, we could again increase the CSOFT P gain back to normal without anything ringing up. 

This current ASC situation seems better than what we've had the last several days, so we probably want to make some of these changes permanent, but we don't want to make any changes in the guardian since ASC engagement has been tricky and we may not have a chance to test things. I think that turning off PRC1Y, and increasing the CHARDY gain from 3 to 6 are definitely improvements.  The change to the CSOFT input matrix I'm not so sure about. 

Summary: BRUCO gave us a few interesting clues about our current noise, including a broadband contribution to the DARM noise that seems to couple through the 9 MHz sideband. 

We had 20 minutes of glitch free data from 1231532246 to 1231533411.  This was before Georgia and Danny fixed the ISS second loop clipping 46412, which improved the noise from 27-15 Hz and around 350-400 Hz. Gabriele and Joe Areda managed to run BRUCO (and update the code to deal with some other updates), a report is here: https://ldas-jobs.ligo.caltech.edu/~gabriele.vajente/bruco_lho_1231532246/

Some interesting features (other than the ISS coherence)

• The peak at 90 Hz which has appeared in our spectrum since December has coherence with accelerometers and ISI sensors around HAM5. This is where PEM injections on the HAM5/6 chamber walls showed some noise that was well below DARM in the past, so we can try exciting those ISIs, or the OFI to see if that has gotten worse. 
  • Some templates used in the past for shaking HAM5 are in /ligo/home/sheila.dwyer/noise_budget/simple-noise-budget/restructured_nb/input/ISI
• The broad coherence of the OMC QPD sums with DARM (OMC QPD A sum).  Some of this is just the intensity noise that is also witnessed by the ISS second loop out of loop sensors (PD7)  The attached screenshot shows a comparison of this coherence before and after the ISS fix, as you would expect the ISS fix reduces the coherence at the frequencies where the ISS noise was reduced, but the broad coherence is unchanged.  The legend in this screenshot is wrong, the gold DARM spectrum is not a 92 Mpc reference.  We aren't sure of our calibration at the moment. 
  • We increased the modulation depth for 45 MHz by 3dB from 5:29 UTC to 6:28 UTC on Jan 16th.  There were many loud glitches in this period, but there is a quiet time that starts at 5:47, the coherence of the OMC QPD with DARM was unchanged at this time.  So this noise isn't coupling through the 45 MHz sidebands.
  • We increased the 9MHz modulation depth by 3 dB from 6:40 UTC to 7 UTC, and the coherence increased, roughly scaling with the sideband power.  (Our range also dropped by about 10 Mpc) Koji reminds us that we've seen a 9 MHz higher order mode in the OMC camera before: 29395  We can't see much in the camera tonight, if we increase the exposure we saturate the camera to the edge of the screen, so if we want to recreate Jenne's images we probably need to go to the new camera housing and move the camera or add a lens.  
  • We could add more whitening to these QPDs, and will plan to do that next time we are unlocked. 
• We hadn't redone the SRCL FF since the increase in power and changes to TCS, so this showed up in BRUCO.  Georgia, Craig and Danny have improved the feedforward above 60 Hz (separate alog).  

A new h1calinj model has been installed on the h1oaf1 machine.  This model will hold all front-end injection handling logic, including the injection EXC test points, under the channel prefix "H1:CAL-INJ_".  The existing injection handling code in the h1calex model ("H1:CAL-PINJX_") has been left as is for now to facilitate transitioning and testing.

The first two attached images are of the top-level contents of the new h1calinj model, and of the contents of the "INJ" CAL_INJ_MASTER2 library part that contains all the core logic.  The top of the latter shows the main injection signal flow.  The two "CW" and "TRANSIENT" filter banks at the upper left hold the EXC inputs and calibration filters for the continuous-wave (CW) and transient (TRANSIENT) injection inputs respectively.  The outputs of the two modules are summed and the overall "MASTER" output goes through an output switch ("MASTER_SW"), and then finally a switch ("END_SW") that determines which of calex or caley receives the injection signal ("H1:CAL-INJ_{X,Y}" via cdsIPCxRFM).  Below the main signal path is logic to determine the presence of signals at various points in the injection path, and bundle that info into a single status word ("STATUS_OUT").

IPC receivers for the INJ_MASTER output sent from h1calinj (H1:CAL-INJ_{X,Y}) were added to the h1calex and h1caley models.  EPICS and acquired test point monitors of the received signals were also added.

NOTE: a single 2**14 Hz, 61 us cycle delay will be added to the injection path because of the IPC needed to carry the signal from the vertex to the ends.

The new acquired fast channels are:

• H1:CAL-INJ_CW_OUT_DQ
• H1:CAL-INJ_TRANSIENT_OUT_DQ
• H1:CAL-INJ_MASTER_OUT_DQ
• H1:CAL-INJ_STATUS_OUT_DQ
• H1:CAL-INJ_X_OUT_DQ
• H1:CAL-INJ_Y_OUT_DQ

The STATUS channels (the H1:CAL-INJ_STATUS_OUT_DQ uint32 fast channel and the H1:CAL_INJ_STATUS slow channel) have the following bits:

0. MASTER
1. CW
2. TRANSIENT
3. CBC
4. BURST
5. DETCHAR
6. STOCH

A value of zero (0) indicates no signal of the specified type is present, and a value of one (1) indicates the presence of the specified signal.

