1635 -1700 hrs. local -> To and from Y-mid Behaviour a little different today. ~5 minutes after opening the LLCV bypass valve 1/2 turn open there was no indication of vapor at the exhaust. It took an additional 5 minutes after opening a full turn before complete. As found, the exhaust pipe was noticeably absent most of the semi-permanent ice build up - warmer ambient temps? Next over-fill to be Tuesday, Feb. 16th before 4:00 pm
In the first attachment you can see that something happened to the TMS osems on Jan 21st that makes the peak to peak noise increse by more than a factor of 10. This is still the case.
I then took a look at the osem spectra, (the fast channels which are not DQ) and saw a 16 Hz comb in RT and SD (second screenshot). Without saving I re ran the measurement including TMSY channels for comparison, and then saw that the comb appeared to have gone away. Spectra of the 256 Hz DQ channels from the time when I originally saw the 16Hz comb are in the third attachment, there is no comb here.
The last attached screenshot shows the osem spectra for both TMSs that I am measureing now, there is some oscillation around 1800 Hz. This is very similar to the behavoir reported in 25062 and 20078
Seems like everything in ETMX chamber is polluted by TMSX OSEM oscillation.
This is an annoyance, some time this week Richard will do the same modification as SR3 mentioned in alog 25516.
{Rana, Evan}
This evening we looked at the coupling of DCPD bias voltage into DARM.
Each DCPD is biased with +12 V from a linear regulator with a 1 Ω output resistor. We wanted to inject extra bias noise, so we exposed the bias lines with a breakout board at the DCPD chassy in the HAM6 rack (see D1300502). Then we summed in the output of an SR785 across an impedance of 10 kΩ in series with 20 µF (see diagram; green shows the normal DCPD electronics and red shows the addition). The 10 kΩ gives a 1:104 ratio of bias fluctuation to drive voltage, and the capacitor ac-couples the drive so that it does not pull on the bias at dc.
With the SR875 we drove a 1 Vpk line first at 187.3 Hz and then at 62.4 Hz (i.e., the bias fluctuation was 70 µV rms). We looked at the response in DARM (at 2 W dc readout, with DARM still controlled by EX). We had 10 mA dc on each PD, in the 400 Ω transimpedance configuration.
For 187.3 Hz (from 02:03:00 to 05:22:00 Z), the magnitude in DARM was 1.13×10−18 m rms, which implies a coupling of 1.6×10−14 m/V. The noise of the regulator at this frequency was 0.9 µV/Hz1/2, which implies a DARM noise of 1.4×10−20 m/Hz1/2.
For 62.4 Hz (from 05:33:00 to 05:56:30 Z), the magnitude in DARM was 1.4×10−18 m rms, which implies a coupling of 2.0×10−14 m/V. The noise of the regulator at this frequency was 1.9 µV/Hz1/2, which implies a DARM noise of 3.7×10−20 m/Hz1/2. Note that there is some room for error here because the DARM control configuration here is not exactly the same as the low-noise configuration. However, the ugf and phase margin should not be too different in the two configurations.
This is very close to the current low-noise DARM noise floor. However, if DARM is limited by voltage noise of the regulators, we should either expect that the DCPD null stream is equal in magnitude to the DCPD sum (it isn't), or the regulator noises are coherent (they aren't).
I am not sure if there are any contradictions in these numbers. DARM calibration at DC is C [W / m] = 4 * pi * G_arm / lambda * sqrt(G_prc * P_in * P_as / G_src) = 1.5e9 and 5e9 for input power 2W and 22W and 26mW at the AS port. This means that the response you measured is ~2e-5 A/V. If the input power is 22W, 1uV/sqHz projects to 4e-21 m/sqHz around 100Hz. This number is what you have in your noise budget plots.
The freerunning DARM channel may not be properly calibrated in the low-power state with EX control, so the numbers I gave originally could be too high.
We directly measured 1.2×10−6 mA rms in DCPD B at 187.3 Hz, which implies a coupling of 0.017 mA/V, which implies the bias noise shows up in DCPD B at 1.5×10−8 mA/Hz1/2. The shot noise with 10 mA on the PD is 5.7×10−8 mA/Hz1/2. So it is a factor of 4 or so below the DARM shot noise, assuming DCPD A is similar. That's close to the dark noise, but the dark noise is flat down to a few tens of hertz while the regulator noise has some negative slope.
Voltage noise of bias (pin #8 of the d-sub), measured with BNC clip doodle and SR785 (with AC coupling).
Sheila, Rana
We continued the investigation. Today we hooked up a DAC channel to the bias to make sweeps and noise injection. No surprises relative to yesterday. Still seems so close to DARM as to be unbelievable. Maybe the 1 Ohm resistor in the bias circuit is not stuffed?
