The operator workstation can now open individual guardian screens from the GUARD_OVERVIEW screen. There was an alias for sitemap for the operator account that needed to be removed.
The following changes were done to the Beckhoff systems:
I reconfigured the network access for monit on all DAQ and front end computers to allow access from the control room workstations. This required restarting monit on all DAQ and front ends.
The AI v4 filter boards at both end stations were traded out for v6 boards. These AI chassis service the ESD.
End X - Removed Filter boards: S1101720-v4 & S1101716-v4. Replaced them with S1104975-v6 & S1300334-v6 Location SUS-XC1-U26
End Y - Removed Filter boards: S1101890-v4 & S1101884-v4. Replaced them with S1201006-v6 & S1203674-v6 Location SUS-YC1-U26
I'll be taking down h1broadcast0 to check the IPMI configuration which is either not working, or incorrect. This will interrupt DAQ data flow to the DMT systems while the machine is down (no other DAQ systems should be affected). I will post a follow up entry when the work is complete.
Goals:
I have been looking into recent data in order to:
- compare the performance form units to units
- correlate ISI stage 1 performance with optical lever motions
- compare winday times with regular times
Plots Desciption:
There are 4 figures attached. Each one display one hour of data.
- The first figure is for the windy afternoon on Saturday October 11th reported by Sheila
- The second figure is for Sunday October 19th at 3am
- The second figure is for Monday October 20th at 3am
- The second figure is for this morning, October 21st at 3am
In each of the four plots attached:
- the three plots in the top row show the ISI longitudinal motion (along the arm axis). Left plot is time series, middle is ASD, right is RMS.
- the three plots in the middle row show the optical lever pitch motion. (Left plot is time series, middle is ASD, right is RMS.)
- the three plots in the Bottom row show the optical lever pitch motion. (Left plot is time series, middle is ASD, right is RMS.)
Comments:
Windy day (first plot):
In the first plot (windy day), the ISI stage 1 RMS motion is about 10 times higher than usual, around 1000 nm/s RMS instead to about 100 nm/s in normal days. The Stage 1 motion is dominated by features below 50 mHz, most likely tilt as reported by Krishna.
The optical lever pitch motion is about a factor of 3 or 4 higher than usual (around 100 nRar instead of a few tens). Though the ISI is shaking at 50 mHz, the RMS of the optical lever is still dominated by the suspension mode at 0.5 Hz.
ETMY test mass pitch is moving twice as much as the others. This can clearly be correlated with the ISI motion at the suspension modes. We need to look into ETMY (something wrong in the blend? not enough loop gain? sensor noise?)
The test masses Yaw motion is dominated by features at the micro-seism, that are probably self inflicted (see comments for the third plot). ISI low performence on ETMY between 0.2 Hz and 1 Hz also affects the performance of the Yaw motion of this test mass.
Regular input motion, night time (second plot):
The ISI motion is pretty consistent from chamber to chamber, except for ETMY that should perform better above 0.25 Hz.
The Pitch motion of all test masses is dominated by the 0.5 Hz suspension mode.
The Yaw motion of all test masses is dominated by the micro-seism.
ISI off:
the third plot (data taken yesterday night) is quite interesting, as it looks like ITMY ISI was damped only.
In this configuration, ITMY pitch RMS motion is 10 times higher than the other units (near 1000 nrad RMS)
The Yaw motion is much lower (tens of nrad), but the three units isolated don't perform better than the unit damped. The unit damped is dominated by suspension modes features. The units isolated are dominated by micro-seism features.
Fourth plot (this morning at 3 amd);
all ISI are ON. They all perform similarly at low frequencies (below 0.5Hz), but optical levers RMS values are high and not consitent from units to units. Maybe some commissioning activities... to be checked.
Conclusion:
- we need to look into ETMY, and improve its performance at the suspension frequencies
- on windy times, and assuming alignment is the priority: for reducing the test masses pitch RMS value, we might want to try to improve the ISI performance at 0.5 Hz at the cost of further increasing the very low frequency motion. For yaw, we need to reduce the self inflicted amplification at the micro-seism.
- on regular times, we need to reduce the apparently self-inflicted yaw motion at the micro-seism
Recall that I haven't yet refined the optical lever calibration on the ETMs. My goal is to get to that during maintenance today (Tuesday, Oct 21 2014).
Fabrice pointed out at a telecon this morning that the performance of ETMY was not as good at 1hz as the other chambers. I looked at the filters I designed, and the gain at 1hz was pretty low (only about 10, see pg 2 on first attachment). This was caused by a too aggressive plant fit, which ate up too much gain at ~1hz, something I gotten smarter about since April when I designed these loops. I've touched up just this loop in matlab (second attachment), but I'll wait for a window to install it. Kiwamu said ths hasn't caused problems for them yet, but it would be nice for all platforms to have the "same" performance.