The h1calinj model also holds the "TINJ" EPICS status bits, nominally set by the INJ_TRANS guardian node.  NOTE: the "H1:CAL-INJ_TINJ_" EPICS records were previously hosted by the ext_alert_ioc.py soft IOC running on the h1fescript0 machine.  The records were removed from the soft IOC and the process was restarted.

The final image attached is a new MEDM screen CAL_INJ_CONTROL2.   All functionality and status bits in the h1calinj model, and the monitors in h1calex and h1caley, are exposed.

As mentioned above, all the existing INJ infrastructure remains in place.  We leave it up to the INJ group to update the INJ_TRANS guardian, the psinject process, downstream monitors, etc.  We would like to schedule the removal of the old INJ infrastructure as soon as possible,

Previously in alog46336 I felt like I was bottomed out on some of the measurements phase delay wise. To get more phase delay so we can be more confident in our measurement I gave 3MHz phase delay box to CLF (I have a feeling that phase delay box hooked up to LO/OMC gives me a bit less phase, not sure if that make sense, I will post more detail alog on the ellipse rotation later). Add CLF sign flip on top of that (which gives us extra 90 degree as Daniel suggested) I was able to go from OK to GOOD then back to OKAY measurement on both squeezing and anti squeezing. Flipping LO sign didn't help (180 deg). And using Q error signal to lock the CLF seems to have given me some extra phase adjustment (not sure why, CLF common mode board input 2 now has Q error signal goes into it). This way I know that I've seen the best sqz/asqz (by monitoring LO Q error signal go up and down while adjusting the phase, we locked with I).

After correcting for some of the minor loss typos and added data taken yesterday after I've acquired more phase delay, here I attached another loss estimate plot. I also give phase noise of 10 mrad this time since assuming that what we measured out of the LO IMON isn't all the phase noise there is. The result hasn't changed. We have more loss compared to when Haocun took a measurement here. We are still in the process of checking red transmission then and now.

Fringe visibility during Jan14 measurement was 99%. In the model I use 97%.

I have been wondering about the true input power, i.e. the input power incident on PRM, ever since the SRCL dither arm power measurement turned out so low.  

Conclusion
With an input power of 1.87 watts according to the PSL photodiode, we probably have around 1.55 watts incident on PRM, assuming an addition 21% unaccounted losses in the REFL A/B LF PD path.

Power Measurement Methods
I aligned PRM and the input optics, and misaligned the rest of the corner with 1.87 W reported from the PSL input power PD going into the IMC.

IM4 Trans
The easiest way would be to get Pinput would be via IM4 TRANS.  IM4 TRANS is calibrated into microwatts, IM4 (aka SM2) transmission is reported to be 2400 +- 200 ppm, and there's what I think is a 90:10 BS (ROM LH1) allowing 10% of the light going to IM4.  However, if we turn the crank here we end up with only 0.2 W, which is not possible.  This means that IM4 TRANS is not calibrated correctly, or the IM4 transmission path is actually around 8 times less transitive that I though.

REFL A/B LF
We believe that the REFL A/B LF PDs are correct based on the calibrated CARM shot noise.  We can use them to calibrate the correct input power.  The main issue here is if we use the REFL optical path (4 50:50 beamsplitters, 1 90:10 beamsplitter, one input faraday isolator, and one PRM) and assume no losses, we only get 1.19 W of true input power on the PRM.  This could be the case, but it's more likely that there are some unaccounted for losses on the REFL path.

PSL and IMC Transmission
If we trust the old numbers from P1500076 on the PSL and IMC transmission, we can make an estimate of the true input power.  We find that by varying the PSL and IMC transmission, we cannot reasonably recover the REFL A/B power levels we see.


PD Powers and Optic Transmissions Assumed
# Most taken from P1500076

# PD calibrated powers
PSL_INPUT = 1.87 # W
REFL_A = 7.3e-3 # W
REFL_B = 6.72e-3 # W
IM4_TRANS = 50.24e-6 # W

# PD uncertainties
dREFL_A = 0.04e-3
dREFL_B = 0.03e-3
dREFL_SUM = np.sqrt(dREFL_A**2 + dREFL_B**2)

# Transmissions
T50 = 0.5
T10 = 0.1
T_PSL = 0.953 # +- 0.013 PSL to MC1 transmission
T_IFI = 0.977 # +- 0.004 https://dcc.ligo.org/DocDB/0119/P1500076/003/Mueller16rsi-AdvIO3.pdf
T_IMC = 0.92  # +- 0.031
IMC_ModeMatching = 0.984 # 0.001
T_PSL_to_IFI = T_PSL * T_IMC * IMC_ModeMatching
T_IM4 = 2400e-6 # +- 200e-6 E070092
T_PRM = 0.031 #

# Trans Uncertainties
dT50 = 0.01
dT10 = 0.01
dT_PSL = 0.013
dT_IFI = 0.004
dT_IMC = 0.031
dIMC_ModeMatching = 0.001
dT_PSL_to_IFI = np.sqrt( T_PSL_to_IFI**2 * ((dT_PSL/T_PSL)**2 + (dT_IMC/T_IMC)**2 + (dIMC_ModeMatching/IMC_ModeMatching)**2) )
dT_IM4 = 200e-6