We also injected random noise into PZT2 to look for upconversion there. We saw the usual kinds of quadratic coupling; noise at a level ~500x above the quiesscent HVmon noise level was able to be just visible in DARM in the baseband as well as by injecting noise close to the 4100 Hz dither frequency.
0554 UTC, leaving it in 'low noise mode'. The BS coil driver is in some mixed low noise state giving us increased DAC noise below 50 Hz, but the noise above there is as good as ever.
Can't get the photodiode D's off the DCC at the moment (outage???), but surely there must be a biga$$ cap (ideally several in ||) from the pd cathode to ground??
Yes, in the in-vacuum preamp, there is a 1 uF capacitor from the Bias to ground. At the LM317T (voltage regulator that sets the bias), there is a 10 uF cap to ground, on the regulator side of the 1 ohm resistor (it's not shown in Rana's sketch above). In any case, the projection given above relies on the 1e4:1 scaling of the injected signal - can't you just measure directly the drive on the bias line, rather than assuming the 1e4:1 ratio?
Bigger caps are unlikely to help much. The noise is due to the internal burried zener diode which serves as the voltage reference. The BW of the regulator is about 10kHz here. A bigger C may reduce this, but may also increase gain peaking. Just use a low noise supply such as the one here. Noise performance: T1000025.
Den, Rana
We directly measured the voltage noise on pin8 and confirmed that our injection through the 9.09k resistor (not 10k as indicated above) scales as expected. It seems that the 1 Ohm resistor is really 1 Ohm (+/- 20%).
From MZ, we have http://www.edn.com/electrical-engineer-community/industry-blog/4422750/2/Simple-circuits-reduce-regulator-noise-floor.
This indicates that some larger caps may reduce the regulator output noise at 100 Hz by a factor of 5 or so. Could be easy to try if we have a spare whitening chassis sitting around.
Rana, Sheila, Den We have measured the bias noise with a better resolution and found that this noise is almost flat below 60Hz. Then made a projection of this noise to DARM when the input power was 22W and power on each OMC PD was 10mAmps. For PD A the coupling is 1.5e-15 m/V while for PD B is 3e-15 m/V. Attached plot shows the projection of bias noise to DARM. It is factor of ~4 below the current sensitivity.
According to the noise budget and measurement of the dark noise, this noise is above the dark noise level. This means that the noise is present only when the DC photocurrent is present.
Suggests QE compression due to bias reduction. For higher P it may be wise to not only improve bias voltage noise, but to fix anode-cathode potential. Either a true transimpedance preamp, or a tracking supply at the rack.
We drove PD_B bias and measured the coupling of the bias noise to OMC NULL channel for different current levels. Plot is attached. The coupling is close to zero when there is no power. The coupling scales linear for small currents and stays almost constant for large currents (>10mA on each diode). We have also double checked that the optical signal scales linear with the current by driving an intensity line.
Kiwamu, Den We have digitized bias channels (LSC:EXTRA_AI_2 channel) and double checked coupling to DARM. Attached plots show coupling of PD_B bias to DARM. There is a calibration line at 62.4Hz of this coupling.
[Evan; Rana; Kiwamu];
We restored the 20 dB gains (that have been commented out for long time) in all the soft loos. We still have some suspicision that the ISS may be doing something bad causing locklosses. In order to separate the issues, we decided to gain them up. If one wants to get rid of them, just comment out lines 1990-1993 in the ISC_LOCK guardian.
The ASC loops had a slow oscillation with these 20 dB gains today, so we commented them out again.
{Kiwamu, Evan}
We had another instance of the interferometer coming into resonance with low recycling gain, even though this was the first lock after an initial alignment.
In this case, the issue was solved by redoing the initial alignment, and letting the green ALS alignment loops integrate about 20 minutes. In this case that might have been overkill; from the attached plot of the ITMX control signals (which contain several instances of offloading to the sliders: 20:54:00, 20:59:00, and 21:08:00), one can see that the control signal took about 10 minutes to reach 90% of its asymptotic value.
We already knew that some of these loops were quite slow, but 10 minutes borders on excruciating. We should try to make these faster.
The CDS web service (cdswiki) has been down since at least 11:50 PST, Evan H has kindly agreed to restart it to see if we can clear the error.
Machine came back after hard RESET using front panel button. Log files show no problems at the time of the freeze up, last log was 03:52 PST which could be close to the time the computer locked up due to the verbosity of the logging.
Thanks again to Evan for the reset.
The ITMX ISI tripped after the most recent earthquake, but the corresponding box on the ops overview screen remained green.
The trip did show up as a diag main guardian notification, however.
Edit: nevermind, it seems that the ops overview screen on video2 is frozen.