PSL weekly report of various parameters.
The relative power noise looks nominal. Better than the reference measurement below 10 Hz by about a factor of 2. About the same everywhere else. The frequency noise measurements (control and error signals) looks better than the reference measurements above ~100 Hz. About the same for frequencies below. The beam pointing looks nominal. Within requirements. The mode scan looks the same as ones conducted before. Higher order mode count, 53. Higher order mode power 4.8%. A little higher than previously. The ISS measurement was obtained with the PMC transmitting (according to the MEDM screen) 23.5 W and reflecting 2.2 W. The measurement is better than the reference measurement below 40 Hz and is the same above 40 Hz. The general level is slightly better than 2.0E-8 out of loop.
The cdsfs0 file server which contains /ligo crashed, reason unknown. Rebooted without power cycle, it came back without any problems.
Copied all the Cal & Sym filters from the HEPI L4CINF banks to the ISI-HEPI L4CINF filters. I did this with cut & paste in foton. Corrected the ISI-HEPI L4C2CART matrix. Safe.snap made. ISI back under Guardian control.
The HEPI matrices are still in need of correcting but that will be more invasive.
model restarts logged for Sun 19/Oct/2014
2014_10_19 03:33 h1fw1
2014_10_19 22:26 h1fw1
2014_10_19 23:28 h1fw1
model restarts logged for Mon 20/Oct/2014
2014_10_20 08:28 h1fw1
all restarts unexpected.
Alexa, Evan, Dan, Kiwamu,
We measured the power recycling gain for the 45 MHz sidebands in the PRMI sideband lock condition. It was estimated to be about 21 according to our OMC scan measurements.
(Method)
We measured the power of a 45 MHz RF sideband at the dark port using the OMC. We measured it in two different configurations -- (1) when PRMI was locked and (2) single bounce from ITMX. We did not measure that of the 9 MHz sidebands because we had a difficulty indentifying it from the carrier light.
Based on the measurement, we estimated the power recycling gain for the 45 MHz sideband. Assumptions we made are:
The measurement were made with 10 W of the light incident on IMC. Note that POP_RF18 was approximately 160 uW during the measurement.
(Result)
Pprmi_at_45MHz = 6.6 (in DCPD_SUM) -- this is the highest we could get by utilizing the dither ASC loops. Without a dither loop, DCSUM would have fluctuated between 3 and 5 presumably due to some alignment fluctuation.
Psinglebounce_at_45MHz = 0.288 (in DCPD_SUM)
(Recycling gain) = Pprmi / Psinglebounce * Tp *Tbs * Rbs / Ri / sin( 2 * pi * lsch * fm / c)^2
where fm is the modulation frequency and it was set to 5 * 9100230 Hz. This gave us a power recycling gain of 21.3.
Actually, I made a small mistake in the calculation as usual -- I did not need to include the ITM reflectivity in the equation because it cancels out between the two measurements.
This does not change the estimation of the recycling gain so much. It still should be Gp = 21.0.
Alexa, Evan, Kiwamu
We observed some new features which are related to the SRC mode hopping.
The SRCL error signal was calibrated in [nm] from a measurement of the open-loop transfer function last night. However the number does not seem right.
Last night, the UGF of the loop was estimated to be 27 Hz, which corresponded to an optical gain of about 5.0 x 1010 [cnts /meters] at the input of the LSC-SRCL filter. Therefore an offset of -800 cnts that we introduced at the SRCL input corresponds to a displacement of 16 [nm] ... which is actually already out of the linear range close to the edge of the linear range (because the linear range is 20-ish nm 40 nm in full width for SRCL). Something is not right.
I made an independent and more accurate calibration for SRCL. The result suggested that my previous calibration was off by roughly a factor of 2. The optical gain of SRCL should be 1.65 x 1011 [cnts/meters].
Therefore the 800 counts offset that we put yesterday should correspond to a displacement of 4.8 nm. We could sweep SRCL up to 6000 counts or 36 nm in one side of the fringe yesterday.
(Calibration method)
In the previous entry, I used the SRCL UGF in order to estimate the optical gain in counts/meters. This time, I used a sideband build-up signal which should give us a direct measure of the SRCL linewdith or liner range.
The plot below shows time series of some signals when we were changing the SRCL offset last night:
As shown in the plot, as we swept the offset of SRCL, the sideband power of SRC observed by AS_RF90 decreased/increased. When the sideband power becomes the half of the maximum, SRCL must be at the point where the linear range ends. Since we already know how big the linear range should be in terms of the SRCL displacement, we can calibrate the optical gain.
The plot below shows a x-y projection of AS_RF90 and SRCL_OFFSET from the same data as shown above:
By performing fitting, I was able to estimate the half-wdith at half-maximum (HWHM). I found the HWHM to be 3300 counts in terms of SRCL_OFFSET. According to galaxy (https://galaxy.ligo.caltech.edu/optics/), the transmissivity of SRM is T_{srm} = 37% for SRM-w14 and this gives a finesse of about 13. Therefore the HWHM should be (1064 nm ) / 4 / finesse = 20 nm.
Finally the calibration is calculated as (3300 counts) / (20 nm) = 1.65 x 1011 [counts/meters].
Kiwamu, Alexa
Today we looked at the demod phases for the REFLAIR RFPDs. The result was inconclusive. We excited PRM_M3_LOCK_L at 3.25Hz with an amplitude of 10000cts. We then examined the magnitude of transfer function of the I/Q error signals to this excitation at 1W, 5W, 7W, and 10W incident laser power. The data was taken with RELFAIR_A_RF45 at -135deg, RELFAIR_A_RF9 at 93deg, REFLAIR_B_RF27 at 107.8deg, and RELFAIR_B_135 at -30 deg following our nominal configuration. Attached show the magnitude vs power for the 1f and 3f signals, along with a linear fit. RELFAIR_A_RF9 behaves as expected with a linear reponse. During the measurement, REFLAIR_A_RF45 was not very coherent with the excitation, which explains why 45Q's response does not look very good. REFLAIR_B_RF27 looks fine as well; but RELFAIR_B_RF135 does not look right; it appears I might have missed a sign flip in the TF. But no major red flags...
Note: At some point after this measurement we had PRMI locked, and we improved the REFLAIR_A_RF45 I/Q ratio. The new phase of REFLAIR_A_RF45 is now 143 deg. This improved the DRMI calibrated noise spectrum of MICH around 10 Hz.
J. Warner, K. Venkateswara
Using Jeff K.'s model modifications, we were able to incorporate BRS tilt-subtraction to the GND_STS which was then output to STS_C channel (which was empty previously). This was then used for sensor correction by changing the cartesian conversion matrix. The tilt subtraction filter uses a simple acceleration to velocity converter, a gain matching and a high-pass filter.
Rich M. wrote for us a neat IIR filter to try out for the sensor correction filter bank. After the first attempt doubled the ground signal, we used this with a gain of -1.0 which worked well :)
We wanted to try out this sensor correction yesterday, but there was a lot of activity at ETMX which was disturbing the BRS as shown in the attached .png file. The good news is that immediately after the activity ceased, the BRS output settled down so the damper continues to work as expected.
This morning we had some time to try it out before EX activity resumed. The file NoSensCorr45mHzblnd.pdf shows the current ISI stage 1 performance with the 45 mHz blend filter. As I've mentioned before, notice the nice coherence with the GND_STS_X and ST1 T240_X between 0.1 to 0.6 Hz. The Y-axis is in counts (velocity).
For the uninitiated, sensor correction adds the ground sensor signal to the position sensor, thus it is effective only below the onboard inertial sensor blend frequency. In this case, since we wanted to do sensor correction in the 0.1-0.5 Hz band, we shifted the inertial sensor blend to 250 mHz. The two files NoSensCorr250mHzblnd.pdf and SensCorrOn250mHzblnd.pdf shows the Stage 1 motion without and with sensor correction respectively. Even with sensor correction on, this configuration seems to be a factor 2-5 worse in the 0.1 to 0.5 Hz band, but it is very interesting that the coherence with GND_STS_X is much smaller. And there is significant coherence with ST1_T240_Z at the microseismic peak. This might be related to what I mentioned before in 14426.
But to our surprise we were injecting noise below 0.1 Hz. I then rechecked our tilt-subtraction filter and it turns out we had one filter gain wrong. I've corrected it and I verified that the tilt-subtracted super-sensor was indeed smaller than the ordinary ground sensor. Unfortunately, EX activity resumed just then and we couldn't verify that the sensor correction was lowering the noise below 0.1 Hz. The ISIs were restored to the 45 mHz blends for commissioning activity.
Summary: Our first attempt at trying sensor correction with the ground 'super-sensor' was not terrible! :) We will do a more refined attempt tomorrow. We will also try doing sensor correction using HEPI which may give slightly different results.
The design of the neat IIR filter Rich wrote to which Krishna refers is detailed in SEI aLOG 594.