Last night, we seemed to have less number of the ETMY DAC saturation events presumably due to a combination of the new modification on the ESD driver electronics (25468, 25485) and moderate seismic and wind conditions. The DACs for driving the ESD now have more range above 152 Hz and therefore can tolerate some of the fast transients without hitting the DAC output range. As shown above, the BNS range plot from last night is cleaner than the typical O1 range plots. Looking at the undisturbed data from yesterday, we had only two saturation events (both of which occurred at the ESD stage of ETMY) out of the 9 hours lock stretch; one at the beginning and the other at roughly 10 minutes before the lockloss. See the attached trend.
A follow up:
I checked the number of saturations from a 14 hrs long stretch from this Wednesday's night as well. I found only two saturation events in the ETMY suspension (both of which happened in the ESD stage. One of them occurred at the ESD and L2 stages). See the attached trend.
The BNS range was already posted by Corey (25502) which showed two deep dips at t = 7 and 14 hours. These two correspond to the saturations at ETMY. I did not study the other small dips.
The new ISS digital AC coupling caused the diffracted power to oscillate by ~1% with a period much less than 1Hz. I tried turning the gain down on the servo, but we lost lock (we were in NomLowNoise).
Later, I locked the interferometer to nominal low noise in order to check the behavior of the ISS. The 2nd loop was able to automatically engage the servo. I did not see a fault behavior with the ISS at least after twenty minutes in full lock.
Perhaps what Jenne saw is a type of oscillation that reported in alog 24665 ? Since the interferometer is locked, I am leaving it undistrubed. By the way OMC DCPDs are seemingly with low transimpedance. Noise in DARM below 50 Hz seems higher than the reference curve. I did not pay attention to the calibration at all.
The attached show an overnight trend (for 12 hours) of the relevant channels. As Jenne pointed out, the diffraction power indeed fluctuated (below 1 Hz) by 1 % or so all the time, but this is a known behavior (alog 24665) -- the 2nd loop changes the operating point of the first loop since the 2nd loop still has some gain at low frequencies even though it is digitally AC-coupled. As Gabriele pointed out, we think this fluctuation is added by the angular motion of the input mode cleaner (alog 24677). If we want to improve fluctuations in the diffraction power, the next attacking point would be the input mode cleaner.
Tagging SEI and CDS in this. If this problem (this aLOG and LHO aLOG 24665) has moved into the "we keep re-discovering it" category, perhaps the ISC team should request that CDS to work with SEI to push fixing HAM3's 0.6 [Hz] oscillation problems, and allow some non-Teusday time to do so.
Kyle, Gerardo We want to know if we can use the measured exhaust flow of CP3 to distinguish between relatively low LN2 pump levels versus relatively high LN2 levels and then exploit this as part of a makeshift control loop and/or justify increasing the time interval between manual over-fills. CP3 was last manually over-filled this past Wednesday afternoon. So this morning, at the ~42 hour mark, we used a plastic garbage bag to capture the exhaust until the bag was taught full (see image). We did this twice and got the following results: 220L/90sec = 145L/min, 220L/100sec = 130L/min. 6 hours later we manually over-filled CP3 at its normal 48 hour scheduled interval and then let the system settle down for 10 minutes or so before repeating the "bag" experiment. This time we got the following results: 220L/115 sec = 115 L/min. CONCLUSION: This result is surprising and not as expected. We may instead need to have Dave B. setup the exhaust PRESSURE signal (the not-blocked sensing line) so that we can trend it (I don't think we can currently for some reason). We already know that the exhaust pressure does vary >50% with slow over-filling so we might make a course control loop that fills CP3 based upon a preset time interval and then shuts off based upon when an exhaust pressure threshold is exceeded.
The discharge line pressure EPICS channel is HVE-MY:CP3_PT201 and it is indeed not in the DAQ at the moment, we'll add it during Tuesday maintenance.
In the mean time it is recorded every hour by the hourly autoburt backup system. Here are the commands to grep for this channel from yesterday for example (the space after the name is important)
david.barker@sysadmin0: cd /ligo/cds/lho/h0/burt/2016/02/12
david.barker@sysadmin0: grep "HVE-MY:CP3_PT201 " */h0vemy.snap
00:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
01:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
02:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
03:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 8.241758241758241e-01
04:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 7.020757020757020e-01
05:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
06:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 7.020757020757020e-01
07:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 6.715506715506715e-01
08:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 6.715506715506715e-01
09:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 7.936507936507936e-01
10:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 6.715506715506715e-01
11:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 6.715506715506715e-01
12:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.494505494505494e-01
13:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.494505494505494e-01
14:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.494505494505494e-01
15:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.494505494505494e-01
16:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 7.020757020757020e-01
17:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
18:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
19:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
20:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
21:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
22:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01
23:00/h0vemy.snap:RO HVE-MY:CP3_PT201 1 5.799755799755799e